Coming from the insurtech and product world, my mantra has always been “move fast and break things”, especially with cutting edge or as yet unproven approaches. That playbook simply won’t fly in a regulated environment, yet it is exactly what’s required to unlock real success with AI.
What is a responsible yet innovative playbook for AI in insurance?
One answer might be the product practice of ‘Continuous Discovery’ – introduced by Steve Blank in ‘Four Steps to the Epiphany’, and made famous by Eric Ries in ‘The Lean Startup’ – where teams continuously engage with customers through interviews or prototypes, to validate assumptions and guide product decisions. Most discovery playbooks are written for software startups.
But insurance doesn’t operate in a “move fast and break things” environment; our regulatory, governance and stakeholder commitments can’t be compromised.
So how do we adapt Continuous Discovery to AI in Insurance?
1. Redefine “fast” : Fast isn’t just shipping code, it is about shortening the path to customer value. Rather than the product-world approach of building ten things to find one that sticks, figure out the shortest path to understanding if the user derives value. Pick a shallow but wide use case, and embed it to get quick feedback.
2. Measure Trust, not just Usability : In the Data and AI world, trust is far more valuable than cool features. In normal discovery we test “Will they use it?”. For AI, we must also test “Will they trust it?”. It’s not enough for a model to be accurate – users and regulators must understand and feel confident in its outputs. Explainability is a non-negotiable part of the user experience.
3. Embed Compliance in the Discovery team : In tech, the holy ‘Discovery Trio’ is the Product Manager, the Designer, and the Engineer. In Insurance, it’s a Quad that includes Risk and Compliance, so risks are surfaced early, before reputational or regulatory damage occurs, and governance is baked in.
4. Use “Opportunity Solution Trees” : Another Silicon Valley framework, OSTs are designed to stop tech teams from leaping straight to solutions, or trying to fit shiny new tech to the problem. Define the Outcome (the “trunk”), identify Opportunities (“the branches”), generate Solutions (“the leaves”). Then test the desirability, usability and feasibility of each solution idea. OST shifts the mindset from “What should we build?” to “What opportunities for value exist, and which solutions best address them?”. This is especially relevant as we try to fit LLMs to every problem, even those that are better solved with more traditional methods.
Insurance doesn’t have to choose between innovation and governance.
In AI, the winners are not the smartest models, but the most trusted experiences. Continuous discovery, adapted thoughtfully, is how we get there.
