Most companies today are rushing to define their AI strategy.
Last week, a Fortune 500 client asked me how to "scale AI" across their 20,000-person organisation. My answer was simple: "You can't, because your teams can't even agree on what 'scale' means."
Most companies don't have an AI problem. They have a coordination problem that AI will expose, not fix.
And I suspect many organisations won't realise the full benefits of AI. Not because the models aren't good enough, but because their companies were never designed for speed and alignment at scale in the first place.
For the past decade, I've been fortunate enough to work across different stages of the tech company lifecycle: from being a founder myself, to operating in pre-IPO environments, to working inside a large publicly-listed company.
Across all of them, I've noticed the same pattern: complexity compounds faster than headcount.
In the beginning, everyone behaves a little like a founder. Teams are small. Context is shared. People optimise for speed. Decisions are imperfect, but fast.
As companies grow, those advantages begin to erode. The shift is subtle at first. People stop solving for the company and start solving for themselves. Taking risks becomes less rewarded than avoiding mistakes. Optimising for outcomes slowly turns into optimising for safety.
This creates predictable second-order effects:
Departments become highly effective at pursuing their own goals, but not always at solving for the company as a whole.
AI doesn't remove these dynamics. If anything, it amplifies them.
In many domains today, AI has dramatically reduced the cost of execution: writing code, reviewing contracts, producing analyses, drafting presentations, answering customer support tickets, synthesising information. Tasks that once took hours now take minutes.
But this creates an interesting dynamic inside large organisations: when execution becomes cheaper, coordination becomes the constraint.
If teams lack alignment, faster execution doesn't necessarily create better outcomes. It simply means more work is produced, more quickly, in more directions.
In healthy organisations, AI compounds leverage. In unhealthy ones, it compounds noise.
People become increasingly productive at their individual tasks while the system itself remains slow:
The bottleneck shifts. It's no longer the production of work, but the organisation's ability to coordinate around what matters.
I started noticing this in my own work. AI made me dramatically faster, but it also surfaced something uncomfortable: the bottleneck was rarely producing the work itself. It was aligning stakeholders, clarifying ownership, navigating dependencies, and deciding what actually mattered.
Most companies don't really have an AI problem today. They have coordination, alignment, and incentive problems. AI simply makes those harder to ignore.
The organisations that benefit most from AI likely won't be the ones adopting the most tools. They'll be the ones that redesign how decisions, context, and execution flow through the company.
So ask yourself: if AI removed most execution bottlenecks tomorrow, would your organisation know what to do with the speed?