Half past six in the evening, the site is closed, and now the second working day begins: writing quotes, deciphering the note from the customer appointment, writing down for the apprentice what needs to go to the site tomorrow, confirming the discussed changes to the customer. No trade gets paid for paperwork, but every trade suffers from it.
Meanwhile everyone is talking about AI agents. This article translates the term into the daily life of a trades business: what an agent really is, which tasks it can take over, and where its limits are.
Chatbot, AI, agent: a quick sorting
A chatbot answers questions, nothing more. An AI agent acts: it documents, files, drafts, reminds. The difference is the one between a reference book and an office helper. What matters is not how clever the model is, but whether it knows your business: your appointments, your customers, your agreements. An agent without that context is like a new employee on their first day, every single time.
Where trades businesses really lose time
The most expensive hour in a trades business is the unpaid one in the office. And it almost always exists for the same reason: what was discussed during the day is written down nowhere.
- The customer appointment for the bathroom renovation: 45 minutes of conversation, three special requests, two changed measurements. Captured: four keywords on a slip of paper.
- The site meeting with the general contractor: who delivers what by when? Lives in the memory of three people, in three versions.
- The phone call with the wholesaler: the promised delivery date becomes a debate two weeks later.
In the evening all of this gets reconstructed, translated into quotes, poured into emails. Exactly this reconstruction work is what an agent can take over, because it consists of speech, and speech is what AI processes best today.
Concrete use cases by trade
Electricians: electrical planning without memory gaps
The planning appointment in a new build is the textbook case: two hours through the shell construction, room by room, with decisions falling in every one of them. Sockets left or right of the bed, a network point in the study, smart home preparation yes or no, an outdoor socket on the terrace. If the appointment is recorded via phone, every room decision is documented afterwards, as a solid basis for the quote and later for the execution. The classic dispute, "but the double socket in the kitchen was agreed", has a readable answer.
Plumbers and HVAC: making special requests binding
Half the project is decided in the bathroom renovation conversation: walk-in shower, fixture series, tile backsplash height, connection for the towel radiator. Exactly these details get lost on the way from the customer's living room into the quote. With a documented appointment, the summary goes to the customer as confirmation, the special request is captured, and later "we agreed something different" debates have one shared source instead of two memories.
Carpenters and joiners: measurements with a hundred details
Built-in wardrobe, kitchen, shop fitting: hardly any trade has more detail decisions per appointment. Wood type, front finish, fittings, handles, soft close, interior layout. The transcript captures everything, including the side remarks ("the bottom drawer a bit deeper, please") that had no room on the measurement sheet and would otherwise resurface only during installation, as a complaint.
Drywall and construction: site meetings with several trades
The weekly site meeting is where commitments are made and lost. Who erects the substructure by when, when does the electrician get into the wall, who files the obstruction notice? The documented conversation becomes a task list with owners and dates that gets tracked, instead of fading away in the site diary. For meetings with several participants, the multi-device recording pays off, more on that in a moment.
The pattern is the same everywhere: conversation in, documented agreements and tasks out.
What this looks like in practice with Sally
Sally is an AI assistant that covers exactly this flow, without new hardware and without switching systems.
Document on site, with the phone that is already there
Start a recording in the app, put the phone on the table, run the appointment. For larger meetings, say the site meeting with four trades at the table, several phones connect into one shared recording: every voice is captured up close and the attribution of who committed to what stays unambiguous. The page about the transcription of in-person meetings shows how. Online appointments in Google Meet, Zoom, Microsoft Teams and Webex are covered automatically by the meeting bot. After the appointment, you get a summary, decisions and recognised tasks with owners, results instead of raw material.
From the conversation into your own tools
Documentation that sits in yet another silo helps nobody. Through 8,000+ connections, the results flow into the programs the business already runs: calendars, task lists, bookkeeping. The task "order material for the Bergers' bathroom" then sits where everyone looks every morning anyway, not in an app nobody opens.
The office afterwards: follow-up and looking things up
The confirmation email to the customer is drafted from the conversation content, with the real agreements and measurements instead of text blocks; why that makes the difference is covered in our article on the follow-up email after a meeting. And all appointments land in a searchable knowledge base: "What did we agree with the Bergers about the tile backsplash?" becomes a seconds-long search query, not a file dig. That also helps when the colleague who ran the appointment is off sick.
Legally clean: consent and data protection
Consent: one sentence at the start
Secret recordings are a criminal offence in Germany (Section 201 of the Criminal Code) and restricted in most countries; recording happens only with everyone's agreement. In a trades business, one sentence at the start of the appointment solves it: "I am recording our conversation so nothing gets lost, is that okay?" Experience shows customers not only accept it but appreciate it, because their wishes are captured bindingly too. If you like, add a note about conversation documentation to your order confirmation, then the framework is clear from the start.
Data residency: customer data stays in Germany
Customer data from conversations is personal data, often including address, building project and budget. With many AI tools it ends up in US clouds, which under GDPR requires a transfer assessment no trades business wants to run. Sally processes exclusively in Germany, GDPR-compliant, operated by Aliru GmbH from Mannheim. Details on the page about GDPR and security.
Staying honest: what an agent cannot do
An AI agent does not lay cables and does not fit kitchens. It does not replace expertise either: whether the documented customer requirement makes technical sense is still the master craftsman's call. And it needs power and connectivity; in a dead zone on the shell construction site, the app's offline recording helps, but the analysis comes once you are back online. The realistic expectation: the agent takes over the writing work around the conversation. That is not a revolution on the site, but it is several office hours per week.
Conclusion
The entry point for AI in a trades business is not the big digitalisation project, it is the next customer appointment: document instead of reconstruct. From there, the use cases grow on their own, from the task list to the basis for quotes. For a broader overview of useful tools, see our article on the 7 best AI tools for tradespeople.
Your own appointment is the fastest way to judge it: test Sally free for 30 days and simply record your next customer appointment.
