Co-Founder at Loomery

Most AI strategies fail for the same reason. They get signed off, greeted with quiet approval and then ignored for the next six months until the contents have been largely forgotten about. This isn’t the result of a failure of ambition or bad strategy, it's what happens when the technology and its adoption are treated as an add-on that sits parallel to the business.

Start with the tools and you get lots of pilots. A chatbot here, an automation there, a hackathon that produces a demo nobody opens again. Plenty of AI activity, but not much you’d call a measurable success. MIT's NANDA initiative found that 95% of enterprise AI pilots in 2025 failed to deliver return and the main barrier was found to be organisational, not the technology itself.

By ditching the separate strategy doc and focusing on the strategy your organisation already has the question changes from ‘what should our AI position be?’ to the more productive ‘how can AI help us do what we’re already doing, faster and better?’ That reframing keeps efforts pointed at something that matters, and gives you a natural way to measure success.

Why this matters now

The pressure behind this isn’t theoretical. Customers are already arriving with AI-assisted research, comparisons and expectations before they touch your product. And soon your customers will likely be agents, if they are not already. With everyone able to access the same frontier models, the advantages that used to take a business decades to build are worth less.

So, what’s left to compete on? We’d say the answer is the speed at which organisations can (1) make the right decisions on what to build and (2) get high quality experiences in front of customers. The technology isn’t the limiting factor - with frontier AI’s capability well ahead of most organisations' ability to use it - the main challenge is ensuring capability sticks.

What this looks like in practice

1. Lead with outcomes

The most common mistake starts with the purchase of the tool. A team gets their Copilot licences or access to Claude, and only afterwards do they ask what they’re meant to be doing with it.

Start with your organisation and your users, decide what you're trying to achieve, then ask how AI helps you get there. Those outcomes can quickly be translated to your evals, the method with which you continuously measure the quality of the experience. Without a clear outcome you run the risk of AI usage being a success metric when
a clear marker of success is the improvement of processes.

2. Get teams making

AI capability isn’t built through an online training course. People learn and gain confidence quickest when they make and build from day one. A designer who spends an afternoon prototyping against a real brief, with a real deadline, hard practical constraints and stakeholder waiting on it, learns more than a week of training modules could teach them.

When we set out to raise AI maturity across Loomery, we established a structured approach to progressing everyone’s capability, across product, design, engineering and all levels of maturity. Teams should have the materials and expertise on hand to get started immediately and progress through these stages and hands-on learning is the fastest route there.

3. Sort out the hard stuff

Tooling, guidelines, access, permissions and data control. None of this is glamorous but it allows teams to move quickly and safely, avoiding any messes someone else will have to clean up later.

A simple way to assess if your guidelines are in place and ready to use is through the eyes of a new starter. On day one, would they get a clear answer on what tools they're allowed to use and the data they can and can't share with them? Any uncertainty means there’s work to be done.

4. Share what works

Meaningful change happens with trust. And that trust is best earned through proof that efforts are working. This is where finding your AI champions comes in. These are the people or teams getting the most value out of AI and whose working methods have visibly changed.

Put them in front of the rest of the team, celebrate their wins publicly and make it clear what’s possible.

Making capability stick

Capability won’t stick from a single initiative. It requires consistent attention and compounds over stages. The first stage is to get people in a room experiencing tools together through hack days, workshops or small group sessions. That’s the ignition point, building curiosity and excitement for the tools. But it’s not the place to stop.

What follows ensures that excitement translates into capability. Teams should have access to hands-on coaching and training in the flow of real work, with good examples set from within the organisation to ensure they continue to work towards the specific outcomes tied to AI enablement.

Finally, a Community of Practice should be established to give teams a forum to share what's working and what’s not, allowing learning to compound long-term instead of staying locked in one team's head.

Conclusion

Embedded capability is built from the bottom up. None of this needs a strategy document, it needs outcomes stated clearly enough to test against, people making real things with the resources to support, the groundwork done before it's needed and proof instead of announcements.

If you’re working out how to make AI stick across your product teams, get in touch, we’d love to talk through what that looks like for you.

Get the latest news and views from Loomery directly to your inbox

Stay ahead of the curve with our monthly newsletter, The Weave.

Newsletter sign up

Why you don’t need an AI strategy

August 2026

We take a bottom-up approach to AI enablement and focus on building capability first. Here's why.

Get the latest news and views from Loomery directly to your inbox

Move fast

Company news

Move fast

Let's talk →

Why you don’t need an AI strategy

August 2026

We take a bottom-up approach to AI enablement and focus on building capability first. Here's why.

Get the latest news and views from Loomery directly to your inbox

Score your team against the 8Cs

Move fast

Company news

Move fast

Let's talk →