Starting point
Weber Painting Co. (fictional) is a typical facade and painting business with five employees. The owner calculates quotes, handles customer conversations and stands on the scaffolding himself several times a week. There’s no office in the classic sense: order intake runs through the owner’s work phone, which mostly sits silent in the company van during working hours.
Over three months, the business analyzed its call log: on average 45 calls a month went unanswered or landed on voicemail, most of them between 7 am and 5 pm, exactly when nobody could pick up. A callback in the evening reached some of the prospects, the rest had already hired another business.
The setup (architecture as a list)
- Phone number: existing landline, forwarded to the AI phone assistant, customers dial the same number as before.
- Knowledge base: service overview (facade, interior painting, coating work), rough price ranges per service, typical lead times for appointments.
- Conversation script: greeting, capturing the request (type of work, approximate area, address), offering an appointment for a callback or an on-site inspection.
- Escalation: keywords like “complaint” or “water damage” put the call straight through to the owner.
- Connection: every answered call creates a case with contact details, summary and next step in the ERP; a calendar slot is reserved.
A Tuesday with the system
7:40 am: the owner is already on the scaffolding. A prospect calls needing a quote for 120 m² of facade. The AI takes down the request, arranges a callback for the next morning at 8:15 am and creates the case in the ERP.
11:00 am: an existing customer reports a stain on the freshly painted wall, a complaint. The keyword triggers, the call is put straight through to the owner, who stops by the same day.
4:30 pm: the owner sees four new cases from the day’s activity in the console, sorted by urgency: no notes, no callback attempt into the void, no forgotten number.
Model calculation
All assumptions are stated openly; we determine your own numbers in the potential assessment.
| Assumption | Value |
|---|---|
| Previously unanswered requests/month (from call log) | 45 |
| Answered & qualified by the AI phone assistant | 31 (69%) |
| Close rate on qualified requests | 25% |
| Resulting jobs/month | ≈ 7.75 |
| Avg. order value (facade/painting work) | €890 |
| Additional order volume/month | ≈ €6,900 |
| Calculated margin on order volume | 35% |
| Additional contribution margin/month | ≈ €2,415 |
| Ongoing cost of the AI phone assistant | €280/month |
| Net contribution margin/month | ≈ €2,135 |
| Setup cost (one-time) | €2,400 |
| Payback period | ≈ 1.1 months → ”< 3 months” |
Important: the €6,900 is additional order volume, not profit. Only after deducting materials, labor and travel costs (here roughly set at a 35% margin) is there an actual contribution margin left, against which the AI’s ongoing costs are calculated.
Limits & learnings
Not every one of the 31 additionally answered calls becomes a job; the 25% close rate is a rough experience-based figure from comparable businesses, not a guarantee. Complaints and disputes were deliberately routed around the system straight to the owner from the start, because human judgment matters here, not AI efficiency. And the knowledge base needed reworking several times in the first two weeks, because the initial price ranges were too coarse and would have led the AI to give unreliable figures; that was corrected before going live, not after.