Why you probably don't need AI... yet.

There's something slightly ridiculous about an AI consultancy telling businesses they probably don't need AI.

But here we are.

At Coal Face AI, we're obviously big believers in what artificial intelligence can do. We've seen enough of the technology, both technically and commercially, to know that it can make an extraordinary difference to how organisations operate.

But we've also seen something else.

Businesses spending money on technology before they've properly worked out what problem they're trying to solve.

And right now, AI is getting an awful lot of that money.

Everyone wants AI. Fewer people know why.

There's a familiar conversation taking place in businesses across the country.

Someone in senior management has been to a conference. They've seen a demonstration, read about what their competitors are doing, or heard that AI is about to transform their industry.

They come back with a perfectly reasonable question:

"What are we doing about AI?"

Unfortunately, that can quickly turn into a less reasonable instruction.

"We need to implement AI."

Implement it where? To do what? For whom? At what cost? And how will we know whether it's actually helped?

Those questions tend to get rather less attention.

IBM's 2025 study of 2,000 CEOs reported some fairly sobering findings about the return businesses were getting from their AI investments.

Think about that for a moment.

We're making some investment decisions because we're worried about missing out, rather than because we've demonstrated a genuine need.

That's not a particularly good foundation for any business investment, let alone one involving unfamiliar technology.

Your problem might not be an AI problem.

Imagine a business where customer orders arrive by email.

Someone opens the attachment, copies the information into a spreadsheet, checks a few details against another system, and then manually enters everything into the company's order management software.

It's slow. It's repetitive. Mistakes happen.

An obvious opportunity for AI, surely?

Possibly.

But perhaps the first question should be why the information needs to be copied between three systems in the first place.

Could those systems already communicate through an existing integration? Could some of the information be captured through a structured form? Could a straightforward automation remove most of the manual work?

You might discover that a relatively simple improvement solves most of the problem.

Or you might discover that AI is genuinely useful for interpreting inconsistent customer documents, identifying missing information and preparing an order for human review.

Both are perfectly good outcomes.

"The difference is that you've worked backwards from the business problem rather than forwards from the technology."

That's where meaningful improvement starts.

The research tells a similar story.

This isn't just an opinion from a small AI consultancy in Yorkshire.

RAND's research into the causes of AI project failure found that one of the recurring problems was organisations failing to properly identify or communicate the business problem they wanted AI to solve.

Sometimes the technology was designed around the wrong objective. Sometimes the necessary data wasn't available. And sometimes the latest technology was being pursued when a simpler approach would have been more suitable.

The UK's own research paints an interesting picture too.

Adoption is clearly accelerating.

But the research also suggests that adoption remains relatively shallow. More businesses are using AI, without necessarily integrating a much broader range of AI technologies into their operations.

McKinsey's 2025 global research adds another useful distinction: while AI was becoming commonplace, only 39% of survey respondents reported any enterprise-level impact on earnings before interest and tax attributable to AI.

Even among those reporting an impact, most attributed less than 5% of their organisation's EBIT to AI.

None of this means AI doesn't work.

It means buying access to AI and making a business materially better are not the same thing.

Sometimes the boring answer is the right one.

If a process is badly designed, automating it doesn't necessarily make it a good process.

It might just mean you've found a more expensive way to do something badly.

And if the information flowing through that process is inconsistent, incomplete or unreliable, adding AI introduces another layer of complexity.

That's not an argument against innovation. Quite the opposite.

It's an argument for getting the foundations right so that innovation actually delivers something.

That might mean cleaning up your data, joining two systems together, removing an unnecessary approval stage, or spending a day with the people who actually perform the work.

None of that sounds particularly exciting in a boardroom presentation.

But it can make an enormous difference.

And importantly, you don't always need an expensive technology project to achieve it.

So when do you actually need AI?

When there's a genuine reason to use it.

AI becomes particularly interesting when the work involves things conventional systems struggle to handle.

Interpreting information that arrives in different formats. Extracting meaning from large volumes of documents. Identifying patterns in complex datasets. Helping people investigate problems, prepare decisions or work through information that would otherwise take hours to assess.

These can all be worthwhile applications.

But the existence of a possible AI use case doesn't automatically make it a worthwhile investment.

You still need to understand the volume of work involved, what it currently costs, how often things go wrong, and what improvement would justify the investment.

You also need to consider what happens when the AI gets something wrong.

Who checks the output? Who is accountable? Can decisions be explained? Is the information appropriately protected?

Those questions aren't obstacles to getting started.

They're part of getting started properly.

Five questions worth asking before spending anything.

If you're considering introducing AI into your business, I'd suggest starting here.

If you can't answer those questions yet, that's absolutely fine.

But you probably aren't ready to commission an AI solution.

You're ready to understand your business a little better.

And that's a worthwhile investment in its own right.

But 'not yet' shouldn't mean 'never'.

There's another side to this argument that's worth recognising.

Businesses shouldn't use a lack of readiness as an excuse to ignore AI indefinitely.

The technology is advancing, competitors are experimenting, and some organisations are already finding genuine efficiencies.

There is a cost to standing still.

But there's an important difference between deciding not to invest in an AI solution today and deciding not to investigate AI at all.

You can start by understanding your workflows, improving your information, training your people and identifying the areas where technology could make a meaningful difference.

You can test a narrowly defined use case without committing to a major transformation programme.

You can establish what success would look like before spending serious money.

In other words, you can get ready.

The businesses that benefit most won't necessarily be the ones that adopted AI first.

They'll be the ones that understood where it belonged.

We would rather tell you that you don't need AI.

There's a reason we called our company Coal Face AI.

We believe the value of technology is realised where the actual work happens. Not in the presentation, the strategy document or the impressive-looking demonstration.

It's realised by the people processing orders, managing enquiries, making decisions, dealing with customers and keeping the business running.

That's where we prefer to start.

Sometimes we'll identify an opportunity for a genuinely useful AI solution. Sometimes we'll recommend a smaller piece of automation. Sometimes the sensible answer will be to improve a process before touching the technology at all.

And sometimes we'll tell you that AI isn't the right investment just yet.

We'd rather have that conversation than sell you something you don't need.