One week into the AI mandate, both people are still calibrating. The founder has moved on to the next item on the list. The delegate has started pulling together a view of where to begin. Neither has said out loud what they’re each privately uncertain about: what exactly was agreed, how much authority the delegate actually holds, and what “keeping me informed” will mean in practice. Both assume the other one is comfortable.
That silence is the most common reason AI mandates stall in month two, and it has nothing to do with the tools or the budget.
What is the week-one AI conversation?
The week-one AI conversation is a deliberate sitting between founder and delegate in the first seven to ten days of the mandate. Its purpose is to make explicit what a handover sentence leaves implicit: what success looks like to the founder, which decisions the delegate can take independently, where the founder stays involved, and what each side needs from the other. A setup call, not a project review.
Many AI mandates begin with something like “I’d like you to lead our AI work.” That sentence contains a job title and nothing else. It does not answer what “AI work” covers, who has final sign-off on tool spend, whether the mandate extends to the whole company or just one function, or what “keeping me informed” means in practice. The conversation this post scripts fills in what the handover sentence left out, and it does so before the first significant decision has to be made.
Why does leaving it implicit cost both of you?
Research on AI adoption points consistently to one failure mode: the leadership work around the tool, not the tool itself. MIT’s NANDA research found roughly 95% of generative AI pilots stall or show no measurable P&L impact; the cause named is a learning gap in workflow integration, not model quality. BCG reports roughly half of organisations stuck before they can scale past proof-of-concept. The technical side rarely fails first.
The failure pattern the skipped conversation creates has a name: verbal delegation but ongoing interference. The founder hands off ownership verbally, then keeps intervening on key decisions because no one drew the line between what the delegate decides and what the founder still touches. The organisation adjusts accordingly: the delegate starts checking before moving, the mandate slows to a stall, and nobody calls it what it is.
Research on reverse delegation and the dynamics of authority identifies the same thread consistently. When the terms of delegation are left vague, people in authority tend to retain informal ambiguous involvement as a hedge. A founder who never specified what “independent decisions” meant has not, in any operational sense, delegated.
What should the delegate bring to the table?
Four questions belong in the delegate’s chair: what does success look like to you in six months (the measure you will actually be glad about, not just the stated goal), what is your real motive for doing this now, which decisions will you still want to be consulted on, and where does my budget ceiling sit. Each answer shapes how the delegate moves for the next quarter.
The motive question carries more weight than it looks. A founder with an eye on exit is measuring this mandate against valuation readiness and whether the business can show it runs without them; owner dependency is consistently cited by M&A advisors as the single largest discount to an exit multiple. A founder who is operationally overstretched is measuring against capacity and workload. The delegate who misreads the motive does excellent work in the wrong direction.
Asking about the budget ceiling is about making your own decisions cleanly. A delegate who does not know whether a £5,000 tool subscription needs sign-off before or after the fact will either check everything (slow) or decide everything and learn the hard way where the limit was. Neither outcome serves the programme.
What should the founder bring to the table?
Four questions belong in the founder’s chair: what do you need from me to do this well, what will you not be able to move forward on without my input, what are you most worried I’ll get wrong from a distance, and when should we check in formally. The purpose is to identify where founder availability genuinely helps and where continued involvement would become an obstacle.
Spencer Stuart’s research on CEO engagement in AI programmes identifies visible, sustained sponsorship as one of the strongest adoption predictors. That visibility means being clear from the start about what you are watching and why, not staying involved in every decision.
The founder who says “I know I’ll find it hard to stay hands-off on the first few tool decisions; let’s agree I’ll flag to you when I’m tempted to intervene rather than just doing it” has done more for the mandate than any project plan. The delegate who hears that is equipped to manage it. The one who does not will spend month two trying to decode what the founder actually wants.
The exit motive is worth naming here too. Owner dependency is the single largest discount to an exit multiple according to M&A advisors, and Kyndryl’s 2024 research found that only 14% of organisations have aligned their workforce, technology, and growth goals. A founder serious about exit-readiness is not just handing off AI; they are rehearsing a mode of operating where decisions happen without them. The week-one conversation is where that rehearsal starts.
How do you close it so the agreement sticks?
The conversation needs a one-page record, kept by both sides, naming what success looks like, what the delegate can decide without checking, where the founder stays in the loop, and when the first formal check-in is. Two people who agreed in the same room and both hold the same page are in a materially different position than two who took separate notes and remembered it differently.
The record does not need to be formal. A shared document, a copied email, a note in a shared notebook. The medium matters less than the act of agreeing on what it says before you leave the room.
One final item belongs on the agenda: name the interference pattern before it starts. Say out loud that verbal delegation with ongoing informal override is the failure mode you are both trying to avoid. Naming it before the programme is under way is far easier than naming it after the stall has already begun and both of you are being careful with each other.
If you are the delegate and the founder sidesteps this conversation, you can ask for it directly. “Before we go further, can we set twenty minutes to agree how decisions get made?” is not overstepping. In week one, it is the most practical move available to you.
