Artificial intelligence was not simply part of the technology stack this year.

Increasingly, it was the risk being insured.

MGA Lab received nine applications focused on AI liability, making it one of the clearest emerging themes in the applicant pool.

As companies deploy autonomous and semi-autonomous AI systems, questions around responsibility become harder.

What happens when an AI agent makes a costly decision? Who is responsible when an automated system gives incorrect advice, causes financial loss, generates harmful content, or acts beyond the expectations of the business that deployed it?

Those scenarios do not always fit comfortably inside traditional liability products.

That creates room for insurers and MGAs capable of understanding the behavior, deployment, and underlying risks of AI systems in a more granular way.

One company selected for the cohort, Ollive, is building purpose-built liability insurance for AI agents and applications.

AI liability is still an early market. But nine applications in a single MGA Lab cycle is a meaningful signal that founders increasingly believe it can become a distinct insurance category.

Auto insurance is not new.

The opportunity may lie in serving the parts of it that broad-market products do not serve particularly well.

Several applicants approached auto through narrowly defined customer groups or operating models.

Three members of the 2026 cohort demonstrate just how different those niches can be.

TaxiFair has built experience in taxi insurance in Ireland and is now preparing to bring that model to New York for taxi and rideshare drivers.

LOOP is focused on personal auto and seeks to price customers without relying heavily on long-standing credit histories.

Shuttlebee approaches commercial auto through student transportation, combining transportation operations, risk-prevention technology, and underwriting.

None of these companies is trying to reinvent the idea of auto insurance.

Instead, each starts with a more specific question:

Can we understand and serve this group of drivers better than a generalist insurer can?

That same pattern appears across several other categories in the cohort.

One of the more interesting findings from this year’s applications was the background of the founders themselves.

Experienced insurance professionals and technologists were roughly equally likely to apply to build an MGA.

That matters because modern MGA formation increasingly demands capabilities from both worlds.

Insurance experience helps founders recognize where existing products, pricing assumptions, underwriting practices, or distribution models are falling short.

Technology can create access to new datasets, automate manual processes, improve risk selection, or make previously uneconomic niches viable.

The opportunity often sits between the two.

Some founders are experienced insurance operators applying new technology to familiar risks. Others are technology entrepreneurs building around new data sources and then developing the underwriting capabilities required to turn those insights into insurance products.

The best MGA businesses may increasingly be built by teams capable of doing both.

One part of the market was noticeably quieter.

Life & Health represented less than 10% of MGA Lab applications.

That is a stark contrast with liability, property, and auto, where founders identified numerous specialized opportunities.

There are plenty of possible explanations. Product development cycles can be different. Distribution is different. Regulation and capital requirements can present different barriers.

What the application data tells us more clearly is simply that fewer entrepreneurs are currently choosing to build there.

The final cohort nevertheless includes two companies exploring the category.

Kaleido Life is developing a model that gives customers access to a portion of their death benefit while they are still alive.

Halo is building around life insurance and tax-advantaged investment products.

Their inclusion illustrates an important point: lower startup activity does not necessarily mean lower potential.

Sometimes it means the barrier to creating something new is higher.