Connect the systems that keep work moving.

That is common, and it is checked during the paid assessment rather than assumed. Plenty of systems still in daily use across manufacturing were never designed to be integrated with.

Where there is no supported API, the workable options are a read-only database view, a scheduled export and import, or file transfer. Carefully controlled UI automation is a last resort, because a layout change or an unexpected dialog can break it.

If the only viable path would be unreliable on a business-critical record, the honest recommendation is not to automate that step. See the integration paths.

No. Estimating judgement stays with the estimator: reading the drawing, deciding process routing, working out whether the job suits the shop, and setting price and lead time.

What can be removed is the work around it — retyping details that already exist in an email, sorting a shared inbox by hand, chasing a quote nobody has updated, and looking the same customer up in three systems.

Anything involving a price or a customer-facing commitment stays under human approval. That is a design rule, not a setting to be switched off later.

Follow-up messages are drafted from templates the client has approved, filled with the quote details, and placed in front of the owner to approve, edit or cancel. Automatic sending without approval is not offered as a default configuration.

The client does. Code, accounts, configuration and documentation belong to the client, and integrations are built in client-owned accounts wherever practical.

The repository is handed over, and so is every credential. Nothing is held hostage on a proprietary platform that would have to be rented indefinitely to keep the work running.

This is a founder-led business, so that risk is real. It is handled by design rather than denied.

Projects are scoped small, built on standard technologies rather than unusual ones, and documented as they are built. The repository, the accounts and the credentials are in client hands, so another developer can pick the work up. Support arrangements state response expectations one person can honestly meet.

A 30-minute conversation about one workflow, followed by a concise written recap, normally within two business days. The recap covers the workflow as described, one or two plausible automation opportunities, the important uncertainties, whether a paid assessment appears worthwhile, and a recommended next step.

It does not include a free architecture, a full process map, a detailed integration specification or an implementation plan. Those are the paid assessment, and pretending otherwise would either waste your time or produce something too shallow to rely on.

Scope comes from the paid assessment rather than from a first conversation. The assessment looks at the real systems and a real sample of requests, and produces feasibility findings per integration point, risks, a future-state workflow, and an implementation scope with its assumptions and exclusions written down.

The proposal that follows states acceptance criteria in business terms, so completion is something you can check rather than something you have to take on trust. See how engagements work.

Pricing and margin decisions. Customer-facing commitments without approval. Interpretation of technical drawings. Engineering, quality and non-conformance judgements. Credit, terms and account decisions.

Anything involving export-controlled or otherwise restricted technical data is out of scope for the public form and the demonstration. Low-confidence classifications and uncertain customer matches always go to a person rather than to a best guess.