“How can we use AI in our business?”

It’s a question many business owners and management teams are asking, and understandably so. The capabilities are developing quickly, new tools are appearing constantly, and there’s plenty of pressure to make sure your business isn’t being left behind.

But I don’t think it’s the best place to start.

Before deciding how to use AI, I’d want to understand something much more fundamental:

What isn’t working properly in the business today?

The answer to that question gives you a useful place to start.

Start with the business, not the technology.

If I were looking for opportunities to improve an established business, I’d start by understanding the systems and workflows it already relies on.

Where are employees losing time? Where is information being entered more than once? Which processes regularly cause delays? Where are customers left waiting? Which tasks require people to move information manually between systems?

I’d also want to understand where growth is pressuring existing processes.

A workflow that was perfectly adequate when it happened a few times a week can become a significant operational problem when the volume increases. Small inefficiencies have a habit of becoming much more expensive when they’re repeated hundreds or thousands of times.

Once those problems are visible, you can start deciding which ones are worth solving.

You may already own part of the solution.

There’s another question I think businesses should ask before buying more technology:

Are we making proper use of what we already have?

Established businesses often accumulate software over time. A CRM is introduced, followed by an accounting platform, booking software, project management tools and other specialist systems.

The individual systems may work perfectly well, but that doesn’t mean the overall operation works efficiently.

Employees can end up moving information manually from one system to another. The developers might not have configured the most useful functionality. A newer system might have been introduced without the old process ever really disappearing.

Before adding another tool, it makes sense to understand whether the technology you’ve already paid for can do more of the job.

The answer might be much simpler than AI.

Once you’ve identified a genuine operational problem, AI becomes one possible solution rather than the objective.

You might discover that two existing systems need connecting properly. A repetitive task might be handled with relatively straightforward automation, or functionality you already own could remove several manual steps.

Sometimes the workflow itself needs redesigning. Sometimes the best answer is to stop doing something that has continued through habit long after its original purpose disappeared.

And sometimes AI is genuinely the appropriate solution.

The important thing is that you’ve arrived at that conclusion because of the problem you’re trying to solve.

AI can’t compensate for weak foundations.

Putting sophisticated technology on top of a poorly designed process can make the underlying problem harder to see.

If the source data can’t be trusted, automating decisions based on that data creates another problem. If employees are already working around a broken workflow, adding

another system can give them another system to work around.

And if a business already has a bottleneck further down the customer journey, using AI to generate more enquiries may send more work into the same bottleneck.

This is why I think understanding the end-to-end workflow matters.

Technology can improve capacity, response times and customer experience, but only when you understand what is preventing those things from improving in the first place.

Put a number against the problem.

Identifying an inefficient process still doesn’t automatically mean you should change it.

I’d want to understand what that inefficiency is actually costing the business.

If a manual task takes five minutes but happens hundreds of times a month, the cumulative cost may be significant. If solving it would release meaningful staff capacity for a relatively small investment, it could be worth prioritising.

Another problem might be frustrating but infrequent, with an expensive or complicated solution. In that case, living with it could be the better commercial decision.

This is where conversations about AI and automation need to shift from capability to return.

The question isn’t only whether something can be improved. It’s whether improving it is worth the time, money and disruption involved.

Where AI fits

I’m hugely positive about what AI can do for businesses, and I expect the range of commercially useful applications to continue growing.

But being positive about AI doesn’t require starting every operational improvement project with AI.

I’d rather understand the business first, diagnose the workflows creating unnecessary cost or limiting capacity, and then decide what type of intervention makes sense.

That could mean improving a process, using existing software more effectively, integrating systems, introducing conventional automation or applying AI where its capabilities genuinely add something useful.

AI becomes much more interesting when it’s attached to a clearly understood business problem.

So before asking “How can we use AI?”, I’d start with “What problem in this business is actually worth solving?”

The technology comes afterwards.