A rough-to-polished diamond ERP with costing, trading and AI insights for K N Gems, Surat
Looking for diamond ERP software in Surat that follows a parcel from rough purchase to certified polished sale? Yukti AI, a software company in Surat, built this system for K N Gems, a natural-diamond trader in the city that buys rough, has it cut and polished through job-work partners and sells polished stones with GIA, IGI, HRD, SGL or GSI certificates. The ERP tracks every rough lot through nine stages with its processing partner, expected and actual yield, weight loss and cost per polished carat; keeps one record per polished stone, linked back to the lot it came from; prices a stone with a seven-component costing calculator that shows margin on price and markup on cost side by side; runs purchase, quotations, sales orders, invoices, dispatch, CRM, follow-ups, a WhatsApp inbox with AI-drafted replies that staff approve, scheduled e-mail reports, accounting with a Tally mapping screen, receivables ageing and fifteen reports; and reads all of it in a rule-based AI layer with a daily briefing, ranked actions, a pricing advisor, anomaly flags and a business chat. A custom ERP from Yukti AI starts from ₹30,000; the final quote for a system of this breadth follows a free workflow audit.
K N Gems trades natural diamonds from Surat. Part of its stock is bought polished; part is bought rough and converted through cutting, polishing and grading partners before it reaches the stone book. The money in that business is decided twice: once when a rough parcel is bought on an expected yield, and again when a polished stone is priced against what it really cost. Yukti AI built one ERP around those two moments: a rough-processing module that records planned shapes, expected yield, job-work issues and returns, actual polished carats, variance in points, weight loss and cost per polished carat per lot; a Diamond Master where each polished stone carries its certificate, 4Cs, origin, purchase cost, labour and charges and a stored total cost; a costing calculator and pricing advisor; and the trading side — purchase, sales, documents, dispatch, CRM, WhatsApp, e-mail, accounting and payments — with five role dashboards and an AI layer that computes from the same records and shows its evidence. It is the working version built around a Surat diamond trader’s workflow, running on a sample dataset with records kept in the browser; the database, logins and live connections to Tally, WhatsApp, e-mail and stock platforms are set up at go-live.
Diamond ERP software in Surat – enquire
Yukti AI, a software company in Surat, builds diamond ERP, stock, costing, CRM, WhatsApp and AI automation for diamond manufacturers and traders in Surat, Mumbai and across India. The case study below shows one real build; this box tells you what you get and how to start.
- What you get: rough-lot processing with yield and cost per polished carat, a certificate-led stone book, a costing calculator, purchase, sales, quotations and invoices, dispatch, CRM and follow-ups, a WhatsApp inbox, e-mail reports, accounting with Tally mapping, receivables, fifteen reports, five role dashboards and a rule-based AI layer.
- Built around your desk: your stages, partners, cost components, approval limits, document formats and roles, not a generic template.
- From price: a custom ERP starts from ₹30,000; the final quote for a system of this breadth is given after a free 30-minute workflow audit.
- Where we are: 220, Leonard Square, Yogichowk, Varachha, Surat – 395006, Gujarat.
- Next step: send us one rough lot’s story — what you paid, who cut it, what came back and what it sold for — and we will show you what one system would have recorded.
Contents
- Quick answer and quick facts
- Screenshots
- The business, who this diamond ERP is for and how the business runs today
- Before vs after and task by task
- How a rough lot becomes polished stock: nine stages, job work, yield and variance
- How to calculate cost per polished carat from rough: a worked example
- Manufacturing and job work with partners
- Diamond costing per stone: seven components, margin vs markup, the ladder
- The certificate-led stone book and the plain-English filter
- Purchase, sales, quotations and invoices and dispatch
- CRM and follow-ups, WhatsApp inbox with AI-drafted replies and e-mail reports
- Accounting and the Tally mapping screen, payments and receivables
- Fifteen reports and five role dashboards, approvals and audit
- AI business intelligence: briefing, ranked actions, pricing advisor, anomalies, chat
- AI and automation for a diamond business: built here, built elsewhere, buildable, not built
- Custom diamond ERP vs ready-made Surat diamond software vs Excel and Tally
- Features checklist, benefits and buying checklist
- Solution architecture, the modules and how it was built
- What was delivered and what this page does not claim
- Technology and cost and timeline
- Diamond ERP in Surat: Varachha, Katargam, Mahidharpura and related diamond systems
- FAQ
What is a rough-to-polished diamond ERP, and who is it for?
A rough-to-polished diamond ERP is one system that records a rough parcel when it is bought, follows it through planning, job work, cutting, polishing and grading with the partner who did each step, receives the polished carats back, works out the yield that was actually achieved against the yield that was expected, divides the full cost of the lot by the polished carats to get a cost per polished carat, and then creates the individual polished stones in a stone book with that cost attached, so every later quotation, sale and profit figure is built on what the stone really cost. Yukti AI, a software company in Surat, built exactly this for K N Gems: rough processing in nine stages with job-work partners and their on-time record, a Diamond Master with one certificate-led record per stone, a seven-component costing calculator with margin and markup, purchase, sales, documents, dispatch, CRM, WhatsApp, e-mail, accounting with Tally mapping, payments, fifteen reports, five role dashboards and a rule-based AI layer. It is for Surat natural-diamond manufacturers-cum-traders who buy rough, get it worked through job-work partners and sell certified polished, and who today keep the kapan register, the stock sheet, the WhatsApp broadcasts and Tally apart. A custom ERP from Yukti AI starts from ₹30,000; a first working version is typically ready in about 7–10 days.
How does AI help a diamond trader in this ERP?
The AI in this build is rule-based: every briefing, ranked action, insight, price suggestion, anomaly flag and chat answer is computed from the stone book, the orders, the lots, the leads and the ledger by fixed rules, and each one shows the records it used. It tells the owner which stones have sat too long without an enquiry, which orders are under a margin threshold, which supplier’s rate jumped, which customer is due to buy again, which job work is late and what to buy next given enquiries against low stock. The WhatsApp assistant drafts replies that list real available stones and never answers a rate request on its own; a person approves before anything is sent. Nothing calls a language model; one can be added later behind the same screens to phrase the same facts. Records first, then automation, then AI, in that order.
Quick facts
| Client | K N Gems, a natural-diamond trader in Surat, Gujarat |
|---|---|
| Business model | Buys polished stones and rough parcels; rough is converted through cutting, polishing and grading job-work partners; certified polished stones are sold to wholesale and retail buyers |
| Stock model | Per stone, certificate-led: stock id, certificate number and lab (GIA, IGI, HRD, SGL, GSI or non-certified), ten shapes, carat, colour D–K, clarity FL–I1, cut, polish, symmetry, fluorescence, measurements, origin, cost, status, location and the rough lot it came from |
| Problem the system is built to answer | What did this rough lot really cost per polished carat; what was my profit this month; which stones are slow; who should I call today; which shape earns most; which supplier gives better margin; how much is outstanding; what should I buy next |
| Solution | A custom diamond ERP: dashboards by role, Diamond Master and live inventory, rough processing, manufacturing and job work, costing calculator, purchase, sales, invoices and quotations, dispatch, parties, CRM, follow-ups, WhatsApp, e-mail, accounting with Tally mapping, payments, reports, AI business intelligence, users and settings — 24 modules on one set of records |
| Core mechanics | Nine rough stages with history; actual yield, variance in points, weight loss and cost per polished carat per lot; stored total cost per stone; cost snapshot per sales line so margin never drifts; discount approval; permissions enforced at the route; every AI output computed from the records |
| Roles | 5: Owner / Super Admin, Sales Manager, Sales Executive, Accounts Manager, Inventory Manager, over 34 permissions |
| Reports | 15, each with a “what stands out” note |
| Automation | 8 WhatsApp rules, 7 scheduled e-mails, 12 recommendation types, 6 anomaly rules, about 15 chat intents |
| Stack | React 18, TypeScript, Vite 5, Tailwind CSS 3, Recharts, React Router 6 |
| Connections | Tally Prime, WhatsApp Business API, SMTP and an external stock platform are shown as connections to be made at go-live; none is live in this version |
| Status | The working version built around a Surat diamond trader’s workflow, sample data, records in the browser; usage and results are not claimed |
What K N Gems’ team works in
Screens from the diamond ERP built by Yukti AI for K N Gems. Figures, stone details, party names and staff names visible in the screenshots are from the sample dataset the working version runs on; they illustrate the screens and are not the company’s business numbers.









In a rough-to-polished business, the profit is decided before the stone exists
A polished diamond has a certificate, a shape, a carat weight, a colour and a clarity, and a price that the trade can argue about within a range. A rough parcel has none of that yet. When a Surat trader buys rough, the price is paid on a guess: this parcel should give around a third of its weight back as polished, in these shapes, at these grades. Then the parcel goes to a planning desk, to a cutting partner, to a polishing partner, to a laboratory, and comes back weeks later as a set of stones that may or may not match the guess. Two points of yield on a parcel is the difference between a good month and a flat one, and most of the cost that decides it — cutting per rough carat, polishing per polished carat, grading, laser inscription, freight, insurance — is spent while the parcel is out of the office.
K N Gems runs both sides of that trade from Surat. It buys polished stones from suppliers and sells them on, and it buys rough, has it worked through job-work partners and sells the output. Each side has its own paperwork: the kapan register for rough parcels and their returns, a stock sheet for polished stones, quotations and invoices as documents, WhatsApp threads with buyers asking “0.50 to 0.70 round F VS, what do you have?”, and Tally for the books. The profit on a rough lot is known only when somebody sits down with all of these at once, which is rarely.
