From Dictation to Reviewed Quote: Why the Workflow Matters More Than the AI

There is work in a trade business owner’s day that nobody pays for: the work at the desk, after hours, when the notes from a site visit have to become a quote. Gathering line items, looking up prices, phrasing, formatting. Anyone doing site visits regularly loses valuable concentration there. And between site and office, something regularly goes missing: the measurement you meant to remember, the extra job the customer mentioned in passing.

This is exactly where we started in the Lab, and it became the Werklinie module “Angebot”. The flow: after the visit, the owner speaks into the phone what needs to be done, the way he would explain it to an apprentice. From that, a structured quote with line items and prices is created. He reviews it, corrects where needed, and sends it. Rework becomes a reviewable flow.

The most important decision is not in the AI

Anyone seeing such a system for the first time marvels at the speech recognition, or at how a stream of talk becomes clean line items. Both are available technology today. The decisions that make the tool fit for daily use lie elsewhere.

The preview before sending. Nothing goes out that the owner has not seen. The system produces a draft, not a finished document. This single stage decides whether the tool builds trust or lands in the bin after the first mistake. AI makes mistakes, that is no secret. A flow that plans for this is robust. One that ignores it is negligent.

The pricing logic stays with the business. What work costs and what margin goes on materials is not decided by a model, these are rules the business defines. The AI structures and phrases, it does not calculate freely. That too is an architecture decision, not a technical given.

The flow adapts to the business, not the other way round. Every business has its own way of writing quotes. An off-the-shelf tool forces the same mask on everyone and charges by user count. A tailored module adopts the existing structure and runs on own infrastructure, with European AI and no per-seat subscription.

Why I show this publicly

For the same reason I prefer showing demos over slides: the difference between a concept and a running system only shows in the running system. The road there was a lesson in itself. The first version, then called Sprechzettel, was a prototype with exactly one purpose: finding out whether the two-stage flow holds up in daily use. It does. Only then did it become a module of Werklinie, alongside company knowledge and site documentation.

The line is live, and pilot businesses are welcome: a business that knows its quoting process and gives honest feedback gets a tool cut to its own workflows in return, not to the industry average.

The core

The bottleneck in the trades is not the order book, it is unpaid administrative time. Dictating instead of typing reclaims a tangible part of it, but only if the flow is cut right: preview before sending, pricing authority with the business, integration instead of an isolated tool. The technology is there. What is usually missing is only the fit to the individual business, and that is smaller than most people think.