I have spent more than 18 years building core insurance platforms for insurers, MGAs and brokers. Here is what that time has taught me about the technology that holds up, and where AI honestly fits.

I have been in insurance software long enough to have watched the same film several times. Only the title changes. First it was “digital,” then “cloud,” now everything is “AI-native.” Each wave arrives with the same promise: that someone has finally found a way to replace the core of an insurance business overnight, and that the people who built it patiently over decades are about to be swept aside. A great many of the companies making that promise raise impressive sums, ship a beautiful interface, win a few names, and then quietly get acquired, pivot, or disappear. The pattern is consistent enough that I have stopped being surprised by it. What still surprises me is how reliably the market forgets it happened.

I think forgetting is the interesting part, because it tells you something about how this industry mistakes motion for progress. Insurance is, at heart, a promise to be there years from now, when something has gone wrong for someone, and everything underneath it, the policy, the pricing, the claim, exists to make sure that promise holds. You build into a thing like that slowly, patiently, and with respect for how hard it is.

Why insurance resists the overnight story

The people promising disruption underestimate exactly this. A policy is not a checkout flow. It is a legal instrument that has to stand up years after it was sold, under regulation, under audit, under a claim nobody anticipated. So the polished new tool performs in the demo and stalls in production, because handling the easy ninety percent was never the hard part.

What a failed core platform actually costs you

And when it stalls, the vendor does not pay for it. You do. A failed migration is your eighteen months and your team’s exhaustion. A core system that cannot keep up is your claims leakage, your slower settlements, and your renewals quietly drifting to a competitor who answered first. The cost of someone else’s broken promise lands on your loss ratio and your book of business, not theirs. That is why insurers, MGAs and brokers buy on ten-year horizons rather than on demo-day excitement. The platform you choose now will shape how your business runs for the next decade. So the real question is not “what can it do this quarter.” It is “Will this still be carrying my business, on my terms, long after the launch?” It is the least glamorous question in the room and the only one that protects your numbers.

I built tigerlab around that question, and I have spent the past 12 years answering it. We have never chased trends; we have built with intent. We made the tiger suite a cloud-native, API-first core platform for the people who actually carry the risk, insurers and MGAs, brokers, underwriters, pricing and claims teams, and we kept our focus on the complex commercial and speciality markets rather than spreading ourselves thin, because depth has always protected customers better than breadth.

How we build, and who we build with

Independence and focus only matter because of what they let us build, and that comes down to one belief we have held from the start. The market loves the shortcut: intelligence bolted onto someone else’s engine, a thin layer that demos beautifully but does not hold, because it never understood the business underneath it. There is nothing wrong with speed itself; the shortcut fails because it stays shallow, sitting on the business without ever learning it. Our platform is API-first and modular, so it can move quickly while still being built into the core of how you work. That understanding is the whole job, and it cannot be retrofitted after launch through a survey and a list of complaints.

So we begin where the real problem is, with the people who will actually use the system. Your teams help shape the process itself, while the software is still being designed and there is still a chance to get it right. We measure our work by your results: higher renewal rates, faster claims handling, a growing book of business. Those are the goals that matter, and we hold ourselves to the same ones.

That way of working only holds if the people doing it can speak both languages. It is easy to find engineers who write excellent code and never grasp why a renewal slips or where a claim leaks. Ours are expected to cross that line, to move from building systems to understanding the business those systems carry, so that real underwriting judgment ends up in the architecture rather than lost in translation. The result is software configurable by your own people without joining a change-request queue, with the logic living inside the system rather than in a layer that passes data between tools. Every decision is traceable, and your data stays within your environment. Under GDPR, DORA and a regulator’s audit, that is not a nicety. It is the difference between a system you can defend and one that becomes a liability the first time you are asked.

The pattern shows up in the numbers. On one book we can point to, gross written premium had been declining month after month. It bottomed in the month of go-live on the tiger suite, then climbed past every earlier figure in the year. Software alone does not move a book of business. But this is the kind of turn the platform is built to support, and it happened where the system met the work, not in a demonstration.

Why I stay calm about AI

This is also why the noise around AI does not unsettle me. I will not claim it will replace your underwriters next year; I have heard that promise about other technologies, and it has never been true. What I do believe is this: most insurance AI pilots fail, and not because the models are weak. They fail because they sit on foundations never built to carry them, and a model that dazzles in isolation is worthless if your platform cannot feed it clean data or act on what it finds.

AI exposes whatever architecture you already have.

For us, AI is not a new direction. It is the latest test of the same conviction. We embed intelligence where the decisions are actually made, inside the rating and underwriting workflow, so its output is traceable, its actions accountable, and the routine administrative weight finally lifts off your underwriters’ desks, the very work that slows your quotes and frustrates your clients. The model only ever mattered for the business it serves.

That is what we mean when we say our purpose is to make insurance intelligent, resilient, and crafted by experts. Intelligent enough to take the routine off your team’s desk. Resilient enough to outlast the next wave, and the one after it. Crafted by people who understand that this industry runs on trust and on the experts who hold it.

If you are choosing a platform to build the next decade on, you are really choosing a partner: someone who will still be standing with you when the launch is forgotten, and the renewals are due, someone who builds your business into the system instead of renting you a layer on top, and someone who measures success by your numbers. We built tigerlab to be that company. I would welcome the conversation.

Tobias Bergmann, Founder and CEO, tigerlab