Yukti AI’s job was to put the whole flow into one system: the rough lot and its stages, the job-work orders and what each partner returned, the polished stones that came out of it with their real cost, the costing of each stone for sale, the enquiry that matched it, the quotation, the order, the invoice, the courier, the receipt and the ledger entry, and an AI layer that reads all of that and tells the owner what deserves attention today.
A diamond ERP for a manufacturer-cum-trader has to carry cost forward: the rough purchase, the job-work charges and the yield that was actually achieved must become the cost of each polished stone, and that cost must travel into every quotation, sale and profit report without being retyped. Billing is the last step, not the system.
Who is this rough-to-polished diamond ERP for?
The system was designed around a Surat natural-diamond house that buys rough parcels and polished stones, gets the rough cut, polished and graded through outside job-work partners rather than on its own factory floor, holds certified polished stock across a vault, a strong room, a Mumbai office, a showroom display and partner premises, sells on credit to wholesalers, jewellers and exporters, and handles most buyer conversations on WhatsApp. If you are a diamond manufacturer-cum-trader in Varachha, Katargam, Mahidharpura or anywhere in Surat, or a Mumbai trader with Surat manufacturing partners, the problems on this page will sound familiar. It is also a worked answer to the search for diamond software in Surat that follows a parcel from rough to polished, not just a stock list with a price column.
It is a good fit if most of these are true for you:
- You buy rough on an expected yield and find out the real yield, and the real cost per polished carat, only when the accountant closes the year, if then.
- Your kapan register knows which parcel went to which partner, but not what it cost per carat when it came back or how that partner compares with the others on yield and delays.
- Polished stock is one Excel sheet with a price column, and nobody is sure whether the price was set from the stone’s real cost or from a rate someone remembered.
- Two salespeople can quote the same stone at different margins, and a discount on WhatsApp is given before anyone checks what it does to the profit on that order.
- Buyer enquiries arrive as trade shorthand on WhatsApp and someone scrolls the sheet by hand to answer, or broadcasts the whole list.
- Outstanding is a phone call to accounts, and Tally knows nothing about which stones or lots the money relates to.
If your business is mainly lot-wise polished trading without rough, see the lot-wise diamond trading ERP built for Arham Exports; if it is assorting in your own premises with parcels moving between custodians, see the diamond assorting ERP with stock custody built for a Surat diamond house; if you need a private buyer portal with holds and per-buyer pricing, see the diamond trading platform for a Surat export house. Yukti AI is in Surat and builds these systems for diamond businesses across Surat, Mumbai and India; see custom ERP software in Surat.
How a Surat diamond trader runs today: kapan registers, Excel stock, WhatsApp broadcasts and Tally
These points describe the gaps a system of this kind has to close, drawn from how Surat diamond houses commonly work and from the questions the build is designed to answer. They are not quotes from the client, and no measurement of the company’s previous way of working is claimed.
- The kapan register stops at the door. A rough parcel is entered when bought and ticked off when the polished comes back. What happened between — who planned it, which partner cut and which polished, how many carats went out and came back at each step, how late each return was — is in notebooks and memory.
- Yield is a feeling, not a number per lot. Everyone knows roughly what a parcel should give. Few can say, lot by lot, expected against actual in points, and which partner consistently returns under plan.
- Cost per polished carat is computed once, late. Rough price plus cutting plus polishing plus grading plus freight, divided by the carats that came back, is a simple sum that is rarely done per lot at the time, so the stones from a lot are priced on a general rate instead of their own cost.
- The stock sheet has a price column with no history. The asking price on a stone is whatever was last typed. There is no record of the cost components behind it, no margin shown, and no way to see what the price would be at a different margin without a calculator.
- Margin and markup get mixed up. A “20 percent” price means one thing when it is 20 percent on cost and another when it is 20 percent of the selling price, and both are used in the trade. Mixing them silently loses four points.
- WhatsApp is the sales floor and the archive. Enquiries come in trade shorthand, replies are typed from memory or the whole list is broadcast, and the requirement is lost when the chat scrolls away. Nobody can list the requirements that were asked for and not met.
- Discounts move faster than approvals. A rate is agreed in a chat before the owner sees what it does to the profit on that order.
- Money is a separate world. Tally has the invoice and the receipt, but not the stone, the lot, the partner or the salesperson, so profit by shape, by supplier, by customer or by lot is a year-end exercise.
- The owner’s questions have no screen. What was my profit this month, which stones are slow, who should I call today, what should I buy next: each one is a request to a staff member and a day’s wait.
How a rough-to-polished diamond business is run now
Before
- Rough parcels in a kapan register; stages in memory
- Yield checked by feel when the polished returns
- Cost per polished carat worked out at year end, if at all
- Polished stock in Excel with a typed price column
- Stone priced on a remembered rate; margin unknown
- Enquiries answered by scrolling the sheet or broadcasting it
- Discounts agreed on WhatsApp before anyone checks the profit
- Outstanding known by calling accounts; Tally knows no stones
- Owner’s questions answered a day later by staff
After (in the ERP)
- Rough lot with nine stages, partner, dates and a stage history
- Expected vs actual yield in points per lot and per partner
- Cost per polished carat recomputed the moment the carats are entered
- One record per stone with certificate, 4Cs, cost components and the lot it came from
- Costing calculator with margin and markup both shown and a price ladder
- Plain-English filter reads the enquiry and lists matching available stones
- Discount changes gross profit live; above the limit it waits for approval
- Receivables ageing, party ledger and a Tally mapping screen on the same records
- Daily briefing, ranked actions and a business chat computed from the records
How does the daily work change, task by task?
The table walks through the real jobs in a rough-to-polished diamond business and what each looks like in the ERP. The “before” column is a general description of the manual way of working, not a measurement, and no time savings are claimed.
| Task | Before (manual) | After (in the system) |
|---|---|---|
| Buying a rough parcel | Entry in the kapan register with weight and price | Rough lot with supplier, purchase cost, rough weight, expected yield %, expected polished carats, planned shapes and expected completion; a purchase voucher of kind Rough with pieces, carats and rate |
| Sending it for cutting | Note of the partner’s name and date | Job-work order for cutting with issued carats, rate per rough carat, due date and status; the lot moves to Job Work Assigned and Cutting with who and when |
| Receiving the polished back | Weigh, tick the register | Returned carats entered on the job-work order; actual yield, variance in points, weight loss and cost per polished carat recompute on the lot |
| Creating the polished stones | New rows typed into the stock sheet | Output stones created in the Diamond Master carrying the lot’s cost and a link back to the lot |
| Pricing a stone | Rate remembered or taken from a general list | Costing calculator: seven cost components, cost per carat, price at a chosen margin or markup, ladder at 5 to 30%, implied margin for a target price; pricing advisor against the reference grid |
| Answering a WhatsApp enquiry | Scroll the sheet or broadcast the list | The requirement is read from the message by rule, available stones are matched, a reply is drafted with up to five stones and rates, and a salesperson approves it |
| Quoting and reserving | Quotation in Word, stone marked by hand | Quotation document with validity and terms from settings; the lead shows which stones were shown; the order carries a cost snapshot per line |
| Giving a discount | Agreed in the chat | Discount % per line with gross profit updating live; a discount beyond the limit goes to the approval queue |
| Invoicing and dispatching | Invoice in Tally, courier slip by hand | Sales invoice from the order; shipment with courier, AWB, insured value, pieces, carats, destination and a checkpoint timeline |
| Collecting | Call accounts, then call the customer | Receipt settled against the order; receivables ageing in buckets; overdue customers oldest first; a payment-reminder rule that waits for accounts approval |
| Keeping the books | Tally, entered separately | Sales and purchase registers, party ledger, P&L and a Tally mapping screen for ledgers and voucher types, with sync logs; the live sync is set up at go-live |
| The owner’s morning | Ask three people | Daily briefing, top-10 actions and the 8 PM owner report, each with the records it was computed from |
Want a diamond ERP built around your own rough-to-polished flow?Send us one rough lot’s story — what you paid, who worked it, what came back, what it sold for — and we’ll show you what one system would have recorded. Free workflow audit, reply the same day.
Chat on WhatsAppHow a rough lot becomes polished stock: nine stages, job-work partners, yield and variance
Search for rough-to-polished diamond software and most of what you find is planning hardware: scanners and planning suites that look inside a rough stone and propose how to cut it. This ERP does not plan the stone. It does the part those tools leave to a register: it records the plan, follows the parcel through every hand that touches it, receives the polished back and works out what the lot really cost per polished carat. If a planning tool is used, its proposed shapes and expected yield are the inputs here.
The nine stages
Every rough lot in the system is at one of nine stages, and moves through them in order: Purchased, Stock Entry, Planning, Job Work Assigned, Cutting, Polishing, Grading, Polished Stock and Sold. Each move is written to a stage history with who made it and when, so a lot’s page shows its whole journey: bought on a date, entered into stock, planned with these shapes and this expected yield, assigned to this partner, cut, polished, graded, returned as these stones, and sold on these orders. The conversion-flow tab shows all open lots as a board by stage, so the inventory manager sees at a glance how many parcels are with partners and which have sat at one stage too long.
Sawing, laser cutting and bruting are not separate stages in this build. Laser marking exists as a job-work process type (alongside cutting, polishing, grading and re-polish), and the assortment of the output is implicit: the polished stones are created one by one in the Diamond Master. For a unit that runs its own factory floor and needs kapan and packet tracking through sawing, laser, bruting and polishing wheels with loss limits at each step, those stages can be added; see the features checklist for what is built and what is not.
