AI in the Trades: How a Heating Company Prepares Quotes in 20 Minutes Instead of Three Hours

A heating and plumbing company in Bavaria with 15 employees now prepares a quote in about 20 minutes. Before, it took about three hours of office work. We do not name the company for confidentiality reasons.

Beforeabout 3 h

Todayabout 20 min

saved

almost 90%less time for every quote

That puts the company in a small minority. According to the German digital association Bitkom, 68 percent of trades businesses already send their quotes digitally, but very few use artificial intelligence:

4 in 100

German trades businesses used artificial intelligence in 2025.

The starting point

Enquiries came in through three channels: website, email and emergency calls. Each landed somewhere else. The knowledge needed for a good quote was scattered as well: which products fit which system, what they cost right now, how long installation takes. Individual people knew it, or it sat in folders. Every quote was handwork.

Why we did not start with AI

An AI assistant that knows neither prices nor customers nor processes writes nicely worded quotes with the wrong content. So we built in four steps, and AI only came in at the third.

1. 1

   Website and leads

   Enquiries arrive in one place.

2. 2

   ERP with Odoo

   One shared data foundation.

3. AI from here

   3AI

   Company brain

   The company's knowledge, separate from the model.

4. 4AI

   Agent

   Answers and acts, a person approves.

The website with about 80 pages bundles contact, quote and emergency enquiries in one place. Odoo holds customers, quotes and orders in one system. The company brain stores products, prices, procedures and documents separately from the AI model. And the agent draws on both.

Today the agent handles four tasks: it answers questions about products, prices and procedures, creates leads directly in Odoo, provides information about invoices and prepares quotes in Odoo. No quote leaves the company without someone from the team approving it.

How three hours become 20 minutes

The AI does not type faster. What disappears is the searching: what did the customer ask for, which products fit, what do they cost, how was a similar case solved last time. The agent puts the draft together from the company's knowledge and the data in Odoo. Someone in the office checks it, adjusts it and approves it. Research turns into review, and review is faster.

What other businesses can take from it

• The process matters more than the model. Without steps 1 and 2, step 4 would not have worked.
• The knowledge belongs to the company. A better model can be used without rebuilding the knowledge.
• Approval stays with people. That builds trust in the team and prevents costly mistakes.
• Start small, but measure. A clearly scoped process with a measurable duration shows what improved.
• One process leads to the next. Because the knowledge base is in place, every further one goes faster.

Next, the technicians get a mobile solution that draws on the same company brain. Anyone on site with a question about a system gets the answer from the same knowledge the office uses.

And the costs?

What a first agent or a company brain costs is shown in our packages. Businesses in Bavaria can receive a grant of up to half of the eligible costs through the Digitalbonus, provided the application is submitted before the order is placed. The full case study is at Case study: trades.

Sources

• Bitkom: Handwerk: Azubis machen Betriebe fit für die Digitalisierung, press release of 27 August 2025, survey of 504 trades businesses