An experiment in better decisions
TableAI is an open, experimental project exploring what happens when you separate input from discussion — and let AI moderate the space between.
Built by a small group obsessed with product, design, AI — and how organizations actually make decisions.
The problem we're exploring
Group decisions are broken by default
You run a team. You need a decision. Getting everyone in a room takes two weeks — and when you finally meet, the same three people talk.
The person who speaks first anchors everyone else — that's The First Speaker Problem. Junior team members self-censor around leadership — The Silence of Seniority. What people actually think diverges further from what they say with each level of hierarchy — The Candor Gap, which grows roughly 40% per management level.
These aren't personality issues. They're structural flaws in how groups communicate. Meetings amplify them. The fix isn't better facilitation — it's separating input collection from group discussion entirely.
The approach
Separate input from discussion
TableAI collects perspectives through private, 1:1 AI interviews. Each participant talks to an AI moderator that asks follow-up questions, probes for reasoning, and extracts the full picture — without anyone else in the room. No anchoring, no conformity pressure, no performance.
Then the AI synthesizes everything: where people agree, where they disagree, and what questions remain open. Sometimes that's enough to decide. When it's not, the synthesis becomes a focused agenda — so the meeting that does happen starts where meetings usually end.
Full transparency
An experiment, not a product
TableAI is an experimental project. There's no company behind it, no funding round, no growth targets. It exists because the question — can AI help groups think together more honestly? — felt worth exploring.
There's no subscription and no paywall, and for now the AI is included too: no API key, no setup. If you prefer to run sessions on your own Anthropic account, you can add your key in Settings; it is encrypted on our end and never shared.
Every session teaches us something. Early users pushed us to deepen the follow-up questions, sharpen the synthesis format, and rethink how disagreements surface. The project keeps evolving — and that's the point.
Teams use it for budget discussions, vendor evaluations, and strategy alignment — anywhere a group needs to think together without the usual meeting dynamics getting in the way.
What we believe
Decisions deserve better than whoever talks loudest
Good decisions come from good input. Good input requires safety, time, and space — things that group settings rarely provide.
Candor requires separation
People are more honest when they're not performing for an audience. Collecting input privately isn't a workaround — it's the point.
Most meetings should have been async
Synchronous time is expensive. Reserve it for the conversation that actually needs to happen — after everyone's thinking is already on the table.
AI should moderate, not decide
The AI's job is to ask better questions and organize what it hears. The decisions stay with the people who have to live with them.
Try it yourself
Create a session, invite a few people, and see what comes back.