What a rough lot records
- Purchase: lot number, supplier, purchase cost, rough weight in carats, purchase date, and the purchase voucher of kind Rough it belongs to.
- Plan: expected yield %, expected polished carats, planned shapes, the processing partner and the expected completion date.
- Charges: labour charges and other charges accumulated from the job-work orders and any direct entries.
- Return: actual polished carats, completion date, and the list of output stones created in the Diamond Master.
What the system computes per lot
- Actual yield = actual polished carats ÷ rough weight.
- Variance = actual yield − expected yield, in points, so a lot planned at 32 percent that returned 30 percent shows −2.
- Weight loss = rough weight − actual polished carats.
- Total cost = purchase cost + labour charges + other charges.
- Cost per polished carat = total cost ÷ actual polished carats.
- Output value = for each output stone, its sold price if sold, otherwise its asking price; summed.
- Profit = output value − total cost, and margin = profit ÷ output value.
When the inventory manager types the polished carats returned on a lot’s page, the yield, variance and cost per polished carat recompute on the spot. The same figures feed the Rough Processing Yield report and the yield-analysis tab.
Yield analysis: lots and partners
The yield-analysis tab charts expected against actual yield by lot and cost per polished carat by lot, so the owner can see which parcels under-delivered and whether a high cost per carat came from the rough price or from a poor return. Below that is a processing-partner performance table: for each partner, the number of lots, average planned yield, average actual yield, the variance between them and the on-time percentage of job-work orders returned by their due date. A partner who is cheap per carat but consistently two points under plan and late is visible as such, which is the conversation most traders have only from memory.
Output stones carry the lot’s id, so every polished stone’s page links back to the rough lot it came from, the partner that worked it and the cost per carat it inherited. That link is what makes profit from rough a per-lot figure instead of a year-end estimate.
How do you calculate cost per polished carat from rough? The yield calculator and its sensitivity table
The rough-processing module has a yield calculator for the moment before a parcel is bought. Its inputs are the rough weight, the rate per rough carat, a yield slider (32 percent by default), a cutting rate per rough carat, a polishing rate per polished carat, other charges for grading, laser and freight, and the expected polished selling rate per carat. The example below runs those formulas on illustrative inputs. Every figure here is computed from the build’s rules on made-up inputs; none is a client figure, a market rate or a partner’s price.
Inputs
| Rough weight | 100.00 carats |
|---|---|
| Rough rate | ₹40,000 per rough carat, so the rough purchase is ₹40,00,000 |
| Expected yield | 32 percent |
| Cutting | ₹1,500 per rough carat, so ₹1,50,000 for the parcel |
| Polishing | ₹6,000 per polished carat (so it rises and falls with the yield) |
| Other charges | ₹80,000 for grading, laser inscription and freight |
| Polished selling rate | ₹1,60,000 per polished carat |
The arithmetic at the expected yield
- Polished carats = rough weight × yield = 100.00 × 0.32 = 32.00 carats.
- Polishing charge = 32.00 × ₹6,000 = ₹1,92,000.
- Total cost = rough ₹40,00,000 + cutting ₹1,50,000 + polishing ₹1,92,000 + other ₹80,000 = ₹44,22,000.
- Cost per polished carat = ₹44,22,000 ÷ 32.00 = ₹1,38,188 (rounded).
- Revenue = 32.00 × ₹1,60,000 = ₹51,20,000.
- Profit = ₹51,20,000 − ₹44,22,000 = ₹6,98,000.
The cost build-up on screen shows each component as a share of the total: here the rough itself is about 90 percent, cutting 3.4 percent, polishing 4.3 percent and other charges 1.8 percent. That is why a trader’s eye goes to the rough price first; the calculator shows why it should go to the yield next.
The sensitivity table: what two or four points of yield do
The calculator then re-runs the same sums at −4, −2, 0, +2 and +4 points of yield and prints the swing. With the inputs above:
| Yield | Polished carats | Total cost | Cost per polished carat | Revenue | Profit |
|---|---|---|---|---|---|
| 28% (−4) | 28.00 | ₹43,98,000 | ₹1,57,071 | ₹44,80,000 | ₹82,000 |
| 30% (−2) | 30.00 | ₹44,10,000 | ₹1,47,000 | ₹48,00,000 | ₹3,90,000 |
| 32% (plan) | 32.00 | ₹44,22,000 | ₹1,38,188 | ₹51,20,000 | ₹6,98,000 |
| 34% (+2) | 34.00 | ₹44,34,000 | ₹1,30,412 | ₹54,40,000 | ₹10,06,000 |
| 36% (+4) | 36.00 | ₹44,46,000 | ₹1,23,500 | ₹57,60,000 | ₹13,14,000 |
The line the calculator prints under the table, for these inputs, would read: four points of yield on this parcel is a ₹6,16,000 swing in profit. Two points is ₹3,08,000. Notice that the total cost barely moves — only the polishing charge changes with the carats — while the cost per polished carat moves from ₹1,23,500 to ₹1,57,071 across the table. The parcel costs almost the same whatever comes back; what comes back decides everything. (Computed from the build’s rules on illustrative inputs; not a client figure.)
What the calculator is for, and what it is not
It is for the buying decision and for the conversation with a partner: at this rough price, this parcel needs at least this yield at this selling rate to be worth it, and a partner who returns two points under plan costs this much. It is not a planning tool; it does not look inside the stone. Expected yields are the trader’s own judgement or a planning tool’s output. Once the parcel is bought, the same formulas run on the real lot as the job-work orders return carats, so the estimate and the actual sit side by side. For the general method of figuring rough cost there are explainers online; this is the diamond-specific version a Surat desk uses, with the sensitivity that matters shown as money.
Manufacturing and job work: issuing to partners, returns, rates and on-time delivery
K N Gems does not run its own cutting floor; the work goes to partners. The manufacturing module is therefore a job-work ledger rather than a factory floor. Each job-work order belongs to one rough lot and one partner and records the process (Cutting, Polishing, Grading, Laser Marking or Re-polish), the carats issued, the carats returned, the rate per carat, the resulting charges, the issued, due and returned dates, a status of Issued, In Progress, Returned or Delayed, and a quality rating for the return.
Three things follow from that record. First, the lot’s labour and other charges accumulate from the orders, so the cost per polished carat is built from what was actually paid to whom, not from a budget. Second, an order past its due date without a return is Delayed, which puts the partner on the owner’s ranked-actions list (“chase delayed job work”) and into the on-time percentage in the partner performance table. Third, the partner performance view shows charges by partner and yield against plan in points, so the question “who should get the next parcel?” is answered from records.
Parties of type job worker sit in the same party master as customers and suppliers, with city, GSTIN, phone, WhatsApp, e-mail, credit terms and notes, so a partner in Katargam is a party like any other, and the ledger shows what has been paid to them. For a unit that wants karigar-level tracking with packets, loss limits per process and automatic jangad numbers on each outward movement, those are additions, listed honestly in the built / buildable / not built table.
Diamond costing per stone: seven cost components, margin vs markup, and the profit ladder
Search for diamond costing software and you get job-costing modules inside large ERPs, generic calculators, and a lot of pages about the 4Cs. None of them explains the sum a Surat desk actually does before quoting a stone. The costing calculator in the K N Gems ERP does that sum in the open, with every component visible and both ways of expressing the margin shown at once, because the trade uses both and mixing them costs money.
1. Seven cost components
Total cost of a stone = purchase price + labour + processing + certification + transportation + additional expenses + other charges. For a stone that came out of a rough lot, the purchase price is its share of the lot’s cost per polished carat; for a stone bought polished, it is the invoice price. The breakdown shows each component as a percentage of the total, so a stone whose certification and freight have quietly become a tenth of its cost stands out. Cost per carat = total cost ÷ carat weight.
2. Margin on price vs markup on cost
This is where most spreadsheets go wrong. Margin on price means the profit is a share of the selling price: price = cost ÷ (1 − margin). Markup on cost means the profit is a share of the cost: price = cost × (1 + markup). A 20 percent margin and a 20 percent markup are not the same price. The calculator lets the user choose either mode, and the result always reports both percentages, so a stone priced at a 20 percent margin is shown as a 25 percent markup, and a 20 percent markup is shown as a 16.7 percent margin. The conversion is margin = markup ÷ (1 + markup) and markup = margin ÷ (1 − margin).
3. The profit ladder
Below the chosen price the calculator prints a ladder at 5, 10, 15, 20, 25 and 30 percent, with the selling price and the profit at each step, so a salesperson negotiating on WhatsApp can see what the next rung costs without redoing the sum. It also runs backwards: type a target selling price and it reports the implied margin and markup, which is what you need when a buyer names a number.
4. Working with stock
A stone can be loaded into the calculator from the Diamond Master, so its stored cost components come across without retyping; calculations can be saved and compared side by side; and the result can be used for a quotation. On every stone the total cost (purchase + labour + other charges) is stored on the record rather than recomputed in each report, so the stock valuation, the sales profit and the AI pricing advisor all use the same number. When a stone is sold, the sales line snapshots the cost at that moment, so a later correction to the stone’s cost does not quietly change last month’s reported profit.
5. The reference grid and the pricing advisor
Settings hold an editable reference grid: a base rate per carat for nine carat bands, benchmarked on a G / VS1 / Excellent / GIA round, with factors for each of the ten shapes (a princess or a cushion at a lower factor than a round, for instance) and further adjustments for colour, clarity and cut. The rates shipped with the working version are illustrative placeholders, marked as such on screen and meant to be replaced with the house’s own. The pricing advisor uses this grid to suggest a price for a stone with its reasoning and to list the stones whose asking price is furthest from the reference, above or below. It is not a market feed; a Rapaport or price-list link is listed as an addition.
