About AI — Before You Spend a Penny

I’ve spent a lot of my life building, fixing and turning around all sorts of businesses.  These include startups, my own small and medium businesses, investor-backed growth companies and multi-site operations with complex cost structures and boards expecting answers.

In all of that time, I have watched businesses make expensive mistakes in remarkably consistent ways. They move before they are clear. They commit before they have defined what success looks like. They buy a solution before they have properly named the problem.

AI is not different. It is just faster, more expensive when it goes wrong, and at the moment, it is surrounded by enough noise that good people are making bad decisions under pressure.

I want to talk about that not from a technology perspective (I will leave that to my co-founder Malcolm, who has been working in AI since the early 1990s) but from the perspective of someone who has sat inside organisations managing cost, capacity, and consequence. Because that is where most of the AI conversation is failing the people who most need it.

The environment you are operating in right now

So what is really happening at the moment. 

Since April 2025, employing people in the UK has become materially more expensive. Employer National Insurance rose from 13.8% to 15%. The threshold at which you start paying it dropped from £9,100 to £5,000 per employee. The National Living Wage increased 6.7% — twice the rate of inflation. It rises again in April 2026.

The Employment Rights Act 2025 became law in December. Its obligations roll out through 2026 and 2027. Every new hire is now a more significant legal and financial commitment than it was two years ago.

And now the Middle East conflict has reintroduced energy and supply chain volatility that most businesses had stopped planning for.

I am not making a political argument. I am describing the operating environment as it actually is and in that environment, the cost of a bad decision, any bad decision, including an AI decision, is higher than it has been in a long time.

This is not the moment to buy a tool because a vendor demo was impressive or because a competitor appears to be moving.  This is the time to be precise and clear about what YOU need.

What most businesses actually need from AI

Here is the first question I would ask before anything else: where in your business is capacity being consumed by work that does not require human judgement?

Don’t ask where AI could be interesting or where you could experiment to produce better outcomes. You need to ask where, specifically, is your most expensive resource, the time of your senior people, being used on tasks that do not actually need them?

In almost every business I have worked with, the answer is the same: Reporting. Chasing. Summarising. Coordinating. Formatting. Updating. This is the layer that sits around the real work and consumes somewhere between a third and half of a senior person's week.

That work does not require human judgement. It requires human time and human time is now your most expensive input.

This is where AI earns its place. Not as a replacement for your people, and not as a system that creates more complexity than it solves. When it is deployed properly, against specific tasks with clear boundaries, it takes the administrative weight off the people who are already there so they can spend their time on the work that actually needs them.

Done correctly, this does not require a large investment, a long implementation, or a restructuring of your workforce. It requires clarity about exactly which tasks you are targeting, what the boundary conditions are, and how you will measure whether it is working.

That clarity is not a bureaucratic hurdle. It is the difference between AI that quietly earns its place in your business and AI that sits on a shelf costing money while nobody quite knows what to do with it.

What AI can actually do well in this environment

Based on what I have seen work, not in theory but in practice, there are three areas where AI delivers genuine value for businesses under cost pressure right now.

Giving back capacity that already exists.

Your best people are spending significant time on work that does not need them. AI deployed against specific, defined tasks returns that capacity without a hire. That is growth without cost growth.

Making existing customer relationships more productive.

Acquiring a new customer in this environment is expensive. Serving an existing one better and faster is not. AI deployed to reduce friction and improve response times within your existing base works quietly in the background. Retention improves. Referrals follow. Revenue grows without the cost that used to come with it.

Protecting what your business knows and keeping it yours.

When a key person leaves, they take with them everything that was never written down. Most businesses understand that risk, even if they do not do much about it. The risk most are not thinking about is quieter and more damaging. It is happening right now in organisations that have started using AI tools without asking the right questions first.

Your data is your wealth. Your client relationships, your operational knowledge, your decision history, your processes — these are assets. You would not hand someone else access to your bank account. But when AI tools are deployed without governance, without encryption, without clarity about where your data goes and who can access it, that is effectively what is happening.

Data sovereignty is not a technical question. It is a commercial one. Before any AI tool is deployed in your business you should be able to answer: where does our data go, who owns it, how is it protected, and what prevents it from being used to train someone else's model?

If you cannot answer those questions, you are not ready to deploy.

The mistakes are as predictable as the opportunities. And in this environment, they are more expensive.

Do not deploy AI to reduce headcount you cannot afford to lose. The Employment Rights Act has made getting rid of people more costly and more complicated. In most cases the disruption and the legal risk will cost you more than you save.

Do not buy tools before you have defined the problem. I have watched businesses do this repeatedly. The tool sits unused, the licence fee keeps coming, and six months later nobody can quite remember what it was supposed to solve.

Do not put AI in front of your customers without understanding what happens when it goes wrong. A slow customer experience is forgivable. A bad one is not. In a tight market you cannot afford to lose customers to a saving that was not worth making.

And do not treat AI as a one-off project. The businesses that get genuine value from it are the ones that build it into how they operate, with proper governance, clear accountability, and explicit protection for their data. That is not a purchase. It is a practice.

The question that should come before everything else

Before you speak to a vendor, before you approve a budget, before you let anyone in your business start experimenting, ask this and answer it honestly.

Where in this business is capacity being consumed by work that does not require human judgement, and what would it be worth to get that capacity back?

If you can answer that with something real, an actual workflow, an actual measure, a clear definition of what success looks like, you are ready. If you cannot, you are not making a decision. You are shopping. And shopping is expensive when every commitment carries more weight and every mistake costs more to fix. The question that should come before everything else

I have spent thirty years watching businesses make decisions they could stand behind and decisions they could not. The ones they could not stand behind almost always had the same thing in common. They were made before the question was properly formed.

AI is no different. In the current climate it is simply the place where that mistake is most expensive.