AI is the biggest opportunity Guernsey has had in a generation, and I want us to grab it with both hands.
For a small island with a tight labour market and an economy built on skilled services, tools that allow one person to accomplish work that previously took several are almost tailor-made for us.
We have spent years worrying about where our future workforce will come from. Guernsey’s working-age population is projected to fall under every official demographic scenario to 2033. Even under a high-migration scenario it falls by about 1.1 per cent, while without migration the reduction would be about 8.1 per cent. At the same time, the number of older islanders that workforce must support increases.
For most of our history, the answer to a shortage of people has been to find more people.
AI gives us another answer: make the people we already have dramatically more productive.
Done properly, this is the closest thing Guernsey has to growing its workforce without growing its population.
And for perhaps the first time in a long time, being small could become an advantage.
A Guernsey accountant, advocate, administrator, retailer or entrepreneur can access technology that would once have required an IT department and a substantial budget. A small business does not need to wait years for an enterprise transformation programme. An individual can begin using these tools today.
That starts to change the economics of scale.
The opportunity for our finance industry is particularly significant. Financial and insurance activities account for around 37% of Guernsey’s annual GDP, and finance is among the sectors most exposed to AI.
If a Guernsey financial services business can use AI to handle routine administrative work faster, analyse information more effectively and allow skilled employees to serve more clients, it can compete with organisations many times its size.
But there is an uncomfortable flip side.
The same technology that allows businesses in Guernsey to serve the world makes it easier for businesses elsewhere to perform work currently done here. The business can survive while some of the work, jobs and economic value move elsewhere.
So our challenge is not to protect Guernsey from AI. It is to ensure Guernsey captures the value AI creates.
And that brings me to government, because here AI creates a very different problem.
Every new weapon changes the battle, and not always in the way you expect.
When armies exchanged spears for rifles and machine guns, they acquired vastly greater capabilities. But technology did not necessarily reach both sides at the same speed, nor did institutions immediately understand how to use it or defend against it.
Something similar is beginning to happen to government.
AI does not simply make answering questions easier. It makes asking them extraordinarily easy.
And at precisely this moment, Guernsey is considering strengthening the machinery through which government can be questioned.
The Scrutiny Management Committee wants changes to our Freedom of Information arrangements that could ultimately give decisions of the independent appeals process legal force. Its argument is that an independent appeals mechanism must be more than advisory if it is to provide meaningful accountability.
I agree with the principle.
Government should be transparent. The public should be entitled to ask how decisions have been reached. Journalists should be able to investigate. Deputies must be able to challenge committees and hold them to account.
But we also need to confront the other half of the argument.
We are considering strengthening one side of the system just as technology is giving that side an extraordinary increase in firepower.
Look at deputies’ written questions. Rule 14 questions to Policy & Resources increased from 64 in 2024 to 143 in 2025 and had already reached 145 by the beginning of September 2026. Since 2024, more than 350 questions had gone to P&R. Health & Social Care had dealt with 115 and Home Affairs 111.
Yet the same Freedom of Information response revealed something that concerns me even more. There is no central log of Rule 14 questions across government. Officer time and cost are not routinely recorded. The involvement of legal advisers, communications staff and senior officers is not centrally logged. There is not even an agreed electronic format for recording them.
That is where we start the AI revolution from.
The questioner has just been issued a machine gun. Government is still holding a bow and arrow.
A deputy can use AI to interrogate a long policy paper and develop possible lines of questioning. A journalist can analyse mountains of published material. An islander can get help turning a concern into a carefully formulated Freedom of Information request.
I think that empowerment is overwhelmingly a good thing. AI is democratising capabilities that were previously expensive or time-consuming.
But generating the question and producing an official answer are fundamentally different propositions. The person asking can adopt AI immediately. Government cannot.
Guernsey has the information-security responsibilities of a small nation, but the administrative capacity of a town.
Our government does not merely possess ordinary office documents and the personal details of local residents. Across Guernsey’s public institutions there is information relating to people and businesses, taxation, regulation and law enforcement, as well as confidential communications and information exchanged with other governments and authorities.
As an international finance centre, the Bailiwick’s public authorities also operate within international tax-transparency, regulatory and financial-crime frameworks. Although Guernsey is a community of little more than 60,000 people, we have many of the information responsibilities expected of a nation state.
That completely changes the problem.
My personal AI assistant can search whatever information I decide to give it. A government AI would operate in an environment where different people have very different rights to see different information.
The danger is not simply that an AI might reveal somebody’s personal details. It is that a tool whose strength is its ability to find, connect and summarise information could cross a boundary it should never have crossed: information about a company, a taxpayer, a regulatory matter, a law enforcement case, confidential correspondence or information received from another authority.
The more powerful AI becomes at finding information, the more important those boundaries become.
Before AI can safely work across government information, we need to know what it can see, who is making the request, what that person is authorised to see, what the AI may do with the information and how its output will be checked.
That is not bureaucracy invented to slow AI down. It is part of being a government.
The questioner can rearm overnight. Government has to rebuild the armoury.
Freedom of Information already demonstrates the strain. The Code received just 14 requests in its first full year. In 2025 it received 164, while only about 41 per cent were answered within the 20-working-day target.
The marginal cost of asking is collapsing while the cost of producing a reliable official answer remains stubbornly human. For smaller departments the consequences can become serious. An officer researching an FOI or preparing a Rule 14 response cannot simultaneously spend that time on another task.
Sooner or later, if the volume keeps growing while the capacity to respond does not, an apparently obvious solution will start to look attractive: stop accepting so many questions. I have heard it said already.
I think that would be the wrong answer.
We should be extraordinarily careful about responding to technology that makes government easier to scrutinise by making government harder to scrutinise. But it is equally wrong to pretend its capacity is infinite.