A worked stone costing: 1.01 carat, seven components, 20 percent margin vs 20 percent markup
The inputs are illustrative. Every figure is computed from the build’s formulas on made-up inputs; none is a client figure or a market price.
| Component | Amount | Share of cost |
|---|---|---|
| Purchase price (share of the lot’s cost per polished carat, or the polished invoice) | ₹3,00,000 | 90.9% |
| Labour | ₹10,000 | 3.0% |
| Processing | ₹8,000 | 2.4% |
| Certification | ₹6,000 | 1.8% |
| Transportation | ₹2,000 | 0.6% |
| Additional expenses | ₹3,000 | 0.9% |
| Other charges | ₹1,000 | 0.3% |
| Total cost | ₹3,30,000 | 100% |
| Cost per carat (1.01 ct) | ₹3,26,733 |
Twenty percent, two ways
- Margin on price, 20%: price = ₹3,30,000 ÷ 0.80 = ₹4,12,500; profit ₹82,500; the same price expressed as markup on cost is 25%.
- Markup on cost, 20%: price = ₹3,30,000 × 1.20 = ₹3,96,000; profit ₹66,000; the same price expressed as margin on price is 16.7%.
The gap between the two “20 percent” prices is ₹16,500 on one stone. On a parcel of thirty such stones quoted by two salespeople using the two conventions, that gap is the month’s profit. This is why the calculator always prints both.
The ladder, margin on price
| Margin on price | Equivalent markup on cost | Selling price | Profit |
|---|---|---|---|
| 5% | 5.3% | ₹3,47,368 | ₹17,368 |
| 10% | 11.1% | ₹3,66,667 | ₹36,667 |
| 15% | 17.6% | ₹3,88,235 | ₹58,235 |
| 20% | 25.0% | ₹4,12,500 | ₹82,500 |
| 25% | 33.3% | ₹4,40,000 | ₹1,10,000 |
| 30% | 42.9% | ₹4,71,429 | ₹1,41,429 |
The ladder, markup on cost
| Markup on cost | Equivalent margin on price | Selling price | Profit |
|---|---|---|---|
| 5% | 4.8% | ₹3,46,500 | ₹16,500 |
| 10% | 9.1% | ₹3,63,000 | ₹33,000 |
| 15% | 13.0% | ₹3,79,500 | ₹49,500 |
| 20% | 16.7% | ₹3,96,000 | ₹66,000 |
| 25% | 20.0% | ₹4,12,500 | ₹82,500 |
| 30% | 23.1% | ₹4,29,000 | ₹99,000 |
Backwards: a buyer names ₹4,00,000
Profit = ₹4,00,000 − ₹3,30,000 = ₹70,000. Implied margin on price = 70,000 ÷ 4,00,000 = 17.5%; implied markup on cost = 70,000 ÷ 3,30,000 = 21.2%. The salesperson sees both before replying, and if the house’s default margin is 20 percent, the order will show that line under target when it is entered. (Computed from the build’s rules on illustrative inputs; not a client figure.)
A certificate-led stone book: one record per polished stone, linked to its rough lot
Stock in this build is per stone, not per lot or parcel. That is the right model for a house that sells certified polished: a buyer asks for a certificate number, not a parcel. Each record in the Diamond Master carries the stock id, certificate number and laboratory (GIA, IGI, HRD, SGL, GSI or non-certified), the shape (round, princess, oval, cushion, emerald, pear, marquise, radiant, asscher, heart), carat, colour D to K, clarity FL to I1, cut, polish and symmetry grades, fluorescence, measurements, table and depth percentages, girdle, culet, origin (Botswana, South Africa, Russia, Canada, Angola, Namibia, Australia or India), the purchase party, date and price, labour and other charges, the stored total cost, the asking price, the status, the location, the rough lot it came from, and how many enquiries it has had.
Status is one of Available, Reserved, Sold, In Process, In Transit or Memo; location is one of the house’s own places — a main vault and a strong room in Surat, the Mumbai office, the showroom display, a job worker’s premises, with a customer on memo, or in transit. Memo here is a status and a location; there is no memo or jangad document, issue, return or ageing in this build, and the page says so plainly in what it does not claim. The diamond software pillar covers memo and holds.
Live inventory: four tabs
- Stock: the book itself, with the plain-English filter described below, column filters and bulk edit.
- Ageing and movement: days in stock per stone, enquiry counts, stones over the slow-moving threshold set in settings, and recent status and location changes.
- External stock platform: the card for a stock-software or listing-platform connection, shown as a connection to be made at go-live, with the publishing channels a house might use.
- Buyer preview: the same book with cost, labour, supplier and remarks stripped out and a published uplift percentage applied, which is the shape of a feed that would go to a listing platform or a private buyer portal. The portal itself is a separate build; see the diamond trading platform with buyer holds.
A CSV import screen exists for bringing an existing stock sheet in; in this version its parse and validation run on a sample file, and the mapping for a client’s own sheet is done at go-live. Natural diamonds only: the model has no lab-grown flag, so this page makes no claim about lab-grown handling.
How the plain-English filter reads “0.50 to 0.70 round F VS” — and why it is not AI
The search box above the stone book accepts the shorthand the trade actually types: 0.50 to 0.70 round F VS, oval above 1 carat D-F VVS, pr 1.5 ct GIA under 5 lakh, available cushion g to h si. It is a rule-based parser, not a language model, and this page calls it that. Under the hood it is a set of patterns:
- Shapes from an alias table: rnd and round, ov and oval, cush and cushion, em and emerald, pr and pear, mq and marquise, and so on.
- Carat as a range (“1 to 1.20”, “0.50-0.70 ct”), a bound (“above 2 carat”, “under 1 ct”), or a single value (“1.5 ct” becomes 1.4 to 1.6).
- Colour as single letters D to K when followed by “colour”, “col”, a comma, a space or the end of the phrase, or as ranges (“D-F”, “g to h”).
- Clarity as exact grades or as the groups VVS, VS and SI.
- Labs GIA, IGI, HRD, SGL and GSI.
- Price caps in lakh, crore, thousand or plain rupees: “under 5 lakh”, “below 100000”, “upto 2 cr”.
- Status words: available, in stock, sold, reserved.
Two design choices make it trustworthy. First, it echoes back what it understood — “Round / 0.5–0.7 ct / F / VS1, VS2” — so a salesperson sees the filter that was applied, not a black box. Second, it never invents a stone: the matching step applies the criteria to the book and returns real records or nothing. The same parser reads requirements out of WhatsApp messages for the draft assistant, out of lead notes for the CRM’s stock matching, and out of questions in the business chat. A language model could be put in front of it later to handle looser phrasing, but the matching would still run on the book, by rule, with the understood filter shown. Saying so is more useful to a buyer than calling a regular expression “AI search”.
Purchase: polished vouchers matched to stones, rough vouchers that become lots
A purchase voucher is of kind Polished or Rough. Polished lines are matched to stones in the Diamond Master, so the invoice price becomes the stone’s purchase price; rough lines are described by pieces, carats and rate and become rough lots. Each voucher records the supplier, the supplier’s invoice reference, additional charges, tax percentage and total, the amount paid, a payment status of Unpaid, Partial, Paid or Overdue, a due date derived from the supplier’s credit days, attached documents (certificate, invoice, packing list, photo), returns, and a purchase approval where the role requires it.
The purchase-analysis tab shows purchase value by month split polished against rough, spend by shape with the average rate per carat paid, and a supplier comparison of the rate paid against the margin the goods went on to earn, which is the figure behind the ranked action “renegotiate with the dearest supplier” and the anomaly rule for a supplier rate jump over the previous purchase.
Sales orders with a cost snapshot per line, quotations, proformas and invoices
A sales order moves through ten stages: Inquiry, Quotation, Negotiation, Confirmed, Invoiced, Paid, Dispatched, Completed, plus Cancelled and Returned. Each line carries the stone, its carat and rate, a discount percentage and a cost snapshot taken at the time of sale, so the gross profit per order is computed from what the stone cost on the day it was sold and does not drift if the stone’s record is corrected later. As a salesperson changes the discount, the order’s gross profit and margin update live; a discount beyond the permitted limit needs the discount-approval permission or goes to the approval queue. Orders show delivery status, courier and AWB once dispatched.
The sales-performance view shows billing and profit by month, each salesperson against a monthly target, customer-wise sales, and cancellations and returns, which feed the Staff Performance and Customer Profitability reports.
Six document kinds
The documents module produces a sales invoice, purchase invoice, quotation, proforma invoice, credit note and delivery challan, each with a status of Draft, Sent, Accepted, Paid, Expired or Cancelled. Document terms and quotation validity come from settings, and the company profile and bank details printed on them are maintained there too. A quotation reserves the stones it names on the lead and in the stone’s status, and the follow-up module raises a reminder before it expires. A memo or jangad document is not among the six kinds; that remains an addition.
Dispatch: shipments with courier, AWB, insured value and a checkpoint timeline
Polished stones move by secure courier or by hand, and the dispatch module records each shipment with its courier (the usual secure logistics names the trade uses, or hand delivery), the AWB, packed, dispatched, expected and delivered dates, a status from Packed to Delivered or Exception, the insured value, pieces, carats, destination and a checkpoint timeline. The shipment links to the order, so the stone’s status becomes In Transit and then Sold, the order moves to Dispatched, and the WhatsApp order-status rule can answer a buyer’s “where is it?” from the shipment record rather than from a phone call to the office.
