When every order is a new product — an AI quotation engine for a custom box maker
How Yukti AI built The Box Factory an 18-module AI-first manufacturing CRM and ERP, built around the reality that every enquiry is a bespoke box — with a flagship AI quotation engine that turns a free-text requirement into a real bill of materials, cost breakdown and quantity-tier pricing.
The Box Factory manufactures decorative and corporate-gift boxes in MDF and rigid board — every job a different size, finish, print and quantity. Quoting meant costing each one by hand. Yukti AI built a single platform where the AI quotation engine produces a real BOM and cost analysis from a plain-language brief (with a margin slider, tier pricing, version history and an approval workflow), and the other 17 modules carry that job through CRM, artwork approval, sample sign-off, production planning, inventory, dispatch, accounting and reports.
Quick facts
| Client | The Box Factory |
|---|---|
| Industry | Manufacturing — decorative and corporate-gift boxes (MDF, rigid board) |
| Business type | Made-to-order manufacturer — every job a bespoke product |
| Core challenge | Every enquiry costed by hand; jobs tracked in the head through design, artwork, sample, production and dispatch |
| Solution | An 18-module AI-first manufacturing CRM + ERP in one platform |
| Flagship module | AI Quotation & Estimation Engine — free-text requirement → real BOM + cost + tier pricing |
| Other modules | AI Dashboard, CRM, Unified Inbox, AI Box Builder, Product Templates, Order Management, Production Planning, Inventory, Purchase Management, Dispatch, Sample Management, Artwork Management, Accounting, Documents, HR, Reports, Settings / Company Profile |
| Stack | Next.js 16 (App Router, TypeScript), Tailwind CSS, Radix UI, Zustand, recharts; light & dark themes |
| Status | All 18 modules built, browser-verified, production-build clean |
| Result | Quoting becomes structured estimation with a margin and tier table; every job tracked as a project from enquiry to payment |
A custom-box maker sells estimates, not products
The Box Factory makes premium packaging — rigid boxes with magnetic closures, decorative MDF boxes, corporate gift sets. There is no catalogue to sell from. A brand comes with a brief: this size, this board, this lamination, this foil, this insert, this quantity. Every one is a new product that has never been made before.
So the first real work on every enquiry is a costing: how much board, how much lamination, how much foil, how many printing passes, how much cutting and pasting labour, what wastage, what margin — and how all of that changes at 100 units versus 500 versus 2,000. Get it wrong high and you lose the job; wrong low and you lose money on it. And once a job is won, it runs a gauntlet — design, artwork approval, a physical sample, sign-off, production stages, dispatch — where a missed step means a reprint or an angry client.
For a made-to-order manufacturer, the quotation is the product until the job is won. Software that treats quoting as picking from a price list is solving the wrong problem.
What The Box Factory was doing by hand
- "Every quote is costed from scratch on a calculator or a spreadsheet." Slow, and inconsistent between whoever prepares it.
- "Tier pricing is guesswork." The price at 500 units versus 2,000 was estimated each time, not modelled.
- "When a customer changes the brief, we re-cost the whole thing." No version history, no way to see what changed the price.
- "Job status lives in people's heads." Where a job is between enquiry, artwork, sample and production is a question you ask around.
- "Artwork and samples get skipped." Occasionally a job went to bulk before the artwork was actually approved.
- "Quoted versus actual is never checked." Whether a finished job made the margin it was quoted at was rarely known.
How quoting and tracking changed
Before
- Each quote costed by hand, differently each time
- Tier pricing estimated, not modelled
- Brief changes mean re-costing from zero
- Job status known by asking around
- Artwork / sample sign-off sometimes skipped
- Quoted-vs-actual margin rarely checked
After
- Free-text brief → real BOM + cost breakdown in the AI engine
- Quantity-tier pricing table generated with the quote
- Version history — re-run the estimate on the edited brief
- Every job a project on a production plan
- Artwork & sample modules gate bulk production
- Reports compare quoted value to actual
Why not a standard ERP or a printing-industry package?
Standard ERP assumes you sell from a catalogue. Printing-industry packages assume offset print runs and standard job types. A premium custom-box maker is neither — it is a design-and-estimate business where the quote has to be built from raw materials every time and the job runs through an artwork-and-sample gate before bulk. The comparison that made the case for a custom build:
| What a custom-box maker needs | Standard ERP / print package | The system we built |
|---|---|---|
| Cost a bespoke job from a free-text brief | Pick from a price list | AI quotation engine — real BOM & cost |
| Quantity-tier pricing on every quote | Manual | Generated with the quote |
| Version history when the brief changes | New quote from scratch | Re-run the estimate, see the delta |
| Artwork approval gate before bulk | Not modelled | Artwork module blocks bulk |
| Physical sample sign-off gate | Not modelled | Sample module blocks bulk |
| Quoted value vs actual per job | Rarely | A dedicated report tab |
What runs without anyone doing the math
Free-text brief → priced quote
Quote approved → job on the floor
Brief change → transparent re-price
One job, enquiry to payment
The 18 modules
1. AI Quotation & Estimation Engine — the flagship
Type or paste the requirement in plain language — "200 rigid boxes, magnetic closure, 12×9×4 in, matte lamination, gold foil logo, foam insert". The engine produces a real bill of materials and cost analysis: board, lamination, foil, printing, cutting, pasting, labour and wastage, not a stub. On top of that: a margin slider, a quantity-tier pricing table, version history that re-runs the analysis when you edit the brief, an approval workflow (draft → sent → approved / rejected → converted), and one-click PDF. Benefit: a consistent, defensible quote in minutes, and a clear record of how a changed brief changed the price.
