AI-powered ERP & CRM for a PP woven-bag manufacturer
How Yukti AI unified stage-wise production, quality control, inventory, orders, finance and business intelligence into one custom system for Splenzo Polyfab Pvt. Ltd. & Inara — and added an AI layer that reads the plant's data and says what needs attention.
Splenzo Polyfab & Inara manufacture PP woven bags, FIBC bulk bags and laminated packaging fabric. Their production data lived across registers, spreadsheets and WhatsApp, so no one could see the whole plant at once. Yukti AI built a seven-module AI-powered business management system covering loom-to-dispatch production, quality against spec limits, inventory with price benchmarking, order-to-cash, and an AI Intelligence Suite — delivered in about ten working days on top of a working prototype.
Quick facts
| Client | Splenzo Polyfab Pvt. Ltd. & Inara |
|---|---|
| Location | New GIDC, Kabilpore, Navsari, Gujarat |
| Industry | Manufacturing — PP woven bags, FIBC / bulk bags, laminated packaging fabric |
| Business type | Manufacturer (loom → tape plant → lamination → bag conversion → dispatch) |
| Core challenge | Production, quality, inventory and finance data scattered across registers, spreadsheets and WhatsApp; problems seen a day or two late |
| Solution | Custom AI-powered business management system (ERP + CRM + AI), mobile-responsive, cloud-hosted |
| Modules | Command Centre, Production & Operations, Supply Chain & Inventory, Business Operations, Collaboration & Workflow, AI Intelligence Suite, Admin & Security — 7 in total |
| AI layer | AI Data Analyst, root-cause analysis, anomaly detection, daily AI briefing, analytics & comparisons |
| Integrations | WhatsApp Business Cloud API, LLM provider (client's own account); accounting integration available as a later phase |
| Delivery | ~10 working days on top of a pre-built working prototype |
| Access | Unlimited user logins, role-based access, full audit log |
| Result | One platform in place of scattered tracking; same-day visibility of loom idle time and quality drift against spec limits; a daily AI briefing of what needs attention |
What a woven-bag plant actually does all day
Splenzo Polyfab and its associate brand Inara make woven polypropylene packaging — the strong, printed sacks used for cement, chemicals, grains, fertiliser and industrial powders, plus FIBC "jumbo" bulk bags that hold a tonne or more. It looks like a simple product. The process behind it is not.
A batch of PP granules becomes a finished bag through a chain of stages, and each stage has its own machines, its own operators and its own way of going wrong:
- Tape plant (extrusion). Polypropylene is melted and drawn into flat tape. Tape denier and line speed decide the strength of everything downstream; a tape line that trips starves every loom after it.
- Weaving. Circular looms weave the tape into tubular fabric. Output is measured loom by loom, and an idle loom is money standing still — but "which looms were idle, and for how long" is usually only known approximately.
- Lamination. A layer of film is bonded to the fabric for moisture resistance. Coating GSM has to stay inside a band — too little fails the customer test, too much wastes material on every metre.
- Bag conversion. Fabric is printed, cut and stitched into bags. Here the plant checks bag weight against the specification, watches die variation, and logs rejects, segregation and waste.
- Dispatch. Finished bags are counted, matched to orders and shipped — and the payment against each order has to be chased and reconciled separately.
Every stage produces numbers that matter: output per shift, downtime reasons, reject percentages, coating weight, bag-weight tolerance, raw-material consumption, price paid versus the market. When those numbers live in separate registers and phone updates, nobody sees the whole plant at once — and the owner finds out about a bad batch or a stalled line when it is already expensive.
The problems, in the plant's own words
Splenzo was already a serious, well-run operation. The gaps were the normal ones that appear when a plant grows faster than its paperwork.
- "Production data is in five different places." Loom output, tape-line hours, lamination and bag-conversion progress were each tracked separately. "How did the plant do yesterday" meant collecting updates from several people.
- "We catch quality problems too late." Bag weight drifting toward the edge of tolerance, die variation creeping up, waste rising on a shift — visible only when someone looked, or when a customer complained.
- "We don't know if we bought raw material well." PP granule prices move constantly. Whether the last purchase was above or below market was a matter of memory.
- "Approvals and follow-ups live on WhatsApp." Task chasing and sign-offs sat in chat threads, so things slipped and there was no record of who approved what.
- "Every role wants a different view." The CEO wants company-wide health, the COO wants plant-wise bottlenecks, the plant head wants live shop-floor status. One shared spreadsheet can't be all three.
- "Reports are prepared by hand." Month-end and review numbers were assembled manually, so they were always a little late and a little inconsistent.
- "There are no recommendations, only data." Even where numbers existed, nothing told the team what to do about them.