CRM built for diamond enquiries: requirements in the buyer’s words, matched to live stock
Leads move through nine stages — New Inquiry, Contacted, Requirement Collected, Diamond Suggested, Quotation Sent, Negotiation, Follow-up, Won and Lost — and come from seven sources: WhatsApp, referral, exhibition, walk-in, phone, website and existing customer. A lead stores the requirement in the buyer’s own words and as structured preferences (shapes, carat range, colour, clarity, budget), a score and a temperature, and the stones already shown. The kanban board is drag-and-drop, and each lead’s card shows live stock matched to its requirement using the same rule-based parser as the stone book, so a salesperson opening a lead sees what can be offered today, not what was in stock when the lead was created. A won lead converts to a party; a lost lead records a reason. The analysis tab shows the funnel, conversion by source and lead handling by staff member.
Follow-ups with a suggested opening line
Follow-ups are of six kinds — New Inquiry, Quotation, Negotiation, Payment, Repeat Purchase and Inactive Customer — and are built from quotations about to expire, payments falling due, each customer’s buying cycle and lead stages, with a priority and a due date. The list is split into today, overdue and upcoming, each item carries a suggested opening line computed from the record (the stone that was quoted, the amount due, the months since the last order), and the workload view shows who on the team is carrying what. Reschedule, complete and history are a click each. For a CRM built around the WhatsApp inbox rather than inside an ERP, see the Surati jewellery CRM and WhatsApp inbox.
A WhatsApp inbox with AI-drafted replies that staff approve, and eight automation rules
Most diamond enquiries in Surat arrive on WhatsApp, so the ERP has an inbox for the sales team with filters for All, Mine, Unassigned, AI handled, Human and Pending reply. Each conversation shows the handler, the requirement extracted from the thread, and the stones already sent. The button that matters is Draft with AI. It reads the last inbound message; if the message asks about order status, dispatch or an AWB, it drafts the answer from the order and shipment records; otherwise it parses the requirement with the rule-based parser, searches available stock and drafts a reply listing up to five matching stones with the rate per carat, in a friendly Gujlish register. If nothing matches, it drafts a reply that promises to look and records the requirement as unmet, which feeds the “requested but unavailable” list in inventory intelligence. In every case a person approves or edits before anything is sent.
The eight rules
| Rule | What it does | Who approves |
|---|---|---|
| Stock availability reply | Reads the requirement and drafts a reply with matching available stones | Sales executive |
| No-match capture | Records the unmet requirement and drafts a “will source” reply | Sales executive |
| Quotation draft | Prepares a quotation from the stones the buyer chose | Sales manager |
| Order status reply | Answers status, dispatch and AWB questions from the records | Sales executive |
| Rate request handover | A request for a better rate is never answered by the assistant; it is handed to a person | Always a person |
| Outside working hours | Acknowledges and promises a reply in hours set in settings | Automatic, text from settings |
| Payment reminder | Drafts a reminder from the receivables ageing | Accounts manager |
| New stock alert | Drafts an alert to buyers whose stored preferences match newly entered stones | Sales manager |
Templates are kept by category, and the working hours and the auto-reply switch are in settings. The connection card shows the WhatsApp Business API as a connection to be made at go-live: it needs a Meta business account and approved message templates, and the conversations in this version are sample data. For the general service see WhatsApp automation, and for what Meta charges per conversation see WhatsApp Business API pricing in India.
Seven scheduled e-mails, including the 8 PM owner report
The e-mail module schedules seven automations: a Daily Owner Business Report at 8 PM with the day’s KPIs, tables and a commentary block computed from the records; a Weekly Performance Summary; a Monthly Financial Report; a Daily Outstanding Report for accounts; a Pending Follow-up Summary for sales; a Low Stock Alert by shape and carat band when live demand runs ahead of the threshold set in settings; and Important Business Notifications twice a day. Templates are assembled from KPI, table, text and commentary blocks, and a delivery log records what went out. The mail server connection is simulated in this version and configured at go-live; the content of every e-mail is already computed from the same records as the dashboards.
Accounting with a Tally mapping screen: what is built, what is simulated, what goes live
The accounting module has eight tabs. Overview shows six months of cash flow, the net cash position, receivables and payables ageing and a twelve-month summary. The sales register and purchase register list every invoice. Receivables and payables bucket what is due as not due, 1–30, 31–60, 61–90 and over 90 days. The party ledger lists every sale, purchase, receipt and payment for a party in date order with a running balance. Profit and loss takes gross profit from the cost snapshots on the sales lines; the overheads shown in the sample are modelled as fixed shares of revenue rather than entered expenses, so this is an operational P&L, not a statutory one, and the page does not describe an expense module because there is not one.
The Tally tab
The Tally integration tab has a ledger-mapping table (each ERP party and account against a Tally ledger name), a voucher-type mapping (sales invoice, purchase invoice, receipt, payment, credit note; the delivery challan is left unmapped in the sample to show what an unmapped type looks like), sync logs and an error list, with messages such as “no matching Tally ledger for the party”. The endpoint on screen is labelled as a simulated Tally connection: no live sync and no file import exist in this version. What is built is the mapping and the error handling that a live link needs; the connection itself — through Tally’s own integration interface or a file exchange — is scoped at go-live. The options and their trade-offs are explained in Tally integration with a custom ERP.
Receipts, payments, journals and receivables ageing
Payment entries are of kind Receipt, Payment, Journal or Contra, by NEFT, RTGS, cheque, cash or UPI, and are settled against a sales order or a purchase voucher, which updates the order’s stage to Paid or the voucher’s status to Paid or Partial. Each entry carries a Tally-synced flag for the day the link goes live. The receivables view lists overdue customers oldest first with the days overdue and the amount, and feeds the payment-reminder WhatsApp rule, the Daily Outstanding Report, the ranked action “collect overdue” and the anomaly rule that flags an invoice past due by more than a set number of days as critical. For reminders by voice call rather than text, see the outbound payment-reminder voice AI built for Sureja.
Fifteen reports, each with a “what stands out” note
Every report runs on the same records as the screens, is exportable and printable, and opens with a short computed note on what stands out in the period.
| Report | What it answers |
|---|---|
| Daily Sales | What was billed today, by salesperson and customer |
| Monthly Sales | Billing and gross profit by month against target |
| Shape-wise Sales | Which shapes sell and at what margin |
| Carat-wise Sales | Which carat bands sell and at what margin |
| Purchase Analysis | Polished vs rough spend by month, by shape, average rate per carat |
| Supplier Analysis | Rate paid vs margin earned, by supplier |
| Profit and Loss | Gross profit from cost snapshots, modelled overheads, net |
| Customer Profitability | Billing, profit and margin by customer |
| Payment Ageing | Receivables and payables by bucket |
| Stock Ageing | Days in stock, enquiry counts, slow movers |
| Inventory Valuation | Stock at cost and at asking price, by shape, band and location |
| Rough Processing Yield | Expected vs actual yield, variance, cost per polished carat, by lot and partner |
| Lead Conversion | Funnel and conversion by source and stage |
| Staff Performance | Sales, leads handled and follow-ups completed per person |
| WhatsApp Performance | Enquiries, response handling, conversions from the inbox |
Who sees what: five role dashboards, 34 permissions, approvals and the audit log
Five roles are defined: Owner / Super Admin, Sales Manager, Sales Executive, Accounts Manager and Inventory Manager, over 34 permissions. Each role lands on its own dashboard. The owner sees the whole business with the AI insights panel; the sales manager sees the team pipeline, open quotations and team performance; a sales executive sees their own leads, inbox and follow-ups; accounts sees receivables, payables, cash flow, overdue customers and the Tally position; the inventory manager sees the stone book, rough lots in progress and the oldest stock with its enquiry counts. Date presets of today, week, month, quarter, year or a custom range apply across them.
Permissions are enforced at the route, not only hidden in the sidebar: a sales executive who types the URL of the accounting screen sees a permission notice, not the page. Purchase cost, financials, approvals and user management are separate permissions, so a salesperson can see a stone’s asking price without seeing what it cost. An approval queue handles purchase, discount, price override and stock adjustment requests; an audit log records before and after values for price and stock changes; and a notification centre collects what each user needs to see. One honest note: in this version the roles are enforced in the application; at go-live they are enforced on the server and the database as well, which is standard for any live deployment.
Not sure which of your rough lots actually made money?We’ll show how rough cost, job-work charges, yield and the polished sale line up per lot in one system. Free workflow audit, reply the same day.
Chat on WhatsAppAI for a diamond trader: daily briefing, ranked actions, pricing advisor, anomalies and business chat — all rule-based
The AI module has eight tabs, and the honest description of all of them is the same: every output is computed from the records by deterministic rules and shows the records it used. Nothing in this version calls a language model or an outside service, and every AI output on screen carries a label saying it is a computed insight. A language model can be added later, behind the same screens, to phrase the same facts more naturally or to take looser questions; the evidence underneath would not change. We prefer to say that than to call a set of rules “machine learning”.
Daily briefing
A generated narrative for the day: billing so far against the month’s target, gross profit, stock value and how it moved, receivables and the biggest overdue, lots with partners and any that are late, enquiries in the inbox awaiting a reply, and the three things most worth doing. Each figure is a link to the screen it came from.
Top 10 actions, from twelve recommendation types
The ranked list is built from twelve rules, each with a priority, a reasoning line, an estimated impact in rupees where one can be computed, supporting data chips and a linked action:
- Move slow stones that are holding capital.
- Collect overdue receivables, oldest first.
- Buy a shape or carat band where enquiries run ahead of low stock.
