A founder of a 32-person services firm sits in a Tuesday-morning leadership meeting. He is presenting the new client proposal he has spent the weekend on. The proposal is good; the leadership team agree. He mentions that he used a custom GPT to do the first draft and saved about three hours. His head of operations asks: “Could you show me how to set up that GPT for the proposals my team writes?”
The founder hesitates. He realises two things at once. First, he cannot show her how to set it up in the next 20 minutes; he half-built it through trial and error and could not document it cleanly. Second, his head of marketing, his head of sales, and his finance lead have not asked the same question, and three of them have not used any AI tool in the last month. He has been the firm’s AI department for nine months. The leverage he thought he was building is concentrated in him.
What the data says about leadership and AI
Only 25 percent of frontline employees report receiving sufficient AI guidance from leadership. Organisations with strong AI upskilling programmes are six times more likely to see adoption that translates to business impact. BCG’s 2025 research finds that the highest-performing organisations do not have more AI talent; they have managers who actively use AI and explicitly model usage for their teams. Talent matters less than modelling.
This is the pattern across the McKinsey State of AI 2025 data and the Anthropic Economic Index too. Leadership is the gating layer. Where leaders use AI visibly, teams follow. Where leaders do not, teams treat AI as optional.
Why founder-only AI usage rebuilds dependency
The mechanism is simple. When the team cannot use AI tools confidently, AI-related decisions and tasks default to the founder. The founder absorbs the workload. The implementation looks productive: outputs go up, drafts get done faster, the founder demonstrates capability. The dependency pattern is unchanged. The longer this runs, the more the founder is the firm’s AI department, with everything that comes with that.
The intended leverage migrates back into the founder rather than out into the team. The founder ends up doing more work than before, because they are now doing their own work plus the firm’s AI work, which they used to delegate to the team or skip entirely.
What formal training actually costs
Three-day “introduction to AI” workshops cost £3,000 to £5,000 per person and deliver generic content that does not map to the team’s actual workflows. Reforge AI Pivot for leaders is six weeks at around £1,500 per person; AI Pragmatist for operators is six weeks at around £1,200 per person. The formal route works but is expensive at scale. For a 30-person firm, the leadership team alone runs to £10,000 to £15,000 in fees plus time cost.
For most SMEs, this is a real number, and the timeline is real too. Six weeks is a long time for a busy SME leader to be in training while still doing their day job.
The AI clinic pattern that works
The pattern that works for SMEs is small, structured, and bounded. Weekly 90-minute “AI clinic” sessions where team members bring real work problems and solve them with AI tools. Each session is bounded to a real problem with a tool the team member will actually use in their job. They leave with a workflow they can deploy immediately. Cost is typically £200 to £400 per session for a facilitator.
SMEs report that a series of eight sessions covering the major roles in the firm produces durable adoption. Finance team runs a QBR simulation. Sales team uses AI for proposal drafting. Operations runs a process documentation pilot. Marketing handles a content brief. Each session is grounded in real work; the team member walks out with something they will actually use that week. After eight weeks, the leadership team has shared experience using AI on their own functions, and they can teach the next layer.
What AI-fluent leadership actually means
Not every leader an AI expert. Each leader can name which parts of their function are mechanical (candidates for AI substitution) and which require judgement. Each has tried at least one AI tool in their actual workflow and felt the time saving or quality improvement firsthand. Each understands the governance for their role: what can be delegated, what requires human review, what requires founder sign-off. Each can teach someone else on their team.
For a 10-person firm, this might mean one half-day workshop followed by four 90-minute practicum sessions. For a 50-person firm with multiple functions, it is more structured: department heads complete a focused leadership programme, then run their own eight-week adoption series within their function. The form scales; the principle holds.
The fragmentation failure mode
In a 20-person firm, perhaps the founder and two senior leaders use AI daily. Three middle managers use it occasionally. The remaining team treats AI as a curiosity. This fragmentation is more dangerous than no AI usage. The founder gets frustrated, takes on more AI work themselves to compensate, and ends up more dependent on AI tools rather than less dependent on the team.
The single most reliable predictor of team AI adoption is whether the leader of that team uses AI visibly in their own work. Buying training and not modelling usage is one of the cleaner ways to spend money without changing behaviour. The founder reading this who is the firm’s AI department has probably arrived there by accident; the move back is leadership-team-first, not team-wide.
What to do this week
List the leadership team by name. For each name, answer two questions. Have they used an AI tool in their actual workflow in the last 30 days? Could they teach someone else on their team to do the same? Honest answers, not what you wish were true.
If the answers cluster around “no” for most of the team, the next move is the AI clinic pattern, starting with the leadership team. Pick one real workflow per leader. Hire a facilitator for an eight-session series. Track what they actually deploy after each session, not what they say they will deploy. Within two months, the leadership team should be using AI confidently on their own functions; within three to four months, the team behind them should be following.
If you want a second pair of eyes on what the leadership-team gap looks like in your specific firm, book a conversation.
