Skip to content

Made-to-order manufacturingClient work

From Email to Workshop: Quote Engine, AI Intake and Production Orders

I built quoting, AI-assisted intake and production-order tooling for a made-to-order manufacturer, with human review and explicit limits on what the tests prove.

The brief

Context

A small made-to-order manufacturer needed a connected path from an enquiry to a checked specification, quote and production instruction. Requests could include an email, attachments and drawings. A manager had to decide what was actually being ordered before the workshop could act on it.

I built the quoting and operations tooling around that review. AI drafts, people approve.

Sounds familiar?

Symptoms

The situation Made-to-order manufacturing
Tick what applies to you

Tick the lines that describe your case.

Describe my task

The investigation

Diagnosis

Recognition, commercial approval and production release are different decisions. A model response cannot stand in for a manager checking a drawing. A confirmed specification also becomes stale if someone changes its contents.

  1. I used a production-order record as the operational spine and kept the source request, recognised draft and reviewed specification separate. The integration contract ties approval to the displayed version and defines how stale results, duplicate deliveries and retries must be handled.

Handover

What I built

Delivery noteDelivered Engineering and staging records, August–September 2026

The server-side quote engine supports configured positions and pricing rules. The configurator and business portal use the same engine; the portal adds its pricing tier and manager workflow.

For intake, I built AI-assisted recognition and a manager workspace that keeps the email and attachments beside the proposed facts. The manager can correct the draft and review the drawing before approving it. The recognition contract records outcomes and failures separately and prevents a fallback model result from being accepted automatically.

The production tooling includes production-order records, versioned blanks, manufacturing jobs, and stock reservation and consumption rules. Those components need controlled release and confirmation checks. They do not make every request production-ready by themselves.

Before → after

Results

Client work Done for real clients. Client details are anonymised.

  • Historical unit suite for a verified staging revision

    Sep 2026

    Before
    Engineering validation
    After
    2,650 unit tests passed

  • Integration suite for the same revision

    Sep 2026

    Before
    Engineering validation
    After
    239 integration tests passed

  • Acceptance suite for the same revision

    Sep 2026

    155/155

    Before
    Synthetic acceptance scenarios
    After
    155 of 155 checks passed
The full results table
ScopeObserved resultRead on
Historical staging revision, unit suite2,650 tests passed29 September 2026
Same revision, integration suite239 tests passed29 September 2026
Same revision, acceptance suite155 of 155 checks passed29 September 2026

Notes on the numbers

These are engineering checks from a specific revision, not the latest lifetime totals. The documented test runs left no fixture rows behind and checked that unrelated data stayed unchanged.

The boundaries matter. An earlier production snapshot had a release guard that prevented accepted requests from reaching manufacturing. Later staging work does not, by itself, prove that this production defect was resolved. The later release rehearsal also found database constraints that blocked a production update.

I have no evidence here of a real order completing the full intake, production, shipment and payment cycle. I do not claim saved hours, throughput or an autonomous workshop.

A note from Daniilbefore you decide

What this means for a similar business

I can connect quoting, intake and production instructions while keeping a person responsible for the specification. The handover needs to show which parts are built, which were tested on staging, and which still need proof against a real operating cycle.

— Daniil

Situations this case answers

All situations
Read another case

A Payment Provider Shopify Doesn’t Support: Finding a Working Path

I mapped 21 payment-integration experiments and built a server-verified payment continuation, with sandbox results separated from merchant acceptance.
Daniil Maximkin

Hi, I’m Daniil.

I work with you from defining the problem to implementation and handover. You talk to the person who does the work. I work in English and Russian.

My story My work

Have a similar task?

Describe your task

The first answer is free, within one working day. Or write directly: next@taskfordaniel.com