What makes a good conversation about AI?
This is a tool for one thing: helping your team have a good conversation about AI that ends in decisions you act on. Three things have to be true at once. It has to be a good conversation. It has to be specifically about AI. And it has to lead to decisions the team will stand behind, with people clearer on where they stand and governance that fits. Most AI strategy work fails one of those three tests. Good Conversations About AI is built around all three.
What makes the conversation good: trust
Trust is the engine. It is not the point of the conversation, but it is what makes people speak honestly and what makes the decisions hold afterwards. Where trust is low, people say what is safe rather than what is true, and the plan quietly unravels once everyone leaves the room.
Trust rests on three things. When it fails, it usually fails on one of them rather than all three. We read which one is straining from what your team writes, and the session goes there.
We build on the trust triangle of Frances Frei and Anne Morriss at Harvard Business School, who hold that trust rests on three things: authenticity (people see the real you), logic (your reasoning and judgement are sound), and empathy (you are there for them, not only for yourself). We have adapted those three drivers into the three pillars the conversation is structured around.
Governance & Integrity
How AI is governed here: the policy, the guardrails, what has been decided, and the red lines that are not up for negotiation. This is also where the law, your sector and your client commitments sit.
People & Coexistence
How AI lands on the people doing the work: their roles, their hopes, and their worries. AI as something that helps people do the work, not a quiet route to replacing them.
Advantage, Innovation & Assurance
What only this team can do, and where AI deepens it. Choosing the few things worth doing and what to leave alone, doing them well, and being able to show it.
What this is not. The reading is a relative one: where trust is straining in your team, against the rest of what your team said. It is not a maturity score, and it does not rank you against anyone else.
Credit: this approach builds on the trust triangle of Frances Frei and Anne Morriss (Harvard Business School), set out in their Harvard Business Review article "Begin with Trust" (2020) and their book "Unleashed" (2020).
The three tests
Why this exists. Miss any one of the three and you can name what you get instead.
A good conversation about AI that everyone takes part in and that leads to decisions you act on.
The third of those failures is the one to watch. A team can be sharp about AI and still end up with a plan nobody bought into, because the conversation that produced it was not a good one. That is the failure trust prevents, and it is why the three pillars come first.
You can tell when it's working:
- Everyone in the room contributes. Quiet voices get heard, not just confident ones.
- People can say what they think, including doubt, dissent, and "I don't know."
- Disagreement is welcome. It's where the useful information lives.
- The room leaves with shared language for things they previously couldn't name.
- Nobody leaves feeling smaller than they came in.
You can tell when it's working:
- Treats AI as a topic the team is allowed to be uncertain about, not a test of who's "ahead."
- Names worries, non-negotiables, and red lines as legitimate inputs, not blockers.
- Distinguishes what's already in place, what's been decided, what's an open question, and what's missing.
- Surfaces the tools people are already quietly using.
- Connects technical reality (stack, data, admin rights) to human reality (skills, fears, hopes).
- Is not a referendum on whether to adopt AI. It's a way to see clearly before deciding anything.
You can tell when it's working:
- Ends with a small number of decisions, each with a named owner and a date.
- Is equally explicit about what the team is choosing not to do, and why.
- Sits inside the organisation's existing governance, and is candid about where that governance needs to change to fit AI.
- Both supports and challenges people: enough safety to be candid, enough push for the decisions to hold.
- Connects to something the business cares about, such as risk, capacity, quality, or customer outcomes.
- Produces a record the team can revisit, share, and build on.
- Makes the next conversation easier, not harder.
Ready to have one with your team?
You can run a session solo to try it, then invite the team when you're ready. Nobody else needs an account.