Running a session
What to expect, in the order it happens. You do not need to be a consultant to run this. The seven steps below are the same seven you see along the top of your session, so you can follow either. It works two ways: as pre-work people fill in before a live discussion, or live with everyone in the room together. The steps are the same and only the timing changes.
At a glance
- 1Set up
- 2Invite
- 3Reflect
- 4Surface
- 5Discuss
- 6Decide
- 7Close
Give the session a team name and a few light bits of context: sector, country, rough size, and what the team runs on. Nothing here is evaluated. It helps the room read its own answers later. You are also asked who sets AI policy and who decides key directions, which is how the tool knows where to send anything the team cannot decide by itself. Choose here whether responses are anonymous. You can also switch a named session to anonymous later, though not the other way round. You get a six-character join code straight away.
What you do: Dashboard → New session.
Share the join code or the direct link. Your team does not need an account. They pick a display name and they are in. It works on phones, so you can drop the link into a chat or read the code out at the start of a meeting.
What you do: Session page → Invite your team.
Each person reflects on their own, across the three pillars: how AI is governed here, how it lands on the people doing the work, and where it builds real advantage. They write what is already in place, what is still open, what is missing, what worries them, what they would not do, and the tools they are curious about. The order shifts with their role, so people start where they have the most to say. Everyone still sees everything. They can do this before the meeting or in the room.
What you do: Session page → Members, to watch reflections come in. The “Check for new” button refreshes the list.
When enough is in, pause submissions so the picture stops moving underneath you. Existing answers stay visible. Then run Surface patterns. The AI reads every contribution and works out which pillar it belongs to from what it says, rather than which box it was typed into. You get a reading for each pillar, the alignments and tensions across the room, the open questions, and the leg where trust is wobbling. This is the analysis. It is not the answer, and it is not the report.
What you do: Session page → Submissions toggle, then Open landscape → Surface patterns.
Name what is steady first. It takes two minutes and it banks the credit. Then spend the time on the wobbling leg, because that is where the conversation is most needed. Work the open questions and the gaps for that pillar. The prompt bank has questions you can borrow if the room stalls, and the team can vote on what matters most.
What you do: Session page → Open landscape, plus the facilitator prompt bank.
Turn the conversation into a small number of decisions the team will stand behind. Each thing the room works becomes one outcome, with an owner where it needs one. Anything above the room's authority is referred up to the person you named at set-up, as the team's recommendation. How each outcome works is set out below.
What you do: Session page → Decisions, or Who decides? on any contribution.
Generate the AI trust report, then close the session. Closing ends submissions for good and locks the picture in place. You can still read it, export it, and archive it later. A clear ending is what turns a conversation into a decision.
What you do: Session page → Generate AI trust report, then Close session.
How decisions work
One rule sits under all of this: an individual's input is kept separate from the group's decision. Someone reflecting on their own can offer a view. They cannot decide for the team. That is why nobody can tag their own contribution "decided" while they reflect, and why decisions only exist once the room has met.
In the room, each thing you work becomes one of these. Pick the one that is honest, not the one that sounds most decisive.
| Outcome | Use it when |
|---|---|
| Agreed | The room has settled it. No one needs to do anything next; it's just true now. |
| Action | Someone is going to do something. Give it an owner and a date, or it won't happen. |
| On the horizon | Worth doing, not now. Deferring on purpose is a decision. Losing it by accident is not. |
| Needs an owner | The team agrees it matters, but nobody is holding it yet. Name that honestly. |
| Does not require a decision | The question came up and, looking at it, there is nothing to decide. Close it rather than leave it hanging. |
| For sign-off above the team | Above the room's authority. The team can recommend, but someone else signs it off. |
Referring something up
Some things are not the room's to decide, and pretending otherwise is how a plan quietly dies. Mark those "For sign-off above the team" and they go to the person you named at set-up, as the team's recommendation rather than a decision. They appear in the report under "Referred up for a decision", so the reader can see what is waiting on someone else.
Nothing gets lost
When you generate the report, the tool lists any open questions the team never recorded a decision against. Select one to work it through, or generate anyway and they appear in the report as open. Either is fine. An open question the team knows about beats a tidy report that hides it.
What you get out of it
Four outputs, doing different jobs. The first is for the room. The second is the one you send.
Where the team seems aligned, where there is tension, the open questions worth the time, suggested next steps, a reading for each pillar, and the leg where trust is wobbling. This is the analysis and it is meant to start a conversation, not end one. Download it as a PDF or copy it as markdown if you want it in your own notes.
Where: Open landscape → Surface patterns.
The read, what the team decided, what was referred up for a decision, the whole landscape, and who took part, with the date and the numbers behind it. It is built only from what your team said. While the session is open it is watermarked DRAFT. Close the session and it comes out clean.
Where: Session page → Generate AI trust report.
Every reflection and every decision in one spreadsheet, with its theme, status, owner and dates. For when you want to do your own analysis, or keep a record outside the tool.
Where: Session page → Export all data (CSV).
The same material as a slide sequence, one focus per slide, so you can talk the team through what came back without scrolling a wall of text. It exports as a one-page-per-slide PDF.
Where: Open landscape → Presenter mode.
If you set the session to anonymous, every one of these respects it. No display name or role leaves the tool.
Two ways to run it
Same steps, different rhythm. Pick whichever fits the team you've got.
- Send the join code a few days ahead. People submit when they have a quiet 30–45 minutes.
- Pause submissions when the meeting starts. The room reads the landscape together.
- Re-open near the end so people can add what only emerged in the conversation.
- Share the code at the start. Give the room a quiet 30–45 minutes to reflect on phones or laptops.
- Pause submissions, then walk through the landscape together — patterns, worries, decisions.
- Re-open briefly at the end if new things surfaced; close the session once the room agrees.
A few principles
- There's no required level of AI experience — including none. Nobody is being evaluated on what they write.
- Worries and non-negotiables matter as much as decisions and tools. Don't rush past them.
- Aim for a small number of steps with named owners, and be explicit about what you're choosing not to do, and why.
- If the room needs cover to be candid, you can make the session anonymous at any point, not just at set-up. It applies to what has already been written too. It cannot be undone, so it protects everyone at once rather than singling anyone out.
- The room makes the decisions. This tool is just a way to see and organise them.