It usually arrives a few weeks in, late at night, after the diary is finally quiet. If this goes wrong, it is on me. You did not pick the technology. You did not set the timeline or sign off the budget. The founder handed it to you because you are the one who makes things happen, and now your name is on the initiative and the board is watching. That sits heavily, and the role carries it alone.
This is the worry that has run underneath the whole series. The first month, the strategy work, the sixty-day plan, the first board update, all of it was about getting the work right. This piece is about getting you through it. You have accountability without full control, and the way through is to name the odds early and own the story, so that when the setbacks come, you are reporting a risk you flagged, not confessing a surprise.
What is actually on the line when you take this on?
Four things are exposed at once. Your professional credibility, as employee doubt hardens into leadership scepticism the moment a pilot stalls. Your standing with the board, since you speak for a topic you may not fully command. Your relationship with the founder, who handed it off and may re-engage hard if results lag. And the organisation around you, where resistance can starve a pilot of goodwill.
That concentration is the real problem, not any single risk. The founder kept control of the resources and the decisions while the accountability moved to you. Paro calls this the delegation trap, and Camille Esq describes the common sequence well. A founder hands AI off because it feels technical, steps back, then re-engages with course-correction demands the moment the numbers disappoint. You are left satisfying a board that wants operational gains and a founder whose real motive is often a cleaner business to sell.
The scale of board attention makes this sharper. Harvard Law School’s review of S&P 500 disclosures found reputational risk named as the leading AI concern by 38 per cent of large listed companies. Those are big firms, but the signal travels down. Boards everywhere are nervous about AI, and a nervous board looks for a person when things go wrong. The named delegate is the obvious candidate, which is why this exposure is worth taking seriously rather than waving away.
Why does the named person become the natural scapegoat?
Because the odds are genuinely against the pilot, and the named person is the easiest place to put the blame. MIT’s research, through Project NANDA, found that around 95 per cent of AI pilots fail to show a measurable P&L impact. That is the load-bearing fact of this role. When something with those odds stalls, a board under pressure looks at the name on the slide.
The failure is usually not yours. MIT and others describe a pilot-to-scale valley where initiatives die for reasons well outside the delegate’s control. Poor data quality, an over-eager vendor, a process that was broken before AI touched it, a workforce feeding bad inputs to prove the tool inferior. You can run a textbook process and still land in the 95 per cent, because the things that sink a pilot are commonly upstream of anything a single lead can fix.
There is real fear sitting behind this, and it is worth naming plainly. Surveys of people in this position find a clear majority worried about their own job if they fail to lead adoption, and a large share of executives fearful of AI-related job loss in the year ahead. That fear is rational. It is also what makes people over-promise, because a confident timeline feels safer in the room than an honest one. The over-promise is what turns a known risk into a personal failure later.
How does naming the failure rate actually protect you?
It converts a future surprise into a managed expectation, which is the single most valuable move available to you. Tell the board on day one that around 95 per cent of AI pilots show no measurable P&L impact, and you have done two things. You have shown you understand the terrain better than the hype suggested, and you have made any later setback something you predicted rather than something you confess.
Honesty about the odds is read as competence, not weakness. A board does not want a cheerleader for a technology that fails nineteen times out of twenty. It wants someone who can tell the difference between a sensible bet and a sure thing, and who will say so. When you set the realistic horizon early, that meaningful return commonly takes twelve to twenty-four months rather than a quarter, you are buying yourself the room to do the work properly. You are also denying the board the surprise it would otherwise hold against you.
The language you choose matters here. Successful delegates talk about iterating based on learnings rather than failing, and they report trending ROI, the early indicators, alongside realised ROI as the financial outcomes arrive. This is not spin. A pilot that produces a clear decision to stop is a successful pilot in process terms, because it bought a real answer cheaply. Framing the work that way, honestly and in advance, is how you stay credible across the months where nothing has paid back yet.
What should you commit to instead of a result?
Commit to the things that sit inside your control. You cannot promise a return, because the data, the vendor, the founder’s patience and the workforce’s goodwill are not all yours to command. You can promise readiness, that the data, skills and governance are in place before money goes in. Learning, that every pilot produces a clear decision whether or not it pays. And decision points where the board chooses to scale, pause or stop.
Those three are defensible because you can actually deliver them. Readiness is a checklist you can complete. Learning is a debrief you can run regardless of outcome. A decision point is a date in the diary where you hand the board a real choice with real evidence, which moves part of the accountability back to where the resources sit. A board that has agreed the decision points cannot later claim it was surprised by a pause, because it built the pause into the plan with you.
Pair this with a dual-ROI frame and you can report progress without claiming returns that have not arrived. Trending ROI shows the board early movement, adoption climbing, hours coming back, error rates falling. Realised ROI shows the financial outcome as it lands over the longer horizon. Reporting both, and being clear which is which, is the difference between looking like you are stalling and looking like you are on track.
How do you carry this without it carrying you?
You hold both truths at once. The exposure is real, and it is manageable. The role does concentrate professional, board, founder and organisational risk onto your name, and the odds on any single pilot are poor. None of that lands as a personal failure if you own the narrative early, name the failure rate before anyone else, and commit to readiness, learning and decision points rather than to outcomes you do not control.
Look back across what this role has actually built. An honest read of where the business stood. A short, ranked plan the board could fund. A first update that told the truth about timelines. None of that depended on a pilot succeeding, which is the point. You have been building the thing that survives a stalled pilot, a credible process run by a credible person, worth far more than a lucky early win would have been.
The founder could not fully hand this off, and that has been the lesson all along. The resources and the final decisions stayed with them, so the accountability was always shared even when it did not feel that way. Your job was never to guarantee the outcome. It was to give the business a clear-eyed way to make good AI decisions, and to make sure the risk was named, shared and understood. Do that, and whatever the pilots do, you come out of this with your judgement intact. If you want a second pair of eyes on how you are framing it, book a conversation.
