Northern Ireland Has an AI Strategy. The Interesting Part Is What Happens Next.

Northern Ireland has published its first draft Artificial Intelligence Strategy. I spent some time reading through the full 48 pages, and there is one sentence buried relatively early in the document that, for me, captures what the whole thing should really be about.

“A focus on what AI can accomplish rather than on how AI works should be the remit of a user-focused strategy.”

That's it.

We've spent the last few years talking endlessly about AI. Large language models. Generative AI. Agents. Prompts. Machine learning. Automation. The terminology keeps changing almost as quickly as the technology itself.

Most people don't need more terminology. They need things to work better.

The draft strategy appears to understand that. It deliberately categorises AI according to its outputs because that is easier for non-experts to understand and keeps the focus on impact and value rather than technical capability. That's a sensible approach.

The Numbers Make the Opportunity Hard to Ignore

One of the headline figures is economic.

The strategy references research from Matrix which estimates that AI could contribute between £0.2 billion and £3.7 billion in additional GVA to Northern Ireland by 2030, compared with 2023. The midpoint estimate is £1.2 billion.

That's a huge range, of course. And forecasts around emerging technologies should always be treated as forecasts rather than promises.

But even the lower end tells us something.

There is an economic opportunity here.

The question is how much of it Northern Ireland manages to capture.

And that won't be determined by how many people have an AI account or how many organisations can say they've introduced an AI policy. It will come down to whether people and organisations actually become more productive because of it.

That distinction matters.

Buying AI is easy. Changing how work gets done is harder.

I Like That This Isn't Really a Technology Strategy

Something else worth understanding is that this document isn't trying to tell every government department which AI platform to buy.

In fact, it explicitly says it doesn't prescribe technologies, platforms or individual digital solutions. It describes itself not as a technical manual or an economic blueprint, but as a framework for safe, ethical and purpose-driven AI adoption.

That's probably the right decision.

Any government strategy that prescribed today's AI technology would risk being out of date before the procurement process was finished.

The tools will change.

The models will change.

The interfaces will change.

What matters more is building organisations capable of adapting to those changes.

That's why some of the less exciting parts of the strategy are probably among the most important.

Data governance. Training. Literacy. Infrastructure. Accountability. Interoperability.

None of those makes for a particularly exciting AI demonstration.

But without them, the exciting stuff doesn't work very well.

Government Has to Fix the Plumbing

One of the recurring mistakes with AI is assuming you can place it on top of an organisation and suddenly make the organisation intelligent.

It doesn't work that way.

If your information is scattered everywhere, your data is badly organised, systems can't communicate with each other and nobody knows which version of a document is correct, adding AI doesn't magically fix the underlying mess.

Sometimes it just helps you make a mess faster.

The strategy recognises this problem.

Its ambition is not simply to deploy AI but to establish the governance, infrastructure, data practices and skills required to use it properly. It also stresses that AI should support human decision-making rather than render it obsolete.

That's important.

Before asking what AI can do for government, government needs to ask whether its information and systems are ready for AI.

The boring work comes first.

Clean up the data. Connect the systems. Decide who owns what. Establish sensible permissions. Train people properly.

Then automate.

The Human Part Matters More Than the AI Part

The strategy is built around eight principles: Human Oversight, Accountability and Redress, Data Governance, Technical Safety and Security, Fairness and Transparency, Sustainability, Societal Benefit, and Training and Literacy.

It's tempting to look at a list like that and see bureaucracy.

Some of it will inevitably become bureaucracy.

But the underlying principle is right.

AI shouldn't remove responsibility.

If an AI-assisted system contributes to a decision affecting someone's healthcare, education or interaction with government, we shouldn't reach a point where nobody can explain what happened because "the computer decided".

A human still owns the decision.

The strategy makes that position quite clear. AI isn't presented as a replacement for human intelligence. It is there to add value to what people do and support human decision-making.

I think that's a much healthier way to think about AI generally.

The interesting question isn't, "How many jobs can AI replace?"

It's, "How much low-value work can we remove from someone's job?"

Those are very different things.

AI Literacy Can't Just Mean Learning About AI

This is where I think Northern Ireland has one of its biggest opportunities.

We need widespread AI literacy, but I would be careful about what we mean by that.

We don't need thousands of people sitting through courses explaining the history of artificial intelligence.

