You’re paying for AI. Is your team making the most of it?

Give people more than a login

Sending a link with “see what you can do with AI” gives people access, but little direction.

They need to understand where it could help in their role, what information they can use and what a good result looks like.

A graphic designer and a finance manager shouldn’t be expected to test it in the same way. Their work, standards and responsibilities are different.

For example:

• A graphic designer could use AI to turn a loose brief into several distinct creative territories, then use those to sharpen the direction before spending time on finished design. The value is not the generated artwork. It’s getting to stronger options faster.

• A head of marketing could combine campaign performance, CRM data, sales feedback and customer behaviour into one analysis, then use AI to surface patterns, contradictions and possible reasons performance changed. The useful output is a better set of questions to investigate, not an automated verdict.

  
  

• A product manager could analyse hundreds of support tickets, reviews and interview notes by customer type or stage of the journey, then surface recurring problems alongside smaller but potentially important signals that would disappear in a simple summary.

  
  

• A finance manager could use AI to investigate variances by bringing together invoices, purchase orders, budgets and transaction notes, highlighting where figures do not reconcile and suggesting likely causes for a person to verify.

These give each person a relevant starting point and the business something specific to assess.

Be clear about what you’re testing

In the AI applications I developed, I worked through the source information, instructions, required output, and where someone still needed to check the work or make a decision.

For each test, agree on:

• The task: What necessary, time-consuming work could AI help with?

• The information: What does it need, and what is appropriate to use?

• The standard: What would make the result correct and ready to use?

• The checks: What must a person verify, investigate or approve?

Try it on real examples, including incomplete or unclear material. One impressive result doesn’t tell you enough about everyday use.

Check whether the investment is worthwhile

Look beyond how often people log in. Include the subscription cost, setup and training time, and the effort spent checking or correcting results.

Is it saving time overall? Improving the work? Helping people do something valuable they previously struggled to do?

If someone rarely uses it, find out why. They may need support, a more relevant application or a different tool.

The aim is for people to use AI where it makes a worthwhile difference to their work.