How much ambition can a company absorb?
On this page
Last week I spoke at SportsPro AI + Tech in London about a question I have been thinking about increasingly often: how much ambition can a company absorb before complexity starts growing faster than capability?
The question came from something very practical. Every new sport, market, channel, product or customer experience creates opportunity on the outside of a company, but also more coordination underneath it. More systems, more decisions, more handoffs, more context to reconcile.
That is not necessarily a technology problem. It is an operating-system problem.
For a long time, companies were designed around two very real constraints: information was hard to gather and coordination was expensive. We built meetings, reporting layers, workflows, handoffs and reconciliation processes partly to compensate for that.
Those choices were rational. But AI is starting to change the constraint.
Systems can increasingly gather authorised context, maintain it over time, reason across it and act within defined boundaries. The interesting question is therefore not simply how much existing work we can automate. It is which parts of the organisation we would design differently if we were starting today.
The visible layer is becoming cheap
One of the frameworks I shared in London was Truth → Structure → Expression.
Truth is what the company knows: data, context, history and decisions. Structure is how the company works: rules, workflows, authority, permissions and logic. Expression is where all of that becomes visible: products, interfaces, experiences and agents.
AI is making expression dramatically easier to create. That makes the quality of the layers underneath more important, not less.
A polished interface cannot rescue weak truth. An intelligent agent cannot compensate for authority the business has never defined. This is why I increasingly think many companies do not really have an AI problem. They have a foundations problem that AI is making harder to ignore.
I have written before about the importance of foundations and why the best technology is often invisible until it matters. This feels like the next extension of that idea: as expression gets cheaper, the quality of truth and structure becomes more consequential, not less.
The economics of ambition are changing
The part of this that interests me most is not productivity in the narrow sense.
It is what becomes reasonable to attempt.
At On, one example was our cloud migration. A programme that would traditionally have been long-running and partner-heavy became something a small internal team could take much greater ownership of using agents. The interesting outcome was not simply that individual tasks became faster. The size of problem the team could reasonably take on changed.
That is a different form of productivity.
I have been writing for a while about work leaving something behind and about capacity becoming less linear. This is the same idea showing up at a larger scale: the best use of AI is not necessarily to save a few hours, but to change the class of problems a team can economically take on.
The same shift appears elsewhere. In design, for example, when producing possibilities becomes much cheaper, the scarce skill moves away from operating the tool and towards judgement and taste.
As execution gets cheaper, knowing what deserves to exist becomes more valuable.
That connects to another argument I have made before: AI raises the price of judgement. The tools may make production easier, but they do not remove the need to decide what is worth building, buying, automating or owning.
Simpler underneath, more ambitious at the edge
Successful companies naturally become more complex on the outside.
More products, more sports, more markets, more channels, more athletes, more consumer segments and more ways of experiencing the brand. That external complexity is often the point. It represents ambition.
The problem is when companies mirror that complexity internally.
A new market produces another process. A new capability produces another platform. A new team creates another version of the data. A new requirement creates another vendor. Almost every individual choice can be reasonable, while the organisation as a whole becomes harder to change.
The principle I keep coming back to is simple:
A company should be able to become more complex to the outside world while becoming simpler underneath.
Football should feel different from running. China may feel different from Europe. A flagship store should not feel like an e-commerce site. Those differences are valuable.
But underneath, we should not need a separate technological universe for every expression of the company. The foundations should become more coherent as the possibilities at the edge expand.
More possibilities at the edge. Less duplication underneath.
The real test
For years, we have measured scale by asking whether our systems can handle more transactions, more customers, more products and more countries. Those measures still matter.
I am increasingly interested in another question: can the organisation attempt more without complexity growing at the same rate?
That is why the closing question from the talk was deliberately simple:
Did doing the last thing make the next thing easier?
If the answer is consistently yes, capability compounds.
The last migration should make the next migration easier. The last major launch should leave reusable capability behind. The last difficult decision should leave better context for the next one. The last incident should improve the system rather than simply disappear into history.
I think that is a better definition of an AI-native company than the number of agents it has deployed.
Not a company with more AI, but a company that becomes capable of more.
I shared this argument at SportsPro AI + Tech in London. A few people asked for the slides, so I turned the presentation into a short external guide with one-page explanations for each slide.