2. AI Box Builder
A guided way to specify a box — type, dimensions, closure, finish, print, insert — that feeds the quotation engine with a clean structured brief. Benefit: even a quick enquiry becomes a properly specified job.
3. AI Dashboard
The state of the business on one screen — open enquiries, quotes out, jobs in production, dispatch due, payments pending — with the numbers that matter surfaced. Benefit: the morning check is one screen.
4–6. CRM, Unified Inbox, Product Templates
Customer records and history; one inbox for enquiries across channels; and reusable templates for the box types the factory makes most often, so a repeat job starts from a known spec. Benefit: the front of the funnel is organised and repeat business is fast.
7. Order Management
An approved quote becomes an order with all its spec, quantity, price and delivery date, linked back to the quote it came from. Benefit: the order and the quote never disagree.
8. Production Planning
Each order scheduled through its production stages — cutting, printing, lamination, foiling, pasting, assembly, QC — with a calendar view of the floor. Benefit: capacity and bottlenecks are visible before they bite.
9–11. Inventory, Purchase Management, Dispatch
Material stock checked against the job's BOM; purchase raised where the plant is short; dispatch on completion. Benefit: a job does not stall halfway because a board size ran out.
12. Sample Management
Sample requests tracked from creation to dispatch to customer feedback — and a job cannot move to bulk production until the sample is signed off. Benefit: no bulk run on an unapproved sample.
13. Artwork Management
Every design version and its approval status against a job. Benefit: the plant prints the approved artwork, and there is a record of who approved it.
14–17. Accounting, Documents, HR, Reports
Payment tracking and basic accounting; a document vault per job (POs, approvals, dispatch docs); HR; and a reports section with eight tabs of charts — including quoted value versus actual per job for real margin visibility. Benefit: the business closes the loop on whether jobs made money.
18. Settings & Company Profile
Configuration, master data, users and the company profile that appears on quotes and documents.
What "AI-first" means here
The Box Factory system is AI-first in a specific, concrete way — the AI is doing the hardest recurring task in the business, not answering FAQs.
- Estimation — the quotation engine reads an unstructured brief and returns a structured bill of materials and cost. This is the AI's main job, and it does real math, not a template fill.
- Specification — the box builder turns a rough enquiry into a clean spec the engine can cost.
- Dashboard intelligence — the AI dashboard surfaces what needs attention across quotes, production and payments.
The same data supports a next phase: quote-win analysis (which brief types, sizes and price points convert), reorder prediction against BOM consumption, and plain-language questions about the pipeline and the floor.
How it was built
Scaffold & design system
The app shell, an 18-module navigation, a design system with light and dark themes, and hand-built Radix UI components.
Model the data
Customers, quotes, orders, jobs, BOM lines, materials, artwork versions, samples — the schema for a made-to-order manufacturer.
Build the quotation engine first
The flagship module — real BOM and cost analysis from a brief, margin slider, tier table, version history, approval workflow, PDF — built and verified end to end.
Build out the other 17 modules
Each wired to the store, browser-verified with zero console errors, with a clean production build re-checked after every module.
Two visual QA passes
Post-build audits caught and fixed real bugs — a font fallback, tinted glass surfaces, table number alignment — verified via computed styles, not eyeballing.
What changed
Factual — no invented numbers.
Technology
- Next.js 16 (App Router, TypeScript) — the application, with an auth-gated route group for the CRM.
- Tailwind CSS + hand-built Radix UI — a consistent component set with light and dark themes, both verified.
- Zustand — the state store, persisted client-side; the architecture is designed to move to a hosted database as the next step.
- recharts — the eight-tab reports section.
The build is production-clean — type-checked repeatedly through development and browser-verified with zero console errors across every module.
If you quote every job from scratch, this is buildable for you
An AI quotation engine plus project-style job tracking fits any made-to-order manufacturer:
- Packaging, printing, labels and signage
- Furniture, joinery and modular interiors
- Fabrication — metal, aluminium, glass (see the ZUNOWA case study)
- Garment and textile job work
- Any workshop where the estimate is the hard part of the sale
Start with a free workflow audit. See Manufacturing ERP Software, Custom ERP Software or AI Business Management System.
Questions made-to-order manufacturers ask
Explore more
Manufacturing ERP Software
Production, quality, inventory, orders and an AI layer — built around your plant.
Case study: ZUNOWA
A fabrication ERP for an aluminium and glass business — measurements to installation.
Case study: Splenzo Polyfab
An AI-powered ERP for a PP woven-bag plant — stage-wise production and an AI intelligence layer.
Custom ERP Software
Why made-to-order manufacturers move off generic ERP to a system built around estimation.
AI ERP
What an ERP does when an AI layer reads the data and does the recurring hard work.
All case studies
Every Yukti AI build — jewellery, manufacturing, trading and services.
Summary
Yukti AI built an 18-module AI-first manufacturing CRM and ERP for The Box Factory, a made-to-order manufacturer of decorative and corporate-gift boxes in MDF and rigid board. The flagship module is an AI Quotation & Estimation Engine that takes a plain-language requirement and produces a real bill of materials and cost analysis — board, lamination, foil, printing, cutting, pasting, labour and wastage — with a margin slider, a quantity-tier pricing table, version history that re-runs the estimate on edited briefs, an approval workflow and one-click PDF. The other 17 modules carry a job from enquiry to payment: AI dashboard, CRM, unified inbox, AI box builder, product templates, order management, production planning, inventory, purchase, dispatch, sample management, artwork management, accounting, documents, HR, and an eight-tab reports section including quoted-value-versus-actual per job. Built on Next.js 16, Tailwind, Radix UI, Zustand and recharts, with all 18 modules browser-verified and a production-clean build. Yukti AI builds comparable systems for any made-to-order manufacturer where the estimate is the hard part of the sale.