You handed the AI work to someone you trust. They will come to the one-to-one prepared. The question is whether you will.
Founders often default to “how is it going?” in these meetings, then accept whatever comes back. The delegate delivers the version of events they have prepared, and the founder leaves no clearer than before. The problem is less about anyone’s honesty than about the questions being asked. Without the right prompts, even a good delegate defaults to a progress narrative rather than an honest account of where things actually stand.
The five questions below are designed to change that. Each one surfaces something a standard status update misses, and each has a tone that matters as much as the words.
What problem are we actually solving here, and would it matter without AI?
This question separates problem-first work from tool-first drift. If the delegate struggles to answer without leading with the tools they’re using, that’s diagnostic. Addepar runs this as a standard executive gate, asking whether the initiative would still matter if AI weren’t involved. A clear yes anchors the work in genuine business value. Hesitation points to a programme that has gathered momentum around the technology rather than the outcome.
A strong answer names a specific operational problem and connects it to a business outcome. “We’re cutting document review time from four hours to forty minutes” is a strong start. “We’re exploring how AI can improve our workflow” is not.
Asking this question early in the meeting signals that you’re tracking outcomes, not demonstrations. It also gives the delegate permission to be honest about a programme that may have drifted from its original intent. Founders who disengaged after the handoff and re-engaged only when things looked questionable are a well-documented pattern in delegated AI work. Your presence in the conversation changes what the delegate is willing to say.
If the answer is thin, stay curious rather than concerned. Ask which problem they would choose first if they had to narrow it to one. That often surfaces what the programme is actually doing faster than any prepared update.
Which indicator is moving, and what are we still waiting to see?
Dual-ROI is the honest framing for AI work that is less than a year old. Trending ROI covers the early signals, time saved, process cycles shortened, error rates down; realised ROI is the financial outcome those signals should eventually produce. Asking about both together forces the delegate into a specific, time-bounded account rather than a holding pattern of “it’s going well” with no numbers behind it.
A strong answer gives you both. “Invoice processing is down from five days to one. That’s our leading indicator. Realised ROI, we’re watching debtor days, and we expect to see a shift by Q3.” That’s specific, time-bounded, and honest about the lag.
The lag itself is worth naming. Meaningful ROI from AI work commonly takes twelve months or more to materialise, regardless of how quickly the early wins arrive. Executive expectations tend to run well ahead of programme timelines, and the delegate is often managing that gap without enough support. When a delegate describes a leading indicator moving but a lagging indicator still flat, that’s honest reporting.
If neither is moving, that is a different conversation, one about whether the right process was chosen in the first place rather than whether the delegate is performing.
What have you decided this fortnight, and what’s waiting on me?
Decision rights are often the invisible stall in delegated AI work. The delegate was handed a mandate but may still be waiting on resource sign-off, supplier approval, or a policy call that only the founder can make. This question surfaces two separate things, how much the delegate is operating with genuine autonomy, and where the founder is inadvertently holding up progress without realising it.
A strong answer on the first half tells you the delegate is actually deciding things. “We chose the vendor this week, signed the data processing agreement, and started scoping the first process with their team.” That’s motion. If the delegate has been in the same holding pattern since you last met, the question about what’s waiting on you is where the real answer usually lives.
BCG’s research on the AI adoption gap found that usage can climb while business impact stays flat, often because decision rights remain unclear. The delegate is moving but the programme is not, because the approvals sit at a level the delegate cannot reach alone.
A good outcome from this question is a specific ask. “I need a decision on the data governance policy before we go live” is actionable. “We’re working through a few things” is not.
What is the team telling you, and what are they not saying out loud?
Many AI programme setbacks don’t announce themselves. Employee resistance to new tools tends to show up as process workarounds, selective input, or simply not using the system outside the founder’s line of sight. Asking the delegate to report both what the team is saying and what they sense but can’t prove opens the adoption picture early, when there’s still time to address it.
A strong answer looks candid. “Three people in the finance team are using it every day. Two in sales have logged in twice. I have a feeling they’re worried it will surface errors to the board, but nobody’s said it outright.” That’s the kind of reporting that lets you have a useful conversation.
Employee fear around job security and the visibility of mistakes is one of the most consistent barriers to adoption across owner-managed businesses. HR Executive’s research on AI and workforce distrust shows that passive resistance, people working around tools rather than openly against them, is harder to address the later it’s left.
Korn Ferry’s work on AI readiness shows that organisations tend to assign AI leadership to strong operators rather than people with AI-specific skills, which means your delegate is often managing resistance they had no ready playbook for. Knowing the picture early helps you decide what support they actually need.
How do you hold this conversation as support rather than interrogation?
A script this direct works only if the delegate reads it as the founder staying engaged, not as a loss of confidence in the mandate. That distinction lives in the framing. Lead with curiosity before assessment. Ask what you can remove from their path before asking what they’ve done. The signal you want to send is that you’re tracking the work because it matters, not because you’re preparing to step back in.
Spencer Stuart’s research on AI leadership recommends that CEOs become power users rather than delegators, because it changes how the delegate reads the conversation. A founder who has tried the tools arrives with genuine curiosity. A founder who hasn’t is asking from a remove the delegate can’t quite read, which makes honest reporting harder.
The conversation runs well when you arrive with a genuine interest in outcomes, a few things you can clear from the delegate’s path, and an opening question about their experience rather than the programme’s progress. “What’s been the hardest part of this in the last two weeks?” asked first changes the register of everything that follows.
Use the one-to-one to signal to the person who took the mandate that you are still behind them.
