Three months ago you put “get my head around AI” on the list. The fortnight after the quarter close, or after the hiring round, or once things settled. The calendar had other ideas. The task went to a good operator. The mandate got delegated. And the question of your own hands-on engagement keeps getting pushed to a date that keeps moving.
That is the pattern. Later has no natural trigger to arrive.
What is the “I’ll do it later” problem for founders?
Founders run close to the edge of their capacity by default. The delegation reflex is a survival mechanism, not a character flaw. When something new and technical arrives, the reasonable response is to hand it to someone who can go deep on it while you hold the rest of the business together. The problem with AI is that this same reflex files it in the wrong category.
Research on why non-technical founders struggle with new technology shows a consistent pattern. AI gets classified under IT, data, or digital, territory that already belongs to a COO or a head of technology. Handing it to them feels natural, even responsible. You put a capable person in charge, the board sees that someone credible owns it, and the question of your own involvement gets pushed downstream.
The categorisation error runs deeper than job titles. Conversational AI is less like learning Excel and more like changing how you approach problems. That shift requires a behavioural change, not a procedural one. You cannot delegate the behavioural part. Your operator can build the systems, run the training, and manage the rollout. The question of how you think about the work is yours.
Later, in this framing, has no natural trigger to arrive. The calendar keeps filling, and the tab stays open.
Why does your personal engagement matter more than you might expect?
The strongest predictor of AI adoption in owner-managed businesses is visible, sustained executive sponsorship. Change management research has established this pattern consistently. Technology rarely fails on technical merits. The people and leadership work is almost always what gets underestimated. When you are not personally using the tools, you send a signal to every person in your business that AI is peripheral to how the business actually runs.
BCG research found roughly half of companies are stuck in stagnating or emerging stages, unable to scale past proof of concept. One pattern the stronger performers share is embedding AI in multiple workflows and in strategic planning, rather than treating it as a standalone IT project. That embedding happens when leaders use the tools, not when they commission someone else to use them.
BrainStorm, a training firm that tracks Copilot rollout data across client organisations, found that businesses reaching 50% activation within 90 days consistently showed active C-suite sponsorship alongside role-specific workflow guidance and ongoing leadership communication. Those without it averaged 28% despite similar licence deployment. That is vendor-reported data and worth treating with some caution, but it aligns with the broader change management evidence.
Your operator can run a good programme. The question is whether that programme runs at full power, or at the power it has when you are visibly behind it.
Where does the deferral actually cost you?
The deferral shows up in a specific place before it shows up anywhere obvious. A founder who has handed over the AI mandate but has not used the tools personally tends to stay involved in decisions about AI without ever quite committing to the work. The programme gets verbal endorsement, which is valuable. What it does not get is the signal that the founder has skin in the game.
The subtler version is what happens to your operator. When the founder has delegated AI but keeps weighing in on individual decisions, the organisation learns a clear lesson. AI is still the founder’s call when it matters. That outcome does not reflect badly on getting involved. It describes what happens when involvement is not grounded in direct use. Your team cannot easily distinguish a founder who is engaged because they understand the territory from one who is engaged because they are anxious about it.
The cost is rarely catastrophic. AI programmes seldom collapse because the founder never used the tools. They stall. They underdeliver. Kyndryl research found that 70% of leaders report their workforce is not ready for AI, and only 14% have aligned their workforce, technology, and growth goals. The alignment gap almost always includes a senior leader who is nominally behind the work but not genuinely practising it.
Does the research say founders need to be power users?
The honest answer is that the evidence is more nuanced than the search-firm playbooks suggest. Visible executive sponsorship is well-supported in the change management research. Staying genuinely behind the programme, using your voice and your calendar to signal that this is how the business works now, makes a measurable difference to adoption. The stronger version of the argument, that the CEO must personally become a hands-on power user, is influential but less formally evidenced.
Spencer Stuart’s playbook for CEOs argues for a 90-day agenda in which the founder actively automates parts of their own work before any company-wide push. Russell Reynolds frames it as “taking responsibility”, a distinction from becoming a deep practitioner. Both firms are pressing in the same direction, but from slightly different positions.
The useful distinction is between delegation and abdication. Delegation means your operator has real authority, clear scope, and your genuine backing. Abdication means you have handed it over and stepped back from any personal engagement with the work itself. The research on AI adoption consistently finds that abdication correlates with the programmes that stall. It cannot tell you whether power-user-level engagement is the variable, or whether the correlation is explained by the sponsorship pattern underneath it.
Your operator needs your sponsorship. Whether they need you to be an expert is a separate question.
What does starting actually look like?
Spencer Stuart’s sandbox for founders is specific and worth taking at face value. Automate one daily task you personally do, before any company-wide rollout, and before you announce anything to the team. The goal is genuine first-hand experience, so your engagement with the broader programme is grounded in something real, not in what your operator has briefed you about what real feels like.
The fortnight framing is deliberate. Pick a task this week, run it with an AI tool for two weeks, and see what you notice. A meeting prep you currently do by re-reading notes. A weekly summary you produce from a set of inputs. An email thread you draft through several iterations. Any of these can be handed to an AI tool in a working session today.
You do not need to tell anyone. The value at this stage is personal. You get a sense of where the tool is confident, where it slips, and what it needs from you to produce something worth keeping. That experience changes the quality of every conversation you then have with your operator about where the programme should go next.
One task, this fortnight. The rest follows more naturally from that than from any planning session.
If you’d like to work through what that first task should be, and how to build from it without another planning session that never quite becomes action, a conversation is the place to start. Book a conversation and we’ll find the right entry point for where you are.