- Contact customers whose buying cycle says they are due to order.
- Follow up quotations that have gone quiet.
- Review orders under a 10 percent margin.
- Renegotiate with the supplier whose rate is dearest for the margin earned.
- Reconsider asking prices that sit above the reference grid.
- Re-engage customers who have gone inactive.
- Reduce holding cost on stock older than six months.
- Propose standing supply plans for repeat accounts.
- Chase job work that is past its due date.
Profit, inventory and customer intelligence
Profit intelligence breaks gross profit by shape, carat band, supplier, customer, salesperson and month. Inventory intelligence compares demand with supply: fast and slow categories, the most requested shapes and carat ranges from leads and WhatsApp, categories that are overstocked and quiet, and requirements that were asked for but not available. Customer intelligence shows billing, lifetime value, a repeat-purchase likelihood from the buying cycle, and days since the last contact.
Pricing advisor
For any stone, a suggested price from the editable reference grid with its reasoning (band rate, shape factor, colour, clarity and cut adjustments), and a list of the stones whose asking price is furthest from the reference in either direction. It is advice against the house’s own grid, not a market valuation.
Risk and anomalies: six rules
An order margin under a threshold (critical under 3 percent); a supplier rate jump over that supplier’s previous purchase; an invoice past due (critical after 45 days); the same certificate number on two stock records; stones past 240 days with no enquiry; and billing more than a set percentage below last month. Thresholds are settings.
Business chat: about fifteen intents
The chat answers questions routed by keyword to about fifteen intents, each with a table, a chart or stone cards and a suggested follow-up question: profit this month, slow stock, who to contact today, best shape, supplier margins, outstanding, sales performance, what to buy next, stock value, fastest sellers, best carat band, top customers, open leads, unmet requirements and stone search in trade shorthand. Ask something outside those intents and it says so rather than guessing. That is the limit of a keyword router and the reason an LLM is the natural next step for the chat, with the same computed answers behind it. For how an agent connected to an ERP works in general, see ERP-connected AI agents.
What AI and automation can do for a Surat diamond trader today: built here, built in other Yukti AI systems, buildable, not built
Nobody ranks a diamond ERP page for “diamond ERP with AI” because the answers are either generic ERP vendors or research notes about rough planning. This table is the plain version for a Surat desk, labelled by what exists. Built here is in the K N Gems ERP; built elsewhere names the Yukti AI system it exists in, with a link; buildable means Yukti AI can build it on your data once the inputs exist; not built means it is not in this system and would be scoped separately.
| Opportunity | Label | What it is, and what it needs |
|---|---|---|
| Daily owner briefing and ranked actions | Built here | Narrative and top-10 list from twelve rules over the records; also the 8 PM owner e-mail |
| Cost per polished carat and yield sensitivity per lot | Built here | Recomputed as job-work returns are entered; partner on-time and variance |
| Plain-English stock filter in trade shorthand | Built here (rule-based) | Regex parser, echoes the understood filter, never invents a stone; also in the trading platform |
| WhatsApp replies drafted from real available stones, approved by staff | Built here | Draft assistant and eight rules; the Business API connection is made at go-live |
| Pricing advice against the house’s own reference grid | Built here | Grid in settings; stones furthest from reference listed |
| Anomaly flags: low margin, supplier jump, overdue, duplicate certificate, dead stock, billing drop | Built here | Six rules with thresholds in settings |
| Business chat over stock, sales, customers and receivables | Built here (keyword-routed) | About fifteen intents; an LLM can be added for open questions |
| Lot-wise buy / stop-buying advice with ageing buckets and broker scorecards | Built elsewhere | Arham Exports diamond trading ERP: AI inventory insights from velocity and lead time, says “not enough data” when it should |
| Copilot over custody, assorting and movements with an AI daily report | Built elsewhere | P. Hitesh & Co. assorting ERP: insight cards over the movement ledger and payroll |
| WhatsApp inbox as the CRM, with AI replies on a live number | Built elsewhere | Surati jewellery CRM: inbox, pipeline and follow-ups built around WhatsApp |
| Payment reminders by outbound voice call | Built elsewhere | Sureja voice AI: calls from the receivables list with a script and a summary back |
| Natural-language chat with an LLM phrasing the computed answers | Buildable | Same services behind the chat; the model phrases, the rules compute; a person still approves actions |
| Certificate lookup and PDF extraction into the stone record | Buildable | Extraction fills report number, shape, carat, colour, clarity; verified against the lab’s lookup; a person confirms |
| New-stock alerts to buyers by preference, sent on WhatsApp | Buildable | Rule exists; needs the Business API, approved templates and consent |
| Listing descriptions and stock feed text generated from attributes | Buildable | From stored fields; no invented grading |
| Invoice and packing-list OCR into purchase entries | Buildable | Extraction plus a review screen; nothing posts itself |
| Memo / jangad lifecycle with automatic numbers, returns, ageing and reminders | Not built | Only a status and a location exist; scoped separately; the pillar page describes it |
| Rapaport / price-list sync and RapNet-style stock feed | Not built | Buyer preview has the feed shape; no live feed or price list |
| Certificate API (GIA / IGI) live verification | Not built | Certificate fields exist on the stone; no lookup |
| Live Tally sync or file import | Not built | Mapping screen and error handling built; the connection is go-live scope |
| Native mobile app | Not built | The web app is responsive; an app is separate scope |
| Barcode / QR / RFID tallying | Not built | Would be added with a scanner-based stock count |
| Kapan / packet stages with sawing, laser, bruting and loss limits | Not built | For an own-factory unit; the nine stages here are a trader’s view |
| Rough planning optimisation and yield prediction | Not realistic in an ERP | That is scanner and planning-suite territory; the ERP records their output |
| Market price prediction | Not realistic yet | Needs licensed market data and proof; not promised |
The order of work is the same for every diamond house we build for: records first, so that yield, cost and margin are facts; automation next, so WhatsApp, e-mail and reminders act on those facts; AI last, so it reads clean data and shows its evidence. For the service pages behind the labels see AI automation for diamond and jewellery businesses in Surat, the AI automation agency in Surat and what an AI agent is; for budgets see AI automation cost in India.
Custom diamond ERP vs ready-made Surat diamond software vs Excel and Tally: which does a trader need?
There are three realistic choices for a diamond manufacturer-cum-trader in Surat. This is a general comparison, not a review of any product, and the ready-made column describes the category as its vendors present it.
| Question | Excel + Tally | Ready-made diamond software | Custom ERP (as built here) |
|---|---|---|---|
| Rough lot to polished stones | Kapan register; stones retyped | Usually included, often packet-based for factories | Nine stages, job-work orders, output stones linked to the lot |
| Cost per polished carat per lot | Year-end sum, if done | Varies; often cost per packet | Recomputed as returns are entered; sensitivity table before buying |
| Stone costing with margin and markup | A formula in one cell | Price list or Rapaport-based pricing | Seven components, both percentages shown, ladder, implied margin |
| Enquiry in trade shorthand | Scroll and filter by hand | Column filters | Rule-based plain-English filter that echoes what it understood |
| The phone | Rarely, sometimes memo alerts | Inbox, draft assistant, eight rules; API at go-live | |
| Memo / jangad | Paper book | Usually a headline feature | Not built here; scoped if needed |
| Books | Tally | Built in or via Tally | Operational registers, ledger, P&L and Tally mapping; live link scoped |
| Fits your roles and screens | No | Partly | Yes: five role dashboards, your stages, your documents |
| AI | No | Sometimes a BI dashboard or assistant | Rule-based briefing, actions, advisor, anomalies, chat on your own data |
| Cost model | Cheap, costly in mistakes | Licence or subscription per user | One-time build with support; from ₹30,000 |
| Time to start | Immediate | Implementation weeks | First working version in about 7–10 days |
When is it time to switch? When the questions in the challenge section — what did this lot cost per carat, which partner under-delivers, what margin did we really give on WhatsApp, who should we call today — take longer than a day to answer, Excel and Tally have stopped being enough. Ready-made diamond software fits a house whose workflow matches the package, particularly a factory that needs kapan and jangad out of the box. A custom build fits when the rough flow, the costing conventions, the approvals and the AI questions are your own.
Features of diamond manufacturing ERP software in Surat, and which ones this build covers
Surat diamond software vendors describe their products in the same vocabulary: rough inward, kapan and packets, planning, job work and karigar, jangad and memo, lab issue and return, Rapaport pricing, barcode and RFID, Tally, mobile. This checklist uses that vocabulary so you can compare any product, including this custom build, feature by feature. “Covered” means it exists in the K N Gems system; “Can be added” means it is not built and would be scoped with the client.
