There is a big difference between having AI tools available and actually adopting AI successfully within a business.

It is easy to buy licences, give everyone access and assume the hard work is done. But if people don't understand why they should use the technology, don't trust the output or simply don't have time to learn another new system, those shiny new tools can very quickly gather digital dust.

The session AI Adoption in Practice: Making It Work in Your Team, led by Dr Stylianos Kampakis, looked at what actually turns a promising AI pilot into a dependable part of everyday work.

Takeaway 1

Buying the licence isn't the same as adopting the technology

One of the simplest points from the session is probably one of the most important. Measuring how many people have access to an AI tool doesn't tell you whether it is actually working for your business.

Instead, adoption should be viewed as a journey.

Takeaway 2

Start with the job, not the AI

"We need to adopt AI" sounds like a goal, but it isn't particularly useful when you're trying to work out what to do on Monday morning.

A much better place to start is with one real task your team already does. Follow the work from beginning to end and look at where time is being spent, where information has to be found, where things get stuck and where people are repeatedly doing the same manual work.

Then give your pilot a measurable goal. That might be reducing the time it takes to respond to a routine enquiry while maintaining the same quality, for example, rather than simply asking people to "use AI more".

Takeaway 3

If people aren't using it, find out what's actually getting in the way

When an AI pilot isn't gaining traction, it can be tempting to assume the team is resistant to change. In reality, there could be a much more practical problem hiding underneath.

The session highlighted five common places where adoption can stall.

Takeaway 4

You probably don't need every AI tool going

With a new AI platform seemingly appearing every five minutes, it is very easy to end up with a collection of tools rather than a useful system.

The advice from the session was refreshingly simple: give each tool a clear job and choose a small stack that your team can realistically support.

Takeaway 5

Choose pilot users who will find the problems

Your first users don't necessarily need to be the biggest AI enthusiasts in the business. In fact, having different perspectives can make your pilot much more useful.

An experienced colleague can spot the awkward exceptions. A newer colleague can tell you whether the process is actually clear. A constructive sceptic will challenge weak answers. And a busy operator will very quickly tell you whether something works in the reality of a normal working day.

Takeaway 6

Make the deal with your team clear

There can understandably be nervousness around AI at work. Will it replace part of my job? Am I going to be judged if I struggle with it? What happens if it gives me the wrong answer?

Ignoring those concerns doesn't make them disappear. A better approach is to be clear about what you're testing, why you're testing it, what people remain responsible for and what support they will have.

Takeaway 7

Train people for the job they actually need to do

Your team doesn't need to become AI experts overnight. They need to know how to use the tool safely and effectively for the work they're actually responsible for.

The session demonstrated how even a focused 30 minute training session can be structured around a real task rather than a long tour of everything the technology can theoretically do.

Takeaway 8

Give people something useful to work from

If everyone is left to figure out AI for themselves, you'll quickly end up with different approaches, different standards and very different results.

A shared starter kit can give the team a much more useful foundation.

What does this look like in practice?

Successful adoption still requires people, time and ownership

The case studies shared during the session were very different in scale, but they all reinforce the same point: successful AI adoption doesn't happen by simply switching a tool on.

Takeaway 9

Automation doesn't mean removing people from every decision

One of the examples in the session looked at Nuuly's customer service workflow, where AI can identify a request, retrieve live information, verify details, take an eligible action and confirm the outcome.

But importantly, sensitive exceptions can still be passed to a person for approval. That's a useful principle for businesses of any size. Automate the defined, repeatable work where it makes sense, while keeping human judgement where the decision genuinely needs it.

Takeaway 10

Measure the real task, not just the impressive number

AI case studies can produce some eye catching figures, but numbers need context.

For example, the session highlighted a BBVA Peru workflow where average query handling time was reported as falling from around 7.5 minutes to around one minute.

That's significant, but it doesn't automatically translate into an annual saving. To understand the real benefit, you still need to consider how often the task happens, how widely the new process is actually being used, whether quality has been maintained and how much checking or rework is required.

Putting it into practice

Use 90 days to prove one useful workflow

Rather than trying to transform the whole business at once, the session suggested using a 90 day planning window to focus on one workflow and gather enough evidence to decide what happens next.

For Home & Gift businesses

What does this mean for our industry?

For many Home & Gift businesses, the best AI opportunities probably aren't going to begin with a huge transformation project. They'll begin with one repetitive piece of work that takes up more time than it should.

It might be drafting product descriptions, responding to common customer enquiries, preparing retailer information, finding information across internal documents, creating first drafts of marketing content or helping teams organise and summarise information.

Choose one. Understand how it works today. Test whether AI can genuinely improve it. Give people the right training and guardrails. Then measure what actually happened before deciding whether to expand.