How the day changed
Before
- Loom, tape, lamination and bag-conversion data in separate registers and sheets
- Quality drift noticed by chance or by complaint
- Raw-material price judged from memory
- Approvals and follow-ups chased on WhatsApp
- One generic spreadsheet shared by every role
- Reports assembled by hand at month-end
- Data, but no guidance on what to act on
After
- One platform: every stage's output and downtime in the same system
- Bag weight and die variation measured live against set spec limits
- RM price paid compared to the market benchmark on every purchase
- Structured approvals and task tracking with a full record
- Separate real-time dashboards for CEO, COO and plant head
- Analytics and period-over-period comparison generated automatically
- A daily AI briefing naming the three or four things to look at first
Why an off-the-shelf ERP wasn't enough
The obvious question — why not just buy a standard ERP? Splenzo looked at it the way most manufacturers do. The honest answer is that a generic ERP solves the accounting and stops roughly where the factory floor begins. It is built for a standard order-to-invoice flow, it has one dashboard for everyone, and it charges per user and per module every year. Loom-wise output, tape-line downtime reasons, bag-weight tolerance and waste-by-shift are simply not concepts it understands without expensive customisation.
A custom system inverts that. The plant's real process is the specification. Below is the comparison that made the case.
| What a woven-bag plant needs | Generic ERP / Tally | The system we built |
|---|---|---|
| Loom-wise output & idle-time tracking | Not a concept | Built in as the Loom Plan screen |
| Tape-line production & downtime reasons | Not covered | Tape Plant module with downtime logging |
| Bag weight / die variation vs spec limits | No quality module without add-ons | Quality Control measured against set limits |
| Segregation & waste tracking by shift | Manual | Logged and trended automatically |
| RM price vs market benchmark | Records purchase price only | RM Prices module with benchmark comparison |
| Role-based dashboards (CEO / COO / Plant) | One generic dashboard | Separate real-time dashboards per role |
| Plain-language questions about your data | Not available | AI Data Analyst answers with numbers |
| Licensing | Per user, per module, yearly | One-time build, unlimited logins |
The deciding factor was fit, not price. A system that makes the plant change how it works gets used for invoicing and ignored for everything else.
How the pieces fit together
Users work in the ERP and CRM. A workflow layer moves things automatically. An AI layer reads everything and advises. Integrations connect the outside world. It all writes to one central business database.
The seven modules
A single mobile-responsive, cloud-hosted system. Every module shares one set of data and one login.
1. Command Centre & role-based dashboards
Each person lands on their plant, not a generic home screen — a CEO dashboard of company-wide KPIs (revenue, production and quality health), a COO dashboard of plant-wise performance and live bottlenecks, a plant dashboard of shop-floor status, and a "My Department & My Tasks" view where each employee sees only their own numbers and pending work. Benefit: the daily status meeting becomes everyone opening the system to the answer.
2. Production & Operations
End-to-end tracking from raw tape to finished bag — Loom Plan (weaving schedule and loom-wise output), Tape Plant (PP tape-line production and downtime with reasons), Lamination, Bag Conversion (cutting and stitching through to finished bags), and Quality Control (bag-weight monitoring, die-variation tracking, issue logging, segregation and waste, all measured against set specification limits). Benefit: the plant head can point to the exact stage that is behind, and why, without walking the floor.
3. Supply Chain & Inventory
RM Stock (real-time raw-material levels), RM Prices (price tracking with market-benchmark comparison), and FG Stock (finished-goods stock and dispatch readiness against open orders). Benefit: purchasing is judged against the current market, and dispatch knows what is actually ready to ship.
4. Business Operations — order to cash
Orders (booking and status, linked to the production that fulfils them), Dispatch (planning and tracking against orders), and Finance (payment tracking linked to every order). Benefit: "which orders are overdue for payment" is one screen, not a separate reconciliation.
5. Collaboration & Workflow
Actions (cross-department task tracking), Approvals (structured workflows with a record of who approved what), Incentive (performance tracking tied to real output), Notifications (alerts routed to the right person), and Report Compliance (compliance reporting and tracking). Benefit: the chase moves out of chat threads into a system that remembers.
6. AI Intelligence Suite
AI Data Analyst, Root-Cause Analysis, Anomaly Detection, Daily AI Briefing, and Analytics & Comparisons — covered in full in the next section.
7. Admin, Access & Security
Unlimited user logins at no per-user cost, role-based access control, master-data management (products, customers, defect types), configurable KPI thresholds so the dashboards reflect Splenzo's own standards, and a full audit log of who did what, and when.
What makes it different from "we added AI"
"AI in an ERP" usually means a chatbot that answers help questions. This is not that. The AI in the Splenzo system reads the plant's own production, quality and pricing data and turns it into decisions the team would otherwise take days to reach — or miss.
Ask your plant a question in plain language
Anyone with access can type a question the way they would ask a colleague, and get an answer backed by the actual figures:
- “Why is bag weight off spec this month?”
- “Which looms are causing the most idle time?”
- “Has our PP granule cost gone above the market rate in the last 60 days?”
- “Which shift has the highest waste, and by how much?”
- “What changed in the plant yesterday that I should know about?”
