Stop Pricing the Work Your AI Is Built to Kill: Simon-Kucher's Andrew Larsen

Here's a problem most insurtech founders haven't sat with long enough: if your AI actually works, your customer needs fewer people. If you charge by the seat, your revenue goes down. Your best product outcome becomes your worst pricing outcome.

That's the tension Andrew Larsen spends his days on. Andrew is a Director in the insurance practice at Simon-Kucher, and he was part of the team leading Simon-Kucher's ITC Vegas 2026 masterclass, "Winning Monetization and Sales Strategies for Insurtechs in the Age of AI." I couldn't sit down with him at ITC this year, so I sent him seven questions instead. His answers are blunt, practical, and worth reading whether you're building insurtech, buying it, or selling it to carriers and agencies.

1. What's the biggest monetization mistake you see insurtechs make when they bolt AI onto an existing pricing model instead of rethinking it?

They add AI to the product without asking what the customer is actually paying for.

The clearest example is keeping a seat-based model when the AI exists to make each employee more productive or reduce headcount. If the product works, seats go down and revenue goes down with them. The pricing model is fighting the value proposition.

The opposite mistake is metering every AI interaction or token. Customers do not care how many tokens your model consumed. They care about the business result, and insurance buyers in particular care about budget predictability.

The right question is which AI capabilities are simply part of the product and which create incremental, measurable value worth a scalable price component. Core AI belongs in the platform price. Genuinely incremental automation can carry a bounded consumption or value metric.

2. Once AI is doing work that used to justify the price (underwriting, claims triage, etc.), how should a growth-stage insurtech rethink its value metric?

Simple rule: do not price on the input your AI is designed to eliminate.

If ten underwriters will soon do the work of fifteen, pricing on underwriting seats is structurally fragile. The same applies anywhere AI compresses manual work.

Move the metric closer to the durable economic activity the platform enables: premium managed, policies processed, submissions evaluated, claims handled, lines of business deployed, programs managed. Something that grows when the customer's business grows.

In practice I prefer a base platform commitment plus one understandable scaling metric over a stack of meters. The metric does not need to capture every dollar of value. It needs to correlate with value, scale with the customer, be auditable, and stay predictable enough to budget.

Outcomes can play a role, but I would not tie the whole commercial model to something like loss ratio. Too many variables outside the software drive that number.

3. You talk about pricing evolving with AI-driven value. What does that actually look like in a real contract or term sheet?

Four pieces:

• A multi-year platform commitment. Budget certainty for the customer, a predictable base for the vendor.
• Core AI capabilities included in that commitment, not billed feature by feature.
• A bounded scaling component tied to business volume, deployed products or lines, transactions, or AI-driven actions. Bounded matters: the customer should be able to model the top end of its spend.
• Separate implementation or high-touch service fees where the vendor is committing scarce resources or accelerating time-to-value.

Then define explicit expansion mechanics. When usage crosses a band, a new line of business launches, or the customer hits an agreed deployment milestone, the commitment steps up or resets at renewal.

What I would not do is hand a CFO an uncapped token meter and tell them to see what the bill looks like in December.

None of this requires exotic contracting. It is familiar enterprise software machinery (commitments, bands, ramps, true-ups, multi-year terms) attached to a value metric that still makes sense in an AI-enabled operating model.

4. Where are you seeing AI create real, defensible pricing power today, versus where it's mostly marketing?

The line is between AI that changes an economic or risk outcome and AI that makes an existing workflow slightly nicer.

In insurance, real pricing power shows up where AI demonstrably improves risk selection, quote-to-bind performance, underwriting throughput, claims handling economics, or the premium a team can manage without proportional headcount growth.

Speed also carries pricing power. If technology compresses an implementation, underwriting, or claims process and the customer starts realizing economics months earlier, they can put a dollar value on that.

Defensibility strengthens when the AI is embedded in the workflow with access to proprietary context across the process. A generic copilot sitting beside the workflow is easy to replicate and hard to monetize for long.

Summarization, generic chat interfaces, and an "AI-powered" label on an existing feature are table stakes. Buyers may like them. That does not mean they will pay extra for them.

5. What's an AI-driven risk to monetization that most commercial leaders aren't watching for yet?

Successful AI can become revenue-deflationary under the wrong commercial model.

Your product automates more work. Customers need fewer seats. Implementations need fewer services hours. Model costs keep falling. You have created far more customer value while shrinking two of the things you traditionally charge for: users and labor.

That is a good product outcome and potentially a bad monetization outcome.

Commercial leaders need to decide now what their durable monetization base will be once "number of humans touching the workflow" stops being a useful proxy for value.

The companies that get this right will have models that participate in the growth or throughput their AI enables, not models that depend on the labor it replaces.

6. When sales incentives and pricing fall out of sync on an AI-enabled product, what's usually the first symptom leadership notices?

Deal inconsistency, well before churn.

Two customers with similar economics end up with materially different pricing because sellers are optimizing for signature, not for preserving the monetization model.

You see more one-off discounts, waived platform fees, free services, bespoke usage exceptions, and reps giving away future AI consumption because comp rewards today's ACV and ignores the quality of the commercial structure.

At that point the price book stops predicting what customers actually pay.

Expansion and margin problems follow, but the early signal is price realization dispersion and a growing pile of exceptions nobody can explain economically.

If the product is meant to land and then scale with usage or realized value, comp has to reward that behavior rather than reward monetizing everything upfront or discounting away the expansion mechanism.

7. What's one monetization pitfall you've seen actually kill an otherwise good insurtech deal?

The "pay twice" problem.

An insurtech can have a strong product and compelling standalone economics, but if the customer keeps paying its incumbent core platform while also paying for the new software, implementation, and integrations, the business case collapses.

This happens when the product is sold as an incremental layer with no credible path to decommission or materially reduce the incumbent cost base.

The vendor thinks it is selling a better product. The CFO sees two software bills, more implementation expense, and no budget freed up.

Discounting rarely fixes it, because sticker price was never the issue. The commercial architecture never gave the customer a way to stop paying for the thing being replaced.

For transformational insurtech products, monetization strategy has to include displacement economics. A technically successful deployment with no economic path off the incumbent is still a bad deal.

The InsNerds take

If you're on the carrier or agency side, use Andrew's answers as a buyer's checklist. Ask your vendors three things: What happens to my bill if your AI works and I need fewer people? What's the most I could possibly pay you in a year? And what am I going to stop paying for once this is live? If they can't answer those cleanly, the problem isn't the product. It's the deal.

About Andrew Larsen
Andrew Larsen is a Director at Simon-Kucher within the insurance practice, with expertise in growth strategy, operating model design, and digital transformation. He has led large-scale initiatives across financial services, insurance, and brokerage, delivering measurable impact in revenue growth, cost optimization, and AI-driven innovation. His experience spans advising Fortune 500 companies, private equity portfolio firms, and emerging tech players on go-to-market strategy, operating model redesign, and AI. Connect with Andrew on LinkedIn and learn more at simon-kucher.com.