What Happens When AI Stops Being the Bottleneck?

Much of the current conversation around AI has shifted towards agents.

It's easy to understand why. The first wave of generative AI showed us what large language models were capable of. The second embedded AI into the tools we use every day through copilots and assistants. Agents feel like something different. For the first time, organisations can see AI moving beyond productivity and beginning to execute work. It feels like the point where AI starts to move from experimentation towards genuine enterprise adoption.

That makes this an important moment. But I'm increasingly wondering whether we're asking the wrong question.

Much of the discussion today focuses on the capability of the technology. Which models will win, how agents will be integrated into enterprise workflows, and which platforms will emerge as the enterprise standard. All important questions, and all logical places to start. However, I suspect the more interesting question is whether agent capability will ultimately determine where enterprise value is created.

History suggests it won't.

Every major technology category eventually reaches the point where the technology becomes good enough. At that moment, value rarely continues to accrue to the technology itself. Instead, it begins to move towards those that enable enterprise adoption. Sometimes that's the organisations able to transform themselves fastest. Increasingly, it's also the businesses that help others navigate that transformation. Cloud computing was never ultimately constrained by infrastructure. Data programmes rarely failed because databases weren't capable enough. The harder challenge was changing the organisation around the technology.

I'm increasingly wondering whether AI is approaching that same inflection point.

This is where the conversation becomes less about AI, and more about organisational change.

As models continue to improve, the technical barriers will inevitably reduce. Agents will become more capable, more reliable and easier to deploy. The harder challenge will become organisational rather than technical. Not whether an agent can perform a task, but whether the organisation around that task is designed in a way that allows AI to create value.

Most organisations weren't built for autonomous systems. Processes have evolved over years rather than being intentionally designed. Critical decisions still rely on experience rather than explicit rules. Governance has developed incrementally, ownership is fragmented, and many workflows depend on tacit knowledge that has never been embedded into systems. Introducing agents into that environment doesn't remove the complexity. It simply makes it visible.

This is one of the reasons I find the rise of Forward Deployed Engineers so interesting.

The obvious explanation is that enterprise AI is technically difficult to deploy, requiring highly skilled engineers to work directly with customers. I wonder whether something more significant is happening. Their role increasingly extends well beyond implementation. The most effective FDEs don't simply connect models to enterprise systems; they help customers understand where AI can genuinely create value and what needs to change for that value to be realised.

Because the work isn't simply connecting models to enterprise systems. It's understanding how decisions are made, where human judgement genuinely adds value, which processes should change, how governance needs to evolve and where organisational friction will prevent technology delivering meaningful outcomes. That's as much an operating challenge as it is a technical one.

Viewed through this lens, the long-term competitive advantage may not come from building the most capable agents alone. Models will improve. Frameworks will mature. Technical capability will become increasingly accessible. The businesses that create the greatest value may be those that combine AI capability with a deep understanding of how enterprises adopt change. Helping customers redesign operating models may become just as important as building the technology itself.

What makes this particularly interesting is that it changes how we think about the next phase of AI.

Today, much of the investment is understandably flowing towards models, infrastructure and agent frameworks. Over time, I suspect an increasing proportion of enterprise value will be created by organisations that understand how businesses change. Those that can bridge technology, process, governance and commercial outcomes. Those that help customers move beyond successful demonstrations and into sustainable transformation.

Which leads to the question I find most interesting.

Will enterprise AI be won by the companies building the best agents?

Or by the companies that become best at helping organisations transform around them?

I suspect the answer is both.

The companies that create the greatest enterprise value won't simply build more capable AI. They'll combine technical excellence with a deep understanding of how organisations change. They'll recognise that deploying agents is only the beginning. Helping enterprises redesign the way they operate around them is where the real transformation begins.

Because agents will undoubtedly change how work gets done.

The bigger opportunity may lie in helping organisations change alongside them.