We don't need everyone to understand how a transformer model works.

And I'm increasingly unconvinced that teaching endless lists of "prompts" is the answer either.

Teach outcomes.

Give someone a real piece of work they currently spend three hours doing and show them how to do it safely in 30 minutes.

Take a repetitive administrative process and simplify it.

Take a pile of information nobody has time to analyse and find useful patterns in it.

Help someone research faster.

Help a manager understand information faster.

Help a business owner communicate better with customers.

That's AI literacy.

The strategy itself makes essentially the same point when it argues that a user-focused approach should concentrate on what AI can accomplish rather than how the technology works.

That sentence should probably be printed above the door of every AI training programme we fund.

And Don't Forget Small Business

The draft is primarily a public-sector AI strategy. That's important context. The Executive says it is designed to establish responsible AI adoption across Northern Ireland's public sector while providing a framework others can use, and the consultation specifically invites businesses and SMEs to participate.

But I think the next stage needs a very strong small-business component.

Northern Ireland doesn't become an AI-enabled economy because a handful of technology companies in Belfast become exceptionally good at AI.

It happens when ordinary businesses become better because of it.

The independent retailer.

The accountant.

The hotel.

The manufacturer.

The tourism business.

The construction company.

The person running a company with five employees who doesn't have an IT department, never mind an AI department.

They don't need an AI transformation programme.

They need someone to show them how to save five hours next week.

That's where this gets interesting.

Because multiply small improvements across thousands of businesses and suddenly you're talking about productivity at an economic level.

Northern Ireland Being Small Isn't Necessarily a Problem

One of the more ambitious claims in the strategy is that Northern Ireland could become a major AI player on the global stage.

It points towards expertise across sectors, connectivity between government, academia and industry, and potential infrastructure opportunities including green data centres and foreign direct investment.

"Global AI player" is a big claim.

But I don't think Northern Ireland needs to compete with Silicon Valley by trying to build the next foundational AI model.

There may be another route.

Become exceptionally good at applying the technology.

Northern Ireland is small enough that government, universities, FE colleges, councils and businesses should theoretically be able to work together relatively quickly.

That's an advantage if we use it.

It's a disadvantage if a small place manages to create the bureaucracy of a large one.

AI moves quickly. Government traditionally doesn't.

That's one of the tensions this strategy will have to overcome.

A three-year programme designed around today's technology could be obsolete long before the programme finishes.

So experiment smaller.

Measure faster.

Keep what works.

Drop what doesn't.

Then scale.

Measure the Outcome, Not the AI

This is probably the biggest thing I'd like to see carried through into delivery.

Don't tell us how many AI projects government has.

Tell us what changed.

Did waiting times fall?

Did staff get hours back?

Did a process that took 20 minutes now take five?

Did an employee stop spending Friday afternoon copying information between spreadsheets?

Did citizens get answers faster?

Did an AI-assisted process reduce errors?

Did a business become more productive?

If nothing got noticeably better, it doesn't really matter how sophisticated the AI was.

The technology isn't the outcome.

The outcome is the outcome.

And ironically, that's one of the most intelligent ideas contained within Northern Ireland's AI strategy itself.

The Strategy Is the Easy Part

The Executive has now opened the strategy for public consultation until 7 October 2026, after which responses will be analysed and a final strategy brought forward for Executive agreement.

There is plenty in the draft worth supporting.

There are sensible principles around human oversight, transparency and accountability. There's recognition that skills and literacy matter. There's an understanding that AI needs good data underneath it. And there's a refreshing emphasis on outcomes rather than obsessing over how the technology itself works.

But writing the strategy was always going to be the easier part.

Now comes implementation.

AI doesn't need another committee discussing what AI might eventually do.

People are already using it.

Businesses are already changing because of it.

Jobs are already changing because of it.

The technology will continue moving regardless of the speed at which government moves.

Northern Ireland therefore has a choice.

We can spend the next few years talking about becoming an AI-enabled economy.

Or we can start becoming one.

Not by chasing every shiny new AI tool that appears.

Not by forcing AI into every process.

And certainly not by measuring success by the number of licences purchased.

Start with the work.

Find the friction.

Find the repetitive tasks.

Find the things taking too long.

Find the information nobody can make sense of.

Then ask whether AI can make any of them better.

That's a much less glamorous AI strategy.

It might also be the one that actually works.