Rough, manufacturing and job work
| Feature | What it does | In this build |
|---|---|---|
| Rough inward and rough lots | Lot with supplier, cost, weight, expected yield, planned shapes, partner, dates | Covered |
| Nine-stage conversion flow with history | Purchased to Sold, who and when at each move | Covered |
| Job-work orders per partner | Cutting, polishing, grading, laser marking, re-polish; issued and returned carats, rate, due dates, status, quality | Covered |
| Yield, variance, weight loss, cost per polished carat | Per lot, recomputed on entry; yield analysis by lot and partner; on-time % | Covered |
| Yield calculator with sensitivity table | Before buying: cost build-up and profit at ±2 and ±4 points | Covered |
| Output stones linked to the lot | Each polished stone carries its rough lot id and inherited cost | Covered |
| Kapan / packet tracking with sawing, laser, bruting stages | Packet-level flow with process-wise loss limits for a factory floor | Can be added |
| Rough planning optimisation | Proposing the cut from a scan | Not an ERP feature; the ERP records the plan |
Stock, certificates and pricing
| Feature | What it does | In this build |
|---|---|---|
| Per-stone, certificate-led stock with 4Cs and more | Ten shapes, D–K, FL–I1, five labs, cut / polish / symmetry, fluorescence, measurements, origin | Covered |
| Status and multi-location | Six statuses; vaults, Mumbai office, showroom, job worker, with customer, in transit | Covered |
| Stock ageing and enquiry counts | Days in stock, slow-moving threshold, enquiries per stone | Covered |
| Plain-English filter | Rule-based parser for shape, carat, colour, clarity, lab, price, status | Covered |
| Costing calculator: seven components, margin vs markup, ladder | Per stone, loadable from stock, saved and compared | Covered |
| Reference grid and pricing advisor | Editable band rates and shape factors; suggestion with reasoning | Covered |
| Buyer preview feed | Book without cost, with uplift, in listing-feed shape | Covered (feed shape; no live feed) |
| Bulk edit and CSV import | Mass changes; import screen validated on a sample file | Covered (import mapping at go-live) |
| Certificate lookup, lab issue and return | GIA / IGI verification; stones out to a lab and back | Can be added |
| Rapaport / price-list sync; RapNet-style feed | Market reference and marketplace publishing | Can be added |
| Barcode / QR / RFID | Scanner-based stock count and packet tallying | Can be added |
| Lab-grown handling | Natural vs lab-grown flag and separate pricing | Can be added |
Sales, memo, CRM and WhatsApp
| Feature | What it does | In this build |
|---|---|---|
| Sales orders with cost snapshot and live margin | Ten stages, discount per line, discount approval | Covered |
| Quotation, proforma, sales and purchase invoice, credit note, delivery challan | Statuses, validity and terms from settings | Covered |
| Dispatch with courier, AWB, insurance and timeline | Shipment linked to the order | Covered |
| CRM with nine stages, stock matched to requirement | Kanban, score, stones shown, lost reasons, funnel | Covered |
| Follow-ups with suggested opening lines | Six kinds, today / overdue / upcoming, workload | Covered |
| WhatsApp inbox, draft assistant, eight rules | Replies from real stock, staff approval; API at go-live | Covered (connection at go-live) |
| Memo / jangad lifecycle | Automatic numbers, issue, return, ageing, reminders | Can be added (status and location only today) |
| Private buyer portal with holds and per-buyer prices | Buyers see their price and hold a stone for a time | Separate build; see the trading platform case study |
| E-invoice, e-way bill, GST returns | Statutory filings and documents | Can be added |
| Multi-currency export invoicing | USD invoices and FX | Can be added (INR only today) |
Accounts, reports, roles and integrations
| Feature | What it does | In this build |
|---|---|---|
| Registers, party ledger, receivables and payables ageing, P&L | Operational books on the same records | Covered |
| Payments: receipt, payment, journal, contra | Settled against orders and vouchers | Covered |
| Tally ledger and voucher mapping with sync logs | Mapping screen and error handling | Covered (live sync at go-live) |
| Fifteen reports with notes | Sales, purchase, profit, ageing, yield, leads, staff, WhatsApp | Covered |
| Five role dashboards, 34 permissions enforced at the route | Approvals, audit log, notifications | Covered |
| Scheduled e-mail reports | Seven automations including the 8 PM owner report | Covered (mail server at go-live) |
| AI briefing, actions, intelligence, advisor, anomalies, chat | Rule-based, evidence shown | Covered |
| Expense entry module | Entered overheads in the P&L | Can be added (sample overheads are modelled) |
| Payroll and attendance | Staff pay and hours | Can be added (built in the P. Hitesh ERP) |
| Native mobile app | iOS / Android | Can be added (web app is responsive) |
Benefits of a rough-to-polished diamond ERP for a Surat trader
Vendors list benefits as outcomes with numbers; this page lists them as mechanisms, because the mechanisms are what was built and the numbers would be invented.
Profit from rough is a per-lot fact
Because the lot carries its purchase cost and its job-work charges, and the returned carats are typed in, the cost per polished carat is a figure on the lot’s page the day the parcel comes back, and the output stones inherit it. Profit from rough is then output value minus lot cost, by lot, not an average over the year.
Partners are compared on records
Average planned yield, average actual yield, variance in points and on-time percentage per partner turn the annual argument about who to trust with the next parcel into a table.
Two kinds of twenty percent stop costing money
Every price shows margin on price and markup on cost together, so the house’s convention is applied the same way by every salesperson, and a buyer’s counter-offer is read back as an implied margin before anyone says yes.
Margin cannot drift after the fact
The cost snapshot on each sales line fixes the profit on that order at the moment of sale. Correcting a stone’s cost later changes the stock valuation, not last month’s sales profit.
Enquiries are answered from the book, in the buyer’s shorthand
The same parser reads the search box, the WhatsApp message and the lead note, lists real available stones and echoes what it understood. Unmet requirements are recorded, so what buyers wanted and could not get becomes a purchasing signal.
Discounts, costs and approvals are separated by permission
A salesperson sees asking prices but not purchase cost; a discount beyond the limit waits in the approval queue; price and stock changes are logged with before and after values; a blocked URL shows a notice rather than the page.
The owner’s questions have a screen
Profit this month, slow stones, who to call, what to buy, outstanding, late job work: a briefing, a ranked list, a chat and an 8 PM e-mail, each computed from the records with the evidence shown.
Nothing is retyped between stock, sales and books
The stone, the lot, the order, the invoice, the receipt and the ledger entry are one chain. The registers, ageing, party ledger and P&L are read from it, and the Tally mapping is ready for the day the link goes live.
Twelve questions to ask before buying diamond ERP software in Surat
- Show me one of my rough lots on screen. Purchase, partner, issued and returned carats, actual yield, variance, cost per polished carat. If the vendor cannot, the register stays in the notebook.
- Where does the cost of a polished stone come from? From the lot it came from and its real charges, or from a rate someone typed?
- Does the price screen show margin on price and markup on cost together? If it shows one number called “margin”, ask which.
- Is profit on a sale fixed at the time of sale? Or does editing a stone’s cost change last quarter’s reports?
- Can I type “0.50 to 0.70 round F VS” and see what the system understood? And does it ever show a stone that is not in the book?
- What does the WhatsApp feature actually send, and who approves it? Ask what happens when a buyer asks for a better rate.
- Memo and jangad: built, partial or not? This build says not; ask every vendor the same and get the answer in writing.
- Which connections are live today and which are at go-live? Tally, WhatsApp, e-mail, a listing platform, certificate lookup. Ask to see a sync log from a real connection.
- Who can see purchase cost? Ask for the permission list and whether it is enforced on the server.
- What does the AI compute, from what, and does it show its evidence? A rule that cites records beats a model that cannot be checked; ask which it is.
- What is the full price, and what is extra? Users, locations, integrations, migration, training, support after go-live.
- Who owns the data and how do I export it? Especially the stone book and the lot history, which are the house’s memory.
One thread, from the rough parcel to the ledger
The modules
Twenty-four modules on one shared set of records — the scope of a full AI-powered business management system shaped around a diamond trading desk with rough processing. See the general service on jewellery and diamond software and custom ERP software in Surat.
Dashboards
Five role dashboards with date presets and the AI insights panel for the owner.
Stock and processing
Diamond Master; live inventory with stock, ageing and movement, external stock platform and buyer preview tabs; rough diamond processing with conversion flow, lots, yield analysis and the yield calculator; manufacturing and job work; the costing calculator.
Trading
Purchase management with analysis; sales management with performance; invoices and quotations; dispatch; parties (customers, suppliers, both, job workers).
Customers and communication
CRM and leads with kanban and stock matching; follow-up management; WhatsApp automation with inbox, draft assistant, rules and templates; e-mail automation with seven schedules and a delivery log.
Money
Accounting with overview, registers, receivables, payables, party ledger, profit and loss and the Tally tab; payments and receivables.
Intelligence and administration
Reports and analytics (fifteen); AI business intelligence (eight tabs); users, roles and permissions with approvals, audit log and notifications; settings for the company profile, document defaults, WhatsApp working hours and auto-reply, default margin, slow-moving days, low-stock threshold, the pricing reference grid, integration cards and dataset controls.
How it was built
One domain model
Stone, rough lot, job-work order, purchase, order, document, lead, conversation, follow-up, payment, shipment, costing, recommendation, anomaly, approval and audit entry were defined once, in one place, before any screen.
Pure calculation services
Costing, analytics and the rule-based AI engine are plain functions over the records with no screen code in them, so the same yield, cost and margin figures appear on every page and in every e-mail.
Cost carried forward, never recomputed loosely
Total cost is stored on the stone; cost is snapshotted on the sales line; lot cost per carat is derived from recorded charges and returned carats.
Interlinked records
The working dataset was generated so that stones, lots, orders, payments, leads and conversations reference each other, which is why stock value, cost of goods, profit, receivables and ageing reconcile across screens.
Permissions at the route
Each role was given its dashboard and its allowed routes, with cost, financials and approvals as separate permissions, so the sales team and the accounts team see different systems.
Honest labels
Every AI output is tagged as a computed insight and every integration card as a connection to be made at go-live, so nobody reviewing the system mistakes sample behaviour for a live link.
What was delivered
What the system does — no invented numbers.
What this page does not claim
- No claim that K N Gems uses the system in daily operation, how many people use it, since when, or that it has gone live. There is no client quote and no outcome figure, because none exists.
