AI in insurance: budget and strategy

Many insurance executives have quietly carried the same worry over the past year: the AI age has arrived, everyone is investing, how much are we investing? What’s the right ratio? Where is our budget flowing? In markets like Turkey this isn’t yet discussed out loud, but in budget meetings, in the quiet bargaining between IT and finance, it’s already there behind the curtain.

There are really two separate budgets

Here’s the point: every insurer’s technology budget is really two budgets. The first is what’s spent keeping old systems alive, the line the industry calls lights on. The old policy-administration system, the old underwriting engine, the old claims process, the old integrations; all of it eats money minute by minute. The second is strategic investment: new architecture, new data infrastructure, new customer experience, and AI.

At most companies the split between these two is 70-30 or 80-20 in favor of lights-on. So only a fifth of your budget goes to the future; the rest keeps today standing. There are companies that have raised this ratio to 50-50 over three years, and without breaking the combined ratio at that; but in this industry that’s a rare example of discipline.

It helps to make this concrete. Lights-on usually means keeping alive a core system built years ago, poorly documented, made more complex with addition upon addition. You can’t stop it, because the policies sit on it; but feeding it gets more expensive every year. That is exactly the silent cost eating into the strategic budget.

A two-sided trap

The trap here runs two ways. On one side are companies that ignore the lights-on burden and launch a big AI project; these crash when they hit the old systems. On the other are companies that spend years only on legacy modernization and lose their capacity to innovate; by the time modernization is done, the technology age has already moved on. Both extremes lose.

The AI age doesn’t ease this balance, it sharpens it. Because AI demands new strategic investment and at the same time grows the lights-on budget to make old systems AI-ready. So the question is no longer will we invest; it’s how we’ll manage the ratio between two budgets. The number that decides is not the size of the budget, but its ratio.

That ratio, not size, is decisive is actually good news for smaller players. A mid-sized company that can devote half its budget to the future can move faster than a giant spending all its revenue feeding old systems. Because transformation is not a spending race, it’s a race of focus and discipline. The right ratio is often worth more than a big budget.

A concrete example of AI growing both budgets is data. For the model to work, data must be clean, organized and accessible; yet data in old systems is often scattered and inconsistent. So before starting an AI project, most institutions unknowingly enter a big data-cleanup and infrastructure job. That means part of the new investment actually goes to fixing the old.

The picture in Turkey

A 2025 survey by Turkey’s AI Initiative (TRAI) laid out, for the first time at corporate scale, the country’s AI pulse in this much detail. The study, with 126 institutions, shows this: while most have put AI on their agenda, institutionalized strategies are still limited. The most-invested areas are chatbots, digital assistants, forecasting and quality-control projects.

The rest of the picture is familiar too. Agentic AI and copilot applications are spreading fast; systems that don’t just chat but execute tasks are being built. Efforts to build an AI-ready workforce are rising, and the most critical headings are awareness, training and cultural change. The survey’s summary is clear: AI transformation moves not through random experiments, but through a clear strategy, the right team and value-creating use cases.

So where to start? The picture says the most value is in visible and frequent problems: speeding up claims, supporting customer service with AI, catching fraud early. These give a measurable return and build the team’s confidence. Rather than a big, abstract AI vision, focusing on a few problems with clearly visible outcomes is the soundest start to transformation.

Another critical point is the team. As the survey notes, the hardest part is not technology but culture and capability. The most advanced tool bought without building an AI-ready workforce just sits on the shelf. What I’ve seen in my own experience: what turns technology into product is not the team that writes the code, but the team that learns to fold that code into its daily work. So part of the investment must always go to people and training.

There’s a danger to name too: pilot purgatory. Many institutions get stuck in trial projects that never reach the field. The demos are impressive, the slides polished; but none of it enters daily operations. The way out is to tie every project to a measurable outcome from the start: which cost will it cut, which process will it speed up, how will we measure it? A project without a clear answer usually stays a showcase that drains the budget.

My reading

These two pictures complete each other. On one side the question of budget ratio, on the other the maturity of strategy and team. The real opportunity, especially for emerging markets, is to shorten the years-long roadmap of mature markets: move straight to ready infrastructure and models, and focus limited resources on the right few problems. Because what wins here is not the biggest investment, but the right ratio and the clearest strategy.

In short, insurance’s invisible test today is not investing in the wrong technology; it’s setting the wrong split between the two budgets. A small but well-aimed start often delivers faster than a scattered, showy budget. And digital transformation is not a destination but a continuously managed balance: building the future while keeping old systems alive is like carrying two pans of a scale at once. The institutions that manage it turn technology from a cost line into a competitive tool.