AI and privacy: you stay in control. dot works for you, not around you.
dot writes drafts and prepares proposals, but sends and changes nothing without you. You switch on whatever works on its own, and every run is in the log.
Always on, when you ask
A person asks for it and approves it
Works on its own, off by default
Nothing happens until you switch it on
Trust is something you should see.So dot shows what it does.
dot works with your permissions, shows what it did and leaves your customers alone until you decide otherwise.
Never by itself
dot writes, a colleague sends.
dot prepares drafts, in the inbox helper too, but an email never goes to a customer by itself. Ask for a task, rule or status change and you get a card first. Only the WhatsApp assistant answers customers itself, if you switch it on, and then it says it's AI.
- Never emails a customer by itself
- Never throws a colleague under the bus in a draft
- Never pretends to be a person
Good morning, thank you for your patience. You'll receive the quote for servicing the 24 boilers tomorrow. Do you have a preferred day for our visit?
The draft for VvE Lindehof is ready. A colleague reads it and clicks Send themselves.
Permissions
dot never sees more than you.
dot works with the permissions of whoever is using it. A technician with the Field worker work profile hears no prices, totals or revenue, and dot never sees cost prices or login details for anyone.
- Only what you're allowed to see too
- Ask dot whether you have access, it tells you why or why not
- External members have no access to dot
Accepted on 1 Oct: an 8 kW air-to-water heat pump, a 100-litre buffer tank, a 200-litre hot water cylinder, the fuse box work, 14 metres of pipework and the fitting.
dot doesn't name prices or totals for the Field worker profile.
Kees hears what's on Thomas' quote, but not what it costs.
Log
Every run is in the log.
Studio lists every AI run, newest first: when, who started it, which tool, which record and whether it succeeded. Open a run for the outcome, how much knowledge dot used and the credits.
- Search by tool, person or record
- Usage per day, per tool and per colleague
- Failed runs cost nothing
Form check
Maintenance - Frank Huisman · automatic
Outcome
Needs completing
Knowledge
2 fragments
Credits
4
Inbox helper
Heat pump - Thomas van Leeuwen · automatic
Draft email reply
Heat pump - Peter Kuijpers · Fleur
Daily brief
Fleur
Form check on Frank Huisman's work order: needs completing, for 4 credits.
Your data
Where your data lives.
Your database is in Frankfurt and the app runs in Germany. For the AI itself, stipt uses models from Anthropic (Claude) and Cohere; that processing goes through the US, under the EU standard contractual clauses.
- Your company's data doesn't train AI models
- An attachment in the chat isn't stored
- Contact email addresses and phone numbers are encrypted
Your data
Database in Frankfurt
The app
Runs in Germany
AI models
Anthropic (Claude) and Cohere
Via the US, under the EU standard contractual clauses
Not used to train AI models
What dot reads, it uses for your answer. Not to train a model.
Cost
Credits from one pot.
All of dot's work comes from one monthly bundle for your whole company: 1,000 credits plus 500 per extra team member. You can top up any time, automatically too, below a threshold you choose.
- The owner is notified at 80% of the bundle
- Purchased credits stay valid for 6 months
- Studio shows usage per tool
1,150 of 4,500
Monthly bundle: 1,000 + 7 × 500 credits
1,150 of 4,500 credits used. At 80%, the owner gets a notification.
How to switch on a tool.Calmly, and in training first.
Everything that works on its own is off by default. When you switch such a tool on, you first choose how far it goes, try it on a real case, and it starts in training. Here, the inbox helper, which reads along with email.
Read along
dot doesn't carry out anything itself
Prepare tasks
Work from a notification becomes a task
Handle on its own
After that task, dot handles the conversation
Email from VvE Lindehof · Fri 2 Oct
Nothing goes to the customer.
Is this right?
92%
agreed with dot, last 30 days
3
lessons for future emails
Off by default
Sorting, urgency flags, the inbox helper, form checks, customer happiness, the WhatsApp assistant and the internal bot are off until you switch them on.
Always on, on request
A draft, a summary or an analysis only starts when someone asks for it, and a person approves it.
Limits per feature
Never promise a price or discount, no legal or medical advice, complaints go to a person. Plus your own instructions.
Training first
While training is on, the inbox helper only reads along. Your team says for each conversation whether dot was right.
Lessons you can see
A correction with a reason becomes a lesson for future emails. You see them on the tool and switch them off one by one.
How often was it right?
On the tool you see what dot did in the last 30 days and how often the team agreed with it.
Tone and form of address
Per customer‑facing tool you choose business‑like, friendly, direct or enthusiastic, and informal or formal.
In the customer's language
Customer‑facing tools answer in the language the customer writes in. Internal output stays in Dutch.
Everything you want to know about dot.Small things that build trust.
Which models, what a tool may do and who sets it up: the details that make AI explainable.
Which models
Mostly Claude from Anthropic. Knowledge base search runs through Cohere. stipt decides which model a feature uses; you only pay for usage.
What a tool may do
Each tool lists the steps it may take itself: an eye for reading, a pencil for changing.
Attachments don't stay
A photo or PDF in the chat belongs to that one message. It isn't stored; only the file name remains.
Settings stay with the admin
If you ask for a settings change as a team member, dot explains how it works and refers you to an admin for the change itself.
Automatic top‑up
When your balance drops below your threshold, stipt tops up through your saved payment method, with a confirmation by email.
Feedback with one thumb
If an answer isn't right, give it a thumbs down. Admins see all feedback in Studio, with the source dot used.
Also in AI.What else dot does for you.
Frequently asked questionsAbout control, privacy and cost.
Does dot email customers by itself?
No. dot writes drafts and a colleague sends them. Only the WhatsApp assistant answers customers itself, if you switch it on. In its first reply it tells the customer they're chatting with AI.
Is my data used to train AI?
No. Your company's data is not used to train AI models.
Where is my data stored?
Your database is in Frankfurt and the app runs in Germany. AI processing goes through the US, under the EU standard contractual clauses. Contact email addresses and phone numbers are encrypted, with a key per workspace.
Can I switch AI features off?
Everything that works on its own is off by default and you switch it on per feature in Studio. Help on request, like a draft or a summary, you only use when you ask for it.
What does AI cost in stipt?
AI runs on credits from one pot per company: 1,000 per month plus 500 per extra team member. You can top up any time, and the owner is notified at 80% of the bundle.
Can I see what dot has done?
Yes. The dot‑log in Studio shows every run with the time, who started it, the tool, the record and the status. Under Usage you see the credits per day, per tool and per colleague.
From the knowledge base
AI you can explain.
In a demo we show what dot would do for you, and what it wouldn't.