Both things can be true. More transparency is good. Government capacity is finite. Scrutiny is essential. Scrutiny has a cost.
We must resist confusing a temporary technological imbalance with evidence that we have too much democracy.
The solution is not to disarm the questioner. It is to arm the government.
But merely buying the same tools will not be enough.
Edison’s practical electric light arrived in 1879, and engineers soon anticipated a revolution in manufacturing. Yet electrification did not immediately generate the productivity transformation they expected.
The economist Paul David explained why. A steam-age factory was built around one enormous engine. Power ran along a central shaft and then through belts and pulleys to every machine. Buildings and production lines were organised around the location of that shaft.
When electricity arrived, many owners did the obvious thing. They removed the steam engine and installed one large electric motor in its place. They kept the shaft, the belts and the old factory layout. It was cleaner and more convenient, but it did not transform productivity.
The breakthrough came when managers stopped asking how to replace the engine and started asking what a factory would look like if every machine could have its own motor. Factories could then be laid out around the flow of work rather than the location of a central source of power.
The redesign, not the motor on its own, created the transformation.
I think we are at the steam-engine-swap stage with AI.
Most of us, and I include myself, use it to perform the same tasks we always performed, only a little faster. We draft the email we would have written anyway. We summarise the report we would have read anyway. We research something that might previously have taken an afternoon.
That is useful. But it is still the big electric motor on the old shaft.
The transformational question is different: if we were designing this process today, knowing AI existed, would we design it this way at all?
Take Freedom of Information and Rule 14. If our response is simply to give an officer an AI assistant and ask it to help draft the same answer through the same process, we will gain some productivity.
But imagine redesigning the entire process. A question arrives and is automatically logged. Previous answers, published documents and related questions are identified. The system respects what the officer is entitled to access and, crucially, what they are not entitled to access. Material is assembled with its sources. Areas requiring judgement are identified. A draft is produced for a human to challenge and approve.
Repeated questions should also tell us something else: perhaps the information should simply be published.
At that point AI is not helping the old process run faster. We have redesigned the process.
The same applies in business. A finance company should not merely ask how AI can write today’s reports more quickly. It should ask why each report exists, who uses it, what decision it supports, where the information comes from, which parts require professional judgement and which parts need a human being at all.
This is a management question far more than a technical one.
And this is where the AI Sprint has made me particularly optimistic.
More than a thousand people took part. For an island of Guernsey’s size, that is a remarkable mobilisation. It came overwhelmingly from the private sector, from people who gave up time they did not really have to spare, shared expertise and worked across organisations.
It reminded me of the spirit Guernsey found during Covid, when the island pulled together because it had to. The difference is that this time we chose to.
Everyone who took part returns to work with new skills, new ideas and new contacts. I expect Guernsey will feel the benefit for years.
The value of the Sprint is not confined to the event itself. Technological change spreads when someone returns to work and asks: why are we still doing it this way?
There is a tendency to think the AI race will be won by whoever can spend the most money on technology. History gives us reason to doubt that. The winners from electrification were not simply those who bought electric motors. The value came when organisations changed around the new capability.
That should give Guernsey confidence.
We cannot outspend London, Singapore or New York. We cannot build larger technology companies than the firms developing the most capable AI systems. Nor do we need to.
Our opportunity is to become exceptionally good at applying what already exists.
A small island where regulators, government and business can sit in one room, as they have during this Sprint, ought to be able to rethink faster than almost anywhere. That helps the businesses already here and gives new ones a reason to come.
The challenge now is not to let that energy dissipate into a thousand people writing their emails slightly faster. It is to turn experimentation into redesign.
That is especially important because our economic constraint is people. Every genuine productivity gain means economic activity that does not require another person to be recruited, accommodated and supported on an island where labour and housing are scarce.
We need AI partly because the people scrutinising government already have it. But we should not spend the next few years building an AI-powered version of today’s bureaucracy.
If we simply bolt an assistant onto existing processes, we may answer questions somewhat faster, produce documents somewhat faster and process correspondence somewhat faster. Useful, certainly. Transformational, no.
Our ambition should be to ask what a small jurisdiction would look like if its public services had been designed in an age when machines could read, retrieve, compare, summarise and draft information at enormous speed.
Some of the answer will be technology. Much of it will be permissions, information architecture, process redesign and leadership.
That makes our task harder. It also makes success vastly more valuable.
Over the next few years there will be an uncomfortable period in which individuals and small organisations can exploit AI much faster than large institutions. It will expose weaknesses in our systems. It may increase government workload before it reduces it. And it will force difficult conversations about where legitimate scrutiny ends and unreasonable demands begin.
We should have those arguments honestly. But we should not solve an information-age capacity problem by rationing democracy. Nor should legitimate concerns about security become an excuse for institutional inertia.
The greatest risk is not that Guernsey adopts AI too quickly. It is that the rest of the world learns to use it while we are still discussing whether we should.
Small jurisdictions have always lived with the disadvantages of scale. AI may be the first major technology in a generation capable of reversing some of them.
It can give a small business capabilities once reserved for a multinational. It can give an ordinary citizen research capabilities once reserved for an institution. And eventually it can give a small government the capability to serve an ageing population with a shrinking workforce.
But for a while, those things will happen at very different speeds.
That is the dilemma we need to prepare for: not fewer questions, but a government increasingly capable of answering them.
The speakers contributing to the Sprint are right about the extraordinary capabilities of these tools. But ultimately technology providers can only sell us the motor.
Redesigning the factory is our job.
Author note: I used AI to assist with parts of the research, development and editing of this article. The arguments, judgements and final text are my own.
You need to be logged in to comment.