AI Disclosure

1. The seats are AI, not people

Every seat in hlpr (the CFO, the CRO, the Challenger, the chief of staff and the rest) is an AI agent. Seats may have a name, a picture and a voice so a team is easy to follow, but no person is behind them. Seat pictures are generated by AI or chosen by your team, and seat voices are synthetic. The people in your company are the only people in your workspace.

2. What the AI does

• Reads the data you connect or upload, within the access you grant, and answers from it.
• Proposes decisions with evidence and a success measure, argues them out with the other seats, and summarizes them for you.
• Drafts work for your approval: emails, plans, documents and dashboards.
• Takes part in meetings in writing or by voice, and writes the minutes and follow-ups.
• Acts on its own only in an area where you have turned that on (see section 4).

3. Models and providers

Seats think and write with Anthropic's Claude models: faster models for routine work and the most capable ones for final write-ups. Live meetings use ElevenLabs for seat voices, Deepgram for speech to text, and LiveKit to carry people's voices between them during a call (we do not store call audio), and seat pictures are generated with Google's Gemini API. Each provider processes only what its part of the work needs, under its commercial terms. The full list is in our Trust Center.

We do not use your data, or data from your connected tools, to train AI models, and our model providers' commercial terms do not allow them to train on it by default.

4. Human oversight

• Seats propose; people decide. Only a company's owner can approve or send back a decision, reopen one, or change what seats may do on their own.
• Seats can never decide items, change their own permissions or reopen a decision. These rules are enforced on our servers, not just hidden in the interface.
• Every seat starts by drafting for approval. An owner can let a seat act on its own in one area at a time once it has a track record there, and can switch that off at any time. People can still give feedback on what a seat did.
• Kill switches stop one seat, a whole company or the whole platform.
• Every decision and change is written to an audit log that cannot be edited.

5. Limits of AI

AI can be wrong, incomplete or out of date, and can sound confident when it is. Seats cite where their figures come from and say when data looks unreliable, but check anything important before you rely on it. What seats produce is information to help you decide, not legal, financial, tax, investment or other professional advice.

6. What you send out

When a seat drafts or, where you have allowed it, sends an email or other message on your company's behalf, your company is the sender. You are responsible for its content and for telling recipients that AI was involved wherever the law or your own policies require it.

7. Meetings

Meetings in hlpr are recorded and transcribed, and every meeting says so when it starts. Speech is turned into text so seats can answer; the transcript is saved with the meeting in your company's workspace.

8. Our blog

Posts on the hlpr blog may be drafted with the help of AI (Anthropic's Claude), and their cover pictures made with Google's Gemini. They are published under the founder's name, and the founder is responsible for what they say.

9. Changes and contact

We update this page when how hlpr uses AI changes, and change the date above. Questions: hello@hlpr.ai, or write to: