You already know who you want. A COO, a head of operations, someone sharp who follows the space and asks good questions in meetings. You are going to hand them the AI brief. The question is what that brief actually contains.
In many founder-led businesses, the handover conversation is about thirty seconds long. Get AI into the business. Make it work for us. Keep me in the loop. That is a wish, not a mandate. The delegate walks away with no clarity on what success looks like in twelve months, no real authority over budget, and no map of which decisions are theirs to make.
Everything gets harder from that point. Decisions come back up. Progress slows. You find yourself in the middle of an AI programme you thought you had handed off.
What is an AI mandate?
An AI mandate is the written brief a founder gives when handing the AI programme to a senior operator. It states what success looks like in plain terms, not which tools to buy or which processes to automate. A mandate defines the outcome, names the decisions the delegate can make independently, and attaches real budget authority. Without those elements, it is an instruction, not a programme with genuine ownership.
The distinction matters because “get AI into the business” says nothing about what better looks like in six or twelve months. BCG’s 2025 research found roughly half of companies stuck in stagnating or emerging stages, unable to move past proof-of-concept. The common constraint is the absence of a defined programme with clear ownership, not a failure of the technology itself.
A mandate is not a strategy document. You are not handing the delegate a roadmap; you are handing them outcome-focused boundaries within which they can build one. The founder’s job is to define the edges. The delegate’s job is to fill the space.
Why does a vague AI brief set your delegate up to fail?
When a mandate lacks specifics, the delegate inherits the gap. They spend the first weeks guessing at success criteria and avoiding decisions that look expensive because they are not certain those decisions are theirs. Kyndryl’s 2024 research found that only 14% of organisations have aligned their workforce, technology and growth goals. The briefing layer is usually where that alignment fails first.
The failure pattern that shows up in practice is too little action, not the wrong action. The delegate checks back on decisions the founder actually wanted them to make. Research on delegation dynamics consistently shows that verbal handovers, without a corresponding written brief, produce what researchers call reverse delegation, where the delegate returns decisions upward rather than absorbing them. For an AI programme, that means weeks of momentum lost to check-in meetings.
MIT NANDA found that only around 5% of generative AI pilots achieve rapid revenue impact. The cause is what researchers call a learning gap in workflow integration, the practical work of embedding AI into how the team actually operates. That embedding work needs a delegate with the clarity and authority to act without constant approval. A vague mandate removes both.
Where does the gap between a wish and a mandate show up?
You feel the gap at the first significant decision, not in week one. The delegate wants to restructure a workflow, bring in a tool that costs real money, or redirect two team members’ time. They ask you. You answer. Then it happens again, and six months in you are still the decision point for a programme you thought you had handed off.
This is the pull-in problem. The delegate, uncertain of their authority, checks back on decisions they should be making independently. The programme does not fail dramatically. It slows to a pace where the founder is still the critical path, just informally rather than by design.
There is a second version worth naming. The founder says the brief has been given, then keeps weighing in on AI decisions, changing direction after the delegate has started moving, or overriding calls the delegate had the authority to make. The team learns quickly that AI decisions flow through the founder regardless. The delegate’s authority is gone before it has done any real work.
What does a well-written mandate actually contain?
A written AI mandate has four components. The first is the outcome, stated in plain terms the delegate can actually measure. The second is a list of decisions they can make without coming back to you. The third is the budget they control outright, without a finance approval at each step. The fourth is the short list of decisions that stay with you.
The outcome framing matters because it tells the delegate what they are optimising for, not just what they are doing. “Reduce the time I spend resolving operational decisions by half within eighteen months” is an outcome. “Implement AI tools across key departments” is a to-do list. The outcome framing also gives you both an honest measurement point at review. Deployed software in four departments is easy to claim. Genuine reduction in founder involvement is not.
Decision boundaries are the element founders find hardest to write, because naming them explicitly requires acknowledging how far you have been in the middle of things. One useful approach is to go through the kinds of decisions the delegate will face and ask honestly whether you want to be consulted in advance or informed after the fact. Consultation belongs on their escalation list. Notification belongs in a weekly update.
Budget authority needs to be real. “Come to me for anything over five thousand pounds” sounds reasonable until your delegate needs to move at pace and every tool evaluation becomes a conversation.
What should sit alongside the mandate?
The mandate is the written brief, but two things need to accompany it. The first is a short conversation about what the AI programme is actually for. The goal is reducing how much the business depends on you personally. Without that framing, the work tends to get built around your preferences rather than a process that holds without you at the centre.
The second is a clear distinction between where you want strategic input and where you are willing to stay out. Founders often want to stay involved in AI decisions without becoming a bottleneck. That is workable, but the mandate needs to name which decisions merit your involvement. Exit-relevant positioning, anything that changes a client relationship you own personally, major spend above the agreed threshold. Everything else sits with the delegate.
The counterintuitive risk worth naming here is what researchers call the dependency paradox. If the AI programme is built to mirror the founder’s instincts and historical decision patterns, the business can become more dependent on the founder, not less. M&A advisers consistently identify founder-centricity as a significant discount to exit multiples; research suggests buyer discounts of 30 to 40% when operations remain founder-centric rather than systematised. The mandate’s outcome should specify a process that works without the founder at the decision point.
Spencer Stuart’s work on AI leadership notes that the founder’s role is to create the conditions where the delegate can run the programme with the founder’s confidence behind them, not their daily involvement.
Writing a mandate takes an hour, maybe two. It saves months of pull-in, weeks of re-litigation, and the frustration of a delegate who has learned to ask for permission rather than act. The brief is yours to write. The programme is theirs to run. If you cannot articulate that difference in a single written document, the handover has not happened yet.
If you want to think through what yours should say, book a conversation.
