We are treating a structural change like a software subscription.
We are watching the emergence of tools that can research, write, analyse, code, design, communicate, interrogate data, automate workflows and increasingly carry out complex pieces of work with far less human intervention than would have seemed realistic only a few years ago.
Yet there are still companies treating AI as if it were an interesting productivity app.
They buy a few subscriptions. Someone uses ChatGPT to rewrite an email. A marketing team generates some content. Perhaps IT runs an “AI initiative.”
Then everyone goes back to working roughly the same way they did before.
I think that fundamentally misunderstands what is happening.
AI is not simply another piece of software to add to the existing technology stack. It has the potential to change the architecture of a business itself: how information moves, how decisions are made, how knowledge is stored, how work is executed and ultimately how many people and how much time are required to produce an outcome.
That creates an enormous opportunity. It also creates a very real threat.
The unit of change is the workflow.
Because the company that properly integrates AI is not merely saving an employee 20 minutes writing an email. It can start redesigning entire workflows.
A process that previously required five people, six systems, several meetings and three days of waiting may increasingly become a coordinated workflow involving software, AI agents and a much smaller amount of human supervision.
- Research that once took a week may take an afternoon.
- Thousands of documents can be analysed rather than sampled.
- Customer enquiries can be understood, classified and acted upon immediately.
- Software can increasingly be created, tested and modified at extraordinary speed.
- Knowledge buried in inboxes and forgotten folders can become usable.
And that is only the beginning.
There are different levels of AI adoption. Giving employees access to AI tools is useful, but it is probably the least interesting level. Next comes connecting models to company information, software and processes so they can actually participate in the work.
Beyond that: companies building their own systems around their own processes, data and intellectual property. At that point, AI stops being a tool employees occasionally use and starts becoming part of the operating infrastructure of the organisation.
The model is not the moat.
The competitive advantage will not necessarily belong to the company with the best subscription to the latest model. Everyone can buy access to broadly similar models.
The advantage will belong to the companies that understand how to combine those models with proprietary data, domain expertise, carefully designed processes, software infrastructure and human judgement.
That is much harder to copy.
And this is why I think some businesses are underestimating the urgency. The gap may compound.
Waiting does not keep you in the same place.
A company that starts seriously adopting AI today does not just become more productive today. It starts learning.
It discovers which workflows can be automated. It improves its data. It builds internal expertise. It creates integrations. It learns where AI fails. It redesigns roles. It develops governance. Its employees become better at working alongside these systems.
A company that waits does none of those things. So when it finally decides that AI matters, it is not starting six months behind. It may be several organisational learning cycles behind.
That is the part I think many leaders are missing.
There are completely legitimate reasons to be cautious. AI can make mistakes. Sensitive information has to be protected. Systems need security, oversight and accountability. Not every process should be automated, and automating a badly designed process simply creates a faster badly designed process.
Caution and inactivity are not the same thing.
The responsible response to a powerful new technology is not to ignore it. It is to understand it deeply enough to use it intelligently.
The question leaders should be asking.
Your competitors have access to the same technology. Some are experimenting aggressively. Some are rebuilding workflows around it. Some are creating AI systems connected directly to their operational data.
They are already asking a very different question from the companies debating whether employees should be allowed to use ChatGPT.
“If we designed this company today, knowing what AI can now do, would we build it the same way?”
For many organisations, the answer will eventually be no.
And that, to me, is the real significance of AI. It is not about replacing every human being. It is not about believing every piece of hype. And it is not about putting an AI chatbot on a website and calling yourself innovative.
It is about recognising that the cost, speed and capability of knowledge work are beginning to change. When one of the fundamental inputs into running a business changes, businesses eventually change around it.
The companies that understand that early will experiment, learn and adapt. The companies that dismiss it will continue operating normally — perhaps for longer than people expect.
Until one day they find themselves competing with organisations that can move faster, understand more, automate more and operate at a fundamentally different cost structure.
By then, “we should probably start looking at AI” may be far too late.
