Before You Buy Another AI Tool, Fix the Process Behind It

The demonstration looks convincing. An AI assistant reads an email, summarises the request and drafts a polished reply in seconds. Someone around the table asks, “How soon can we get this?”

It is an understandable reaction. When your team is stretched and customers expect faster answers, anything that promises relief deserves attention.

But before approving another subscription, ask a less exciting question:

What, exactly, is slowing the work down?

If the answer is unclear responsibilities, missing information or unnecessary approvals, a faster writing tool may leave the real problem untouched. Business process improvement starts with understanding how work happens—and where it gets stuck.

Why another AI tool may leave the bottleneck untouched

Consider this hypothetical customer service team.

An enquiry arrives in a shared inbox. An employee copies the details into a spreadsheet, forwards the message to another department and waits for approval. Meanwhile, the customer sends a follow-up, which a different employee picks up.

An AI tool could help draft both replies. But who owns the enquiry? Which record is current? Does the response really need approval?

Until those questions are settled, the team could produce messages faster while customers continue waiting.

This distinction matters in 2026. The OECD’s 2026 survey of SME digitalisation reports growing AI adoption alongside uneven integration into business operations. Its findings come from a non-representative sample of more than 2,000 SMEs across 12 countries, so they should not be treated as a universal picture. Nevertheless, the implementation challenges offer useful questions for business leaders.

Buying access is one decision. Making it useful is another.

1. Describe the problem without naming a product

“We need AI” is a proposed solution. It does not explain the business need.

Start with something observable:

• Customer enquiries wait too long for an owner.
• Employees enter the same information into several systems.
• Managers spend hours reconciling conflicting reports.
• New starters arrive without the access they need.

Then establish a baseline. How long does the task take? How often does it happen? Where do errors or delays appear?

For example: “We want to reduce the time between receiving an enquiry and assigning responsibility, while maintaining response quality.”

That statement gives you something to investigate and measure. It also leaves room for solutions you may not have considered.

2. Map what happens on an ordinary working day

Invite the people who perform the work to walk through a recent example, from beginning to end.

Ask what happens when information is missing, a colleague is absent or a request falls outside the usual rules. These exceptions can reveal more than a neatly written procedure.

A useful process map records the steps, decisions, handovers and waiting points. You can begin with sticky notes; specialist software can come later.

Simply, if you need to explain, improve or standardise how how works gets done, process maps helps. This is done by observing and recording the actual sequence of work, with the people involved. It doesn’t mater which industry, the method offers a practical starting point for examining workflows.

The important question is simple: does the map reflect what people actually do?

3. Remove avoidable work before automating it

Once the workflow is visible, examine each step.

Why is the same detail entered twice? What purpose does that approval serve? Could a clearer form prevent repeated follow-up questions?

Returning to our hypothetical service team, an initial improvement might involve assigning one enquiry owner, agreeing an escalation rule and maintaining one reliable record.

Those changes would not automatically solve every problem. They would, however, create a clearer workflow against which to assess a tool.

Every step you remove is a step you no longer need to automate, maintain or explain to a new employee.

4. Match the technology to the task

Different problems call for different approaches.

A straightforward rule—such as routing an enquiry according to its selected category—may be handled by an existing system or conventional automation.

A task involving varied language, such as summarising lengthy customer messages, may be worth exploring with AI.

For each proposed use, ask:

• What information will the tool need?
• How will we check its output?
• What happens when it gets something wrong?
• Who remains responsible for the result?

The NIST AI Risk Management Framework organises risk management around governing, mapping, measuring and managing AI risks. It also emphasises testing before deployment and during operation. For a business manager, that means understanding the context and responsibilities alongside the capability.

5. Run a small test and count the whole cost

Choose a limited workflow and agree the success measures before starting.

For an enquiry-handling pilot, you might track assignment time, response quality, correction rates and staff effort. Include the time spent checking AI output, resolving exceptions and maintaining the setup.

Compare the pilot with your baseline. Where practical, also compare it with the simplified process operating without AI. This helps you distinguish the benefit of the tool from the benefit of cleaning up the workflow.

A faster draft is useful. A faster, accurate and properly handled customer response is a more meaningful business outcome.

Before your next purchase, follow one piece of work

You do not need to review the entire organisation before trying AI.

Start with one recurring frustration. Follow it from arrival to completion. Speak to the people involved. Identify the delay, simplify what you can and define what improvement would look like.

Then you will be in a stronger position to judge the demonstration—and ask whether the tool solves your problem.

Need help identifying where your workflow gets stuck? Contact me to discuss your process improvement challenge. Together, we can clarify the problem and explore practical next steps.