How it works today
- Search for an old quote that roughly fits — 10 min
- Copy-paste, adapt names and line items, miss one spot — 45 min
- Rewrite descriptions because this client is different — 25 min
- Read it again, because you know you missed something — 10 min
How it works afterwards
- Enter the key facts: client, service, scope, deadline — 3 min
- A draft appears in your structure, with your text modules — 1 min
- Review, sharpen two sentences, approve — 15 min
What we actually build
1. Analyse your quotes
We take five to ten of your best quotes from the last two years. Not as a template, but as material: How do you structure them? Where do your prices sit? Which phrases keep coming back? Which sentences do you use to justify a discount? That's your style — and that's what we teach the model.
2. Build the quote assistant
From that analysis we create a fixed prompt with your structures, text modules and pricing logic — plus the questions it asks you before it writes. Not a chat where you explain who you are every time. A tool that knows you.
3. Plug it into your workflow
The assistant lives where you already work — not in another tab you'll forget. Depending on your setup: as a saved assistant, as a template in your document system, or connected directly to your client data.
4. Run alongside for two weeks
You create the first five quotes together with me. Then we sharpen what doesn't sit right yet. A tool that ends up in a drawer after the session has achieved nothing.
Tools — and where it doesn't help
Usually Claude or ChatGPT as the language model — which of the two depends on your writing style, not on preference. For connecting to existing data, an automation layer may be added depending on your setup.
Where it doesn't help: for quotes that require genuine technical calculation every time, AI saves little — the calculation work remains. The gain then lies in the accompanying text, not the core. I'd rather tell you that before than after.
Before client data goes into a model
Quotes contain client names, prices and sometimes confidential details. Which of those may go into which tool is governed by the revised Swiss data protection act — and the answer is rarely "all of them". In practice it can be solved cleanly: anonymise templates, insert client data only in the last step, or choose a model that doesn't use your data for training.
→ How we bring data protection and AI together
What this means in time
If you write ten quotes a month and gain 70 minutes per quote, you gain almost twelve hours a month. At an internal hourly value of CHF 120, that's around CHF 1,400 per month — counting only the time, not the quotes that never went out because there was no evening left to write them.
Frequently asked questions
Doesn't an AI quote sound generic?
If you set it up generically, yes. The whole effort in step 1 serves exactly the opposite: the model works with your sentences from your previous quotes. Clients usually don't notice a difference — because it is your text, just assembled faster.
May I enter client data into ChatGPT?
Flatly no, with nuance often yes. It depends on which data, which subscription and which settings. Business tiers of the major providers don't use inputs for training by default — consumer tiers partly do. We check this for your specific case before the first real quote goes through.
How long does the setup take?
A working quote assistant usually takes two to three 60-minute sessions spread over two weeks. You need the time in between to test it in daily work — that's where the real sharpening comes from.
Do I need technical knowledge?
No. If you write quotes in Word or your CRM today, that's enough. I build the tool, you learn to use it — not to program it.
Bring your last five quotes
In the free intro call we look at them together and in twenty minutes you'll see how much of it can be automated. No sales pitch.