Every morning, the AI briefing answers
- Which stage is behind schedule, and what is causing it
- Which quality parameters are drifting toward their limits
- Which raw material is close to running out
- Which orders are overdue for payment
- What management should focus on today
Root-cause analysis
When a quality miss or output gap appears, the system links it to its likely cause automatically — a loom that was down during the shift in question, a lamination batch outside the GSM band, a tape lot with a different denier. Instead of "bag weight was off", the team gets "bag weight was off on these batches, which trace back to this cause".
Anomaly detection
The AI watches for patterns that are unusual for Splenzo specifically — a reject rate stepping up, an RM price out of line with its own recent history, a loom drifting below its normal output — and flags them before they become a complaint or a month-end surprise.
AI is woven into daily decisions — root-cause, anomaly alerts, a morning briefing — not bolted on as a separate chat window. The team never has to "go and use the AI"; it comes to them. The AI runs on an AI-provider account in Splenzo & Inara's own name, so usage is billed to them directly with no markup.
What runs without anyone pressing a button
Low raw material → purchase recommendation
Quality parameter out of tolerance → alert & root-cause
Order milestone → customer update
What each person gets when they log in
CEO / Owner
Company-wide KPIs, quality and production health, and the AI briefing — the state of the business in one view.
COO
Plant-wise operational performance, live bottleneck visibility, and cross-plant comparison.
Plant head
Live shop-floor status by stage, loom plan, downtime, and the day's quality numbers against spec limits.
Production / QC
Data entry for their stage, issue logging, segregation and waste, and their own task list.
Purchase
RM stock levels, price-vs-benchmark comparison, and purchase recommendations awaiting action.
Accounts
Order-linked payments, outstanding by customer, and finance tracking against every order.
What changed
Described honestly — what the system does, not invented percentages.
Technology stack, and why
- Next.js (App Router, TypeScript) — a fast, mobile-responsive web app that works on any phone, tablet or desktop browser with no app to install on the shop floor.
- Supabase (PostgreSQL, Auth, Storage, Row-Level Security) — a proper relational database for production and quality data, with role-based access enforced at the data layer.
- Tailwind CSS + shadcn/ui — a clean, consistent interface non-technical plant staff can use without training overhead.
- Vercel — secure cloud hosting with daily automated backups.
- WhatsApp Business Cloud API — for automated order updates and payment reminders.
- LLM provider API (client's own account) — powers the AI Data Analyst, root-cause analysis, anomaly detection and daily briefing.
Security & data architecture
- Role-based access control — each user sees only what concerns their role, enforced in the database, not just hidden in the UI
- Full audit log — a trail of who created, edited or approved what, and when
- Daily automated backups and secure cloud hosting with SSL
- Client owns all data and can export it at any time; Yukti AI retains the source-code framework (handover negotiable separately)
- AI usage runs on the client's own provider account — no data resold, no markup
From plant visit to handover
Yukti AI does not start from a blank page. A working prototype that already reflected Splenzo's real loom-to-dispatch flow was built first.
Business discovery
We mapped the real flow — tape plant, looms, lamination, bag conversion, dispatch — and the numbers and quality parameters each stage already produces.
Working prototype
A functioning system reflecting that flow, so the conversation was "move this, add that" against something real, not a document arguing over screens that don't exist yet.
Requirement freeze (Day 1)
A joint session where every requirement was discussed, understood and approved before development continued.
Development (Day 2–8)
All seven modules built against the approved requirements, with storage and the AI-provider account connected and the whole system tested.
Deploy & train (Day 9–10)
Deployment on a live server, then training across CEO, COO, plant and department users, with a user manual and walkthrough.
12 months of support
Year 1 support — hosting, backup, security updates and bug fixes — is included. Minor changes after go-live continue as part of that support.
Questions manufacturers ask
Explore more
Manufacturing ERP Software
Stage-wise production, quality against spec limits, inventory, orders and an AI layer — built around your plant.
Custom ERP Software
Why growing Indian businesses move off generic ERP and Tally to a system built around their process.
AI ERP
What an ERP does when an AI layer reads the data — questions, root-cause, anomaly alerts, a daily briefing.
AI Business Management System
Operations, inventory, sales, accounts, tasks and AI in one custom platform for any industry.
Case study: Maven Lift
A business management system for an elevator manufacturer — digital site survey straight into an auto-quotation.
All case studies
Jewellery, manufacturing, trading and services — every Yukti AI build in one place.
Summary
Yukti AI built a customised AI-powered ERP and CRM system for Splenzo Polyfab Pvt. Ltd. & Inara, a PP woven-bag and FIBC manufacturer in Navsari, Gujarat. The system unifies stage-wise production tracking (tape plant, looms, lamination, bag conversion), quality control against specification limits, raw-material and finished-goods inventory with price benchmarking, order-to-cash operations, and collaboration workflows — across seven modules in one platform. An AI Intelligence Suite reads the plant's data to answer plain-language questions, run root-cause analysis, detect anomalies and generate a daily briefing. It was delivered in about ten working days on top of a working prototype, with unlimited role-based logins, a full audit log and twelve months of support. Yukti AI builds comparable manufacturing ERP systems for packaging, plastics, textile, engineering and fabrication businesses across Surat, Gujarat and India.