A draft, not a send
No AI answer goes to the customer on its own. The business case comes back as a draft you edit — heading and text — and what you copy or add to the deal room is the draft as you left it.
Two tools in the app: a business case for the customer's decision-makers, and a meeting summary whose next steps become tasks in the action plan. It is part of the Pro tier of Offert & e-signering only, and where no language model is configured it refuses rather than invents.
What it is
Think of a quote to a housing cooperative or a property manager, where the board or procurement has to understand why. That is where a draft saves time. It does not replace reading it.
No AI answer goes to the customer on its own. The business case comes back as a draft you edit — heading and text — and what you copy or add to the deal room is the draft as you left it.
AI is granted by the proposals.ai capability, which exists only in the Pro tier. There is no AI in Start or Väx, and none in scheduling or invoicing.
Where no language model is configured, the system answers 503 and generates nothing. A mock that writes fluent invented text is exactly what must not reach a buyer.
Every generating call needs the module, the Pro tier and the permission to use AI. The permission follows the role, and not every role has it.
AI sessions per organisation have a monthly cap, with a meter in the app. Identical input is served from the cache instead of being sent again, so the same question twice costs one session.
In and out
All three run against a deal room. This is what each takes in and hands back.
How it runs
The site visit, the board meeting, the walk-through with the property manager. Closepact does not record or transcribe; you need a transcript or your own notes.
Open the room for the deal and paste the text. The names of the customer's stakeholders come from the room, so that next steps can be tied to the right person.
You get a summary, decisions and proposed next steps with an owner and a due date. Read it as a proposal, not a record.
One click adds them to the action plan as a milestone with one task per step — the customer's steps on their stakeholders, yours on your users. The customer sees the plan in the room, so change or remove wrong tasks there straight away.
The tasks are followed in the action plan like any others. The generation is stored against the room and the person who ran it.
Data
What you enter, and the details from the deal listed above, are sent to the configured provider to produce the answer. The provider that is implemented is OpenAI, listed as a sub-processor with the region United States on our sub-processor list.
With us, each generation is stored against the deal room and the person who ran it: the input, the answer and the number of tokens. A transcript you paste is part of the input. There is no separate retention period for generations today; they stay until the organisation is deleted.
Nothing in the input is masked before it is sent. So do not paste personal identity numbers, health details or anything else the draft does not need.
Tier and price
AI is part of Pro, 429 kr per user per month, together with deal rooms, analytics and priced lines. It is not in Start, 199 kr, or Väx, 379 kr, and not in the other two modules. Prices exclude VAT on an annual agreement.
AI sessions have a monthly cap per organisation. An answer served from the cache does not count, and a call that fails at the provider gives the session back.
Example: four users, Väx or Pro
Excl. VAT, annual agreement. The difference buys deal rooms, analytics and AI drafting.
Honest about the edges
A language model writes fluently whether or not it is right. These are the edges to know about.
No language setting. The instructions to the model are written in English, so expect the draft may come back in English even when the material is Swedish.
The business case has a section on expected outcomes with figures. The model does not have your costings; check every figure before the text reaches the customer.
What you add to a published room is visible to the customer at once. The business case is edited before it is added; next steps pushed from a meeting summary go straight into the action plan, which the customer sees.
The model does not read the room's documents, quotes or chat.
No recording, transcription or meeting integration. You paste the transcript yourselves.
The text block exists in the API but has no button in the app yet.
Common questions
The answers describe what is built today. If your question is missing, write to sales@closepact.com.
No. AI exists only in the Pro tier of Offert & e-signering, 429 kr per user per month. A call from an organisation without Pro is refused by the server, not just hidden behind a button. If you have moved down from Pro you can still read what was generated before — reading needs only the module — but not generate anything new.
It is sent, with details from the deal, to the configured provider — OpenAI, region United States, today, which is on our sub-processor list. With us the generation is stored against the deal room and the person who ran it. Nothing is masked automatically, so do not paste more personal data than the draft needs. Where no provider is configured, nothing is sent at all.
We cannot promise it. There is no language setting, and the instructions to the model are written in English. Expect that the answer may come back in English, and plan to read and rewrite before the text reaches a Swedish customer. That goes for the business case and the meeting summary alike, even when the transcript is in Swedish.
Yes; a language model always can. It gets the deal's name, stage and value, the customer and the stakeholders — not your costings, quotes or documents — and words the figures in the expected-outcomes section itself. That is why nothing goes to the customer automatically, and why the system refuses rather than falling back on a mock when no language model is available.
A session is a generation actually sent to the provider. There is a monthly cap per organisation, and the meter in the app shows the month's usage. Identical input is served from the cache and costs no new session, and a call that fails at the provider does not count. Tokens are counted for reporting, not as the limit.
Also in this module
Twenty minutes. We set up a quote of yours before the call.
This page describes one capability of the module. What is bought is the module — from 199 kr per user per month.