- No claim of time saved, yield improved, margin protected, collections sped up or enquiries converted.
- The working version runs on sample data with records kept in the browser. Every stone, lot, party, user, lead, conversation, rupee figure, carat, yield percentage and price in it is fictional, including the rates in the reference grid, and none of it is K N Gems’ data.
- No live connections: Tally Prime, the WhatsApp Business API, the mail server and any external stock or listing platform are shown as connections to be made at go-live, and the sync logs on those screens are generated from the sample records. There is no certificate lookup, no Rapaport or price-list feed, no barcode or RFID, no mobile app, no e-invoice or e-way bill.
- Memo / jangad is not built. “Memo” is a stock status and “with customer (memo)” is a location; there is no memo or jangad document, issue, return, ageing or reminder.
- Sawing, laser cutting and bruting are not tracked stages; laser marking exists only as a job-work process type. There is no planning optimisation. Natural diamonds only; there is no lab-grown flag.
- The AI is rule-based. No language model, machine-learning model or forecast is used anywhere; the plain-English filter is a regular-expression parser.
- The profit and loss is operational: gross profit from cost snapshots and modelled overheads, not entered expenses and not a statutory book or a GST return.
- The line on the external stock card about the client’s existing stock software and books is the build’s own screen copy, not a statement from the client, and this page does not repeat it as a fact.
- Figures and names visible in the screenshots are from the sample dataset and illustrate the screens only.
The working version built by Yukti AI around a Surat diamond trader’s workflow, on sample data, with records in the browser. Anything not listed as built above is described as buildable, not as delivered.
Technology
The build uses React 18 with TypeScript, Vite 5, Tailwind CSS 3, Recharts for the charts, React Router 6 and Lucide icons. There is one domain model, a deterministic interlinked dataset, three pure services (analytics, the rule-based AI engine and costing) and a global store persisted in the browser, under 24 page modules. The costing service is small and self-contained; the analytics and AI services are the bulk of the logic, and none of them know about the screens. The interface uses a navy and emerald palette, Inter with a monospace face for figures, a light theme meant for daytime trading-floor use with a dark mode, comfortable and compact density, and a collapsible sidebar.
For a production deployment the records move to a server database, roles are enforced on the server, logins are added, the approved connections (Tally, WhatsApp Business API, mail, any stock platform) are made, and backups and an audit trail outside the browser are set up. That is the normal path from a working version to a live system, and it is the plan for any diamond house we build for.
What can be added next
- Memo / jangad lifecycle with automatic numbers on outward movement, returns, ageing and reminders.
- Certificate lookup and PDF extraction into the stone record; lab issue and return.
- Rapaport or price-list reference and a live stock feed to a listing platform or a private buyer portal.
- Live Tally sync or file exchange on the mapping already built.
- WhatsApp Business API connection with approved templates, so the eight rules send.
- Kapan and packet stages with sawing, laser and bruting, and process-wise loss limits, for a unit with its own floor.
- Barcode or QR labels with a scanner-based count; a native mobile app.
- An expense module, payroll and attendance, multi-currency export invoicing, e-invoice and e-way bill.
- A language model behind the business chat and the WhatsApp draft, phrasing the same computed facts.
How much does a diamond manufacturing ERP cost in Surat, and how long does it take?
Yukti AI does not publish a price list, because the price follows the scope. Two lines are approved and can be stated plainly: a custom ERP from Yukti AI starts from ₹30,000, and a first working version is typically ready in about 7–10 days. A diamond ERP of this breadth — rough processing, stone book, costing, trading, CRM, WhatsApp, e-mail, accounts and AI — is toward the larger end of what we build and takes a few weeks, and its quote depends on these factors:
- Which modules you need first (rough processing and the stone book only, or trading, CRM and accounts too).
- How many users, roles, offices and stock locations.
- Migration: how much stone, party and opening-balance data must come across from Excel and Tally, and how clean it is.
- Integrations: Tally, WhatsApp Business API, mail, a listing platform, certificate lookup.
- Whether memo / jangad, kapan stages, barcode or a mobile app are in the first scope.
- Hosting, security and support expectations after go-live.
The way to start is a free workflow audit: you show us one rough lot’s story and one polished sale, we show what one system would have recorded, and you get a quote. See custom ERP software cost in India for the general drivers and contact us to book the audit.
Diamond ERP software in Surat: Varachha, Katargam, Mahidharpura and the trader’s desk
Diamond ERP software in Surat is one system that records a rough parcel, follows it through job-work partners, costs the polished stones that come back, answers buyer enquiries from the book, raises the quotation, order, invoice and shipment and keeps the books and the owner’s questions on the same records — built by a team in the same city as your office. Yukti AI, at Leonard Square in Varachha, builds this kind of ERP for diamond houses across Surat, Mumbai and India.
Surat’s diamond trade is spread across its own geography: manufacturing units and job-work partners around Katargam and Varachha, trading offices and the daily market around Mahidharpura, and the newer office blocks on the city’s edge, with Mumbai’s bourse as the other end of many parcels’ journeys. The K N Gems system reflects that geography in its stock locations — Surat vaults, a job worker’s premises, a Mumbai office — and in its assumption that the people who cut and polish are partners, not employees. A house in any of these areas can have its ERP built by a team it can visit, which makes the workflow audit, the screen changes and the staff training a matter of an afternoon rather than a ticket.
What a Surat house should expect from a build like this: the rough flow and the costing conventions written the way the desk already thinks about them (kapan, yield in points, rate per carat, margin or markup as the house prefers); documents in the formats its buyers are used to; WhatsApp as the first channel, with a person approving every reply; Tally kept as the statutory book with a mapping that is ready for the link; and an AI layer that answers the owner’s questions from the house’s own records. For the broader Surat picture see diamond software in Surat, jewellery software in Surat and about Yukti AI.
See your own rough lot and stone costing in a working system firstWe build on your real stages, partners, cost components and documents, so you judge something working, not a slide deck. Free workflow audit, reply the same day.
Chat on WhatsAppRough-to-polished diamond ERP, costing and AI: questions Surat traders ask
Explore more
Diamond Software in Surat
Diamond ERP, stock, memo, holds and AI for Surat traders and manufacturers.
Case study: Arham Exports
Lot-wise diamond trading ERP with ageing, buy advice and broker scorecards.
Case study: P. Hitesh & Co.
Diamond assorting ERP with stock custody, movements, payroll and accounts.
Case study: diamond trading platform
Private buyer portal with per-buyer prices and timed holds for a Surat export house.
AI for Diamond & Jewellery
Where AI and WhatsApp automation help a Surat diamond or jewellery business.
Jewellery Software in Surat
ERP, CRM and automation for jewellery manufacturers and showrooms.
Case study: Surati jewellery CRM
CRM built around a WhatsApp inbox for a jewellery factory.
Custom ERP Software in Surat
ERP built around how your Surat business actually runs.
Tally Integration
Connect a custom ERP to Tally, or move off it in stages.
WhatsApp Automation
Stock replies, follow-ups and reminders from your own records.
ERP-Connected AI Agents
Agents that read your ERP, cite the evidence and let a person approve.
Custom ERP Cost in India
What drives the price of a custom ERP and how to budget.
All case studies
Every Yukti AI build across diamond, jewellery, manufacturing and trading.
Summary
Yukti AI, a software company in Surat, built a rough-to-polished diamond ERP for K N Gems, a natural-diamond trader in Surat that buys rough and polished, has rough cut, polished and graded through job-work partners and sells certified polished stones. Rough lots move through nine stages (Purchased, Stock Entry, Planning, Job Work Assigned, Cutting, Polishing, Grading, Polished Stock, Sold) with a stage history; job-work orders per partner record issued and returned carats, rates, charges, due dates and status; each lot computes actual yield, variance in points, weight loss, total cost, cost per polished carat (total cost ÷ actual polished carats), output value, profit and margin, and a yield calculator shows profit at ±2 and ±4 points of yield before a parcel is bought. Stock is per stone and certificate-led (GIA, IGI, HRD, SGL, GSI, non-certified; ten shapes; D–K; FL–I1; origin; six statuses; seven locations), each stone linked to its rough lot with a stored total cost. A costing calculator adds seven components, reports margin on price and markup on cost together, prints a 5–30 percent ladder and the implied margin for a target price, with a pricing advisor against an editable reference grid. A rule-based plain-English filter reads trade shorthand such as “0.50 to 0.70 round F VS”, echoes what it understood and never invents a stone; the same parser reads WhatsApp messages and lead notes. Purchase (polished and rough), sales orders with a cost snapshot per line and discount approval, six document kinds, dispatch with courier and AWB, a nine-stage CRM with stock matched to requirements, six kinds of follow-up with suggested opening lines, a WhatsApp inbox with AI-drafted replies that staff approve and eight rules (rate requests always go to a person), seven scheduled e-mails including an 8 PM owner report, accounting with registers, ageing, party ledger, P&L and a Tally mapping screen, payments, fifteen reports, five role dashboards over 34 permissions enforced at the route, approvals and an audit log complete it. The AI layer is deterministic: daily briefing, top-10 actions from twelve rules, profit, inventory and customer intelligence, pricing advisor, six anomaly rules and a business chat with about fifteen intents, each showing its evidence; no language model is used. Memo / jangad, sawing / laser / bruting stages, certificate lookup, Rapaport sync, barcode, a mobile app and live Tally, WhatsApp and mail connections are not built and are listed as additions. It is the working version built around a Surat diamond trader’s workflow, on sample data with records in the browser; usage and results are not claimed. A custom ERP from Yukti AI starts from ₹30,000, with a first working version in about 7–10 days.