You’re weighing an AI programme that would make your week lighter. The meeting prep, the first drafts, the client reports that take two hours you haven’t got. On the face of it, there’s nothing to question.
Except there is a question beneath it that many founders have not asked. Are you building a business that runs better with you in it, or a business that runs without you? The two pull in opposite directions. An AI programme scoped for one goal will often work against the other.
What is the fork between building to sell and building to run?
A run-it-well business has the founder at the centre; client relationships, key decisions, and institutional knowledge all route through them because that is what makes the operation work. A sell-ready business has each of those functions moved into a system, a person, or a documented process that does not depend on the founder being present. Similar from the outside. Different at the level of architecture.
The difference becomes visible in a single question from John Warrillow’s Built to Sell. Would the customer remain if the founder left? In a run-it-well business, the honest answer for the top two or three clients is probably not. Those clients chose the firm partly because of the founder, they communicate with the founder, and their confidence in the service rests on that relationship continuing. That is entirely rational as a business model when the founder intends to stay. It is a real problem when the founder is trying to sell, because a buyer mapping customer-defection risk will find those clients sitting directly beneath the founder’s departure date.
Why does this choice affect what your business is worth?
The valuation gap between a founder-dependent business and a founder-independent one is not theoretical. Founder-dependent services businesses in the lower-middle market typically achieve 3 to 4 times EBITDA on exit. Their founder-independent equivalents in the same sector and revenue band achieve 7 to 8 times. The difference is pure risk penalty; the buyer is pricing the probability that revenue, relationships, and capability do not survive the founder’s departure.
Applied to a £3 million EBITDA business, the gap between 4 times and 7 times equates to approximately £9 million in enterprise value. That figure comes not from theoretical discounting but from observed market behaviour across thousands of lower-middle market transactions. William Buck, a chartered accountancy firm active in private equity-backed M&A, places the formal key-person discount at 10 to 25 percent of enterprise value, applied as a direct adjustment to the acquisition price. UK analysis puts the exit value reduction for owner-managed businesses in the £3 million to £30 million revenue band at 20 to 40 percent below what a systematised equivalent achieves. The mechanism is the same in each case. The buyer is acquiring a set of customer relationships, a decision-making capability, and an institutional knowledge base. Where all three are concentrated in the founder, the buyer is also acquiring the risk of losing all three.
Where does an AI programme reveal which business you’re building?
Three places, and each one tracks directly to what acquirers examine in due diligence. Client relationships are the first. Decision-making authority is the second. The location of institutional knowledge is the third. In each case, the question is whether the AI is making the founder more central to that function or whether it is moving the function to something that survives the founder’s departure.
If an AI programme makes the founder better at staying close to clients, generating more personalised communication, or managing more accounts directly, it makes the founder more central. The buyer’s due diligence team will ask which clients are at risk of defecting if the founder leaves. A founder with a more efficient AI-assisted relationship management system is still the relationship.
On decision-making, acquirers map what requires founder approval by examining contracts, pricing sign-offs, and what stops moving when the founder is unavailable for two weeks. An AI that routes better data to the founder for faster decisions has not changed the dependency. The founder remains the decision-maker.
The knowledge problem is the subtlest of the three. If the AI stack is trained on the founder’s judgement, their communication style, and their historical decisions, the business now runs on a model that only the founder fully understands. The knowledge has moved from the founder’s head to a tool that only the founder knows how to use or trust.
When does a run-it-well AI stack create a new dependency?
The new-dependency trap is specific. A founder who builds an AI stack to make themselves faster and more effective has done something genuinely useful for today’s operation. But if that stack is not designed to be used, maintained, and trusted by the team without the founder’s involvement, it has replaced one bottleneck with another. The only person who understands how the tools work is still the founder.
McKinsey’s 2025 research on AI adoption patterns shows that capability tends to concentrate in individuals rather than distributing across organisations unless implementation is specifically designed to transfer ownership to the team. In practice, this means that a founder who builds and operates a capable AI stack but does not build the team’s ability to run it independently has achieved efficiency for themselves without reducing founder dependency.
A buyer’s due diligence does not ask whether the business uses AI. It asks whether the business makes decisions, holds customer relationships, and executes its processes in a way that survives the founder’s departure. An AI programme that makes the first question look better while leaving the second unchanged has not solved the problem.
What to ask before you scope the programme
The choice between building to sell and building to run is not a moral one. Some founders intend to run their business indefinitely, and optimising it around their own effectiveness is entirely rational. The question is whether the choice has been made deliberately. An AI programme scoped without answering it will typically reflect the founder’s present preferences rather than their longer-term goals.
The Axial Dead Deal Report analysed 75 failed transactions in 2025 and found that non-QoE diligence findings, typically customer concentration linked to the founder’s personal relationships, were the leading cause of broken letters of intent, appearing in more than one in four failed deals. An exit-ready founder encounters none of this as a surprise. The founder who has not decided which business they are building typically discovers the answer during due diligence, at the point where it is expensive to change.
Before briefing the delegate who will run the AI programme, ask this question. Is the goal to make the founder better at doing what they do, or to build the business’s ability to do it without them? The first brief produces a run-it-well stack. The second produces a sell-ready one. Both are legitimate. Confusing them is expensive.
