Not every operational problem needs AI.

Sometimes it doesn’t need any new technology at all.

When a business identifies a bottleneck or inefficient workflow, it can improve it in several ways. The right answer depends on what is actually causing the problem.

That distinction matters because if you start by saying, “We want to implement AI,” you’ve already chosen the solution before you’ve properly investigated the problem.

I’d rather start somewhere else: Where is the business losing time, capacity or revenue, and why?

Once you understand that, you can decide what deserves to be solved.

1. Change the process

Some operational problems exist because the process itself no longer makes sense.

A workflow may have developed gradually over several years. New steps get added, responsibilities change, and exceptions become part of everyday work. Eventually, people can find themselves following a complicated process without anyone being particularly sure why every step still exists.

Talking to the employees doing the work can reveal that some steps could be removed, combined or handled differently.

If changing the process solves the problem, introducing more technology could add unnecessary cost and complexity.

Sometimes the best solution is a better process.

2. Connect the systems you already have

Another common problem is that a business already owns most of the technology it needs, but the systems don’t work together effectively.

An enquiry arrives in one system, and somebody manually enters it into another. Customer details are copied into the accounts software. Information gets exported, manipulated in a spreadsheet and then uploaded somewhere else.

The individual systems may be perfectly capable. The inefficiency exists in the gaps between them.

In that situation, an integration may be more valuable than another piece of software.

Connecting existing systems through available integrations or APIs can let information move automatically and reduce the time employees spend acting as the connection between different parts of the technology stack.

3. Automate a predictable workflow

Some processes follow clearly defined rules.

When a particular event happens, the next action is known. If a customer selects one option, one workflow should follow. If they select another, something different should happen.

I’ve built these kinds of workflows in my own business for years.

Client enquiries, bookings, contracts, and communications could move through defined stages without me manually pushing every client from one stage to the next.

When inputs are reliable, and you can express your staff decisions through predictable rules, conventional automation can be extremely effective.

You don’t necessarily need AI to solve a problem that already has clear logic.

4. Use AI where interpretation adds value

AI becomes more interesting when a workflow requires something difficult to handle with straightforward rules.

That might involve interpreting less structured information, analysing content, identifying patterns, or helping make decisions where too many variables make traditional

logic inefficient.

This is where AI can extend what automation can do.

But I’d still want to understand the surrounding workflow before introducing it.

If the source information is unreliable, the process is poorly designed, or another bottleneck exists immediately afterwards, adding AI to one stage may not improve the overall result.

The technology has to make sense within the wider operation.

Start with the commercial problem.

This is why I think operational improvement should begin with diagnosis rather than technology selection.

If an audit identifies a bottleneck, I want to understand how often it occurs, who is involved, what it costs and what effect it has elsewhere in the business.

Then we can compare that cost with the effort and investment required to improve it.

A process change requiring little investment might produce an excellent return. Connecting two existing systems could release hundreds of hours of staff capacity.

Conventional automation may handle a repetitive workflow perfectly well.

In another situation, AI might provide capabilities that weren’t previously practical or affordable for the business.

Those are very different solutions, and that’s the point.

Choose the simplest solution that delivers the result.

I’m positive about the opportunities AI creates for established businesses. But using AI shouldn’t measure of whether an operational improvement project is ambitious enough.

The objective is to improve the business.

That means understanding where time, capacity or revenue is being lost, identifying the underlying cause and choosing an appropriate response.

Sometimes that’s AI. Sometimes it’s automation or an API connecting systems you already own. Sometimes the answer is simply changing how the work gets done.

The cleverest solution isn’t necessarily the best investment.