Insurtech on the Silicon Prairie: AI Is Inevitable - ROI at Scale Isn't (Yet)
On stage in Omaha, a simple question cut through the hype: Who's getting positive ROI from AI at scale? No one raised a hand. That wasn't defeatism - it was an honest baseline from CIOs who are working hard to get there. The takeaway: adoption is real, but meaningful returns still require focus, infrastructure, and patience.
That tone set the pace for this year's Insurtech conference, where 540 insurance leaders, regulators, and founders converged. The mood was clear: AI will matter more in insurance, but progress comes from execution, not slogans.
What Will Actually Move the Needle
Leaders are prioritizing AI that improves productivity and reduces friction across existing workflows. Think agents that assist CSRs, generate illustrations, and keep customers informed on open requirements. These are measurable wins, not science experiments.
If you're leading an insurance team, the question isn't "Should we use AI?" It's "Where can we compress cycle time and costs without risking compliance or customer trust?"
The Trust Gap Is Real - AI Can Help Fix It
Consumers don't love insurance. Complexity, poor communication, and slow processes erode confidence. That's not news to anyone in this industry.
AI won't fix trust on its own, but it can simplify choices, personalize education, and make service feel responsive. Explain the product in plain language. Show what's missing on a claim. Keep them updated before they ask. That's how you rebuild credibility.
Startups: Don't Try to "Disrupt" the Carrier-Broker Model
The carrier-broker-customer model isn't going away. The industry is a fortress. The fastest path to revenue is improving how the fortress operates - underwriting, claims, distribution, and service - not trying to blow it up.
Founders who win in insurance learn how decisions actually get made. Ask your buyer what they're measured on and what their bonus depends on. Then build for that.
Practical steps for startups
- Sell efficiency first: cycle-time cuts, straight-through processing, loss-adjustment expense reduction, premium growth per producer.
- Design pilots with strict scopes: one LOB, one process, 60-90 days, clear baselines and success criteria.
- Avoid one-off custom features that break your roadmap. Multi-tenant first, optional extensions second.
- Prep for security and vendor risk reviews upfront: SOC 2, data handling, model governance, PII controls.
- Price on outcomes where possible (per claim closed, per illustration, per bound policy) to align incentives.
- Map the buying committee early: IT, data, compliance, legal, business owner, and procurement.
Carriers: Be a Better Partner (and Protect Yourself)
Speed helps everyone. Give quick "no's" when there's no fit. Create a pilot lane with pre-approved legal, security, and procurement guardrails so teams can test faster without cutting corners.
And be careful not to customize a startup into a corner. If they can't sell to anyone else, you may lose the vendor right when you need them.
Practical steps for carriers
- Publish problem statements with sample data, KPIs, and constraints so vendors can self-qualify.
- Stand up a pilot program: capped scopes, fixed timelines, sandbox data, and clear exit criteria.
- Set non-negotiables early: explainability, audit trails, retention, model update cadence, and SLAs.
- Score vendors on viability: referenceable clients, unit economics, roadmap alignment, and funding runway.
- Limit bespoke builds; prefer configuration over code. Protect portability and avoid lock-in.
Regulators Are Allies - If You Bring Them In Early
State commissioners made it clear: they want innovation that treats customers fairly. The worst move is coming in as an adversary or surprising them with something that feels like a black box.
Bias risk is top of mind with AI. If a policy gets canceled or priced in a way that raises questions, someone needs to explain how the decision was made. That requires documentation, monitoring, and accountability.
Compliance checklist for AI initiatives
- Document model purpose, inputs, training data sources, limitations, and decision boundaries.
- Run bias and disparate impact testing; monitor drift; keep an audit trail of model changes.
- Disclose AI use where it influences pricing, underwriting, or claims decisions; provide paths to human review.
- Align with frameworks like the NIST AI Risk Management Framework (NIST AI RMF) and AI guidance from NAIC (NAIC AI Principles).
- Review third-party risk: data residency, subcontractors, incident response, and right-to-audit clauses.
Where ROI Is Most Achievable Right Now
Focus on high-volume, rules-heavy processes where quality and compliance are measurable. Small improvements at scale drive real dollars.
- Claims: intake, FNOL routing, document understanding, subrogation discovery, SIU triage.
- Underwriting: appetite and triage, broker submission cleanup, prefill, quote-to-bind assist.
- Service and distribution: agent co-pilots, knowledge retrieval, illustration generation, next-best action.
- Operations and IT: contract review, requirements traceability, code assistance during core modernization.
Metrics that matter
- Cycle time and touch reduction per policy or claim.
- Straight-through processing rate and rework percentage.
- Loss ratio or leakage impact where applicable.
- Cost per claim/policy, producer productivity, CSAT, and first-contact resolution.
- Regulatory exceptions, complaint rates, and audit findings.
The Bottom Line
AI at scale with clear ROI is still rare in insurance, but the path forward is known. Pick focused use cases, measure hard outcomes, engage regulators early, and build partnerships that survive procurement and pilots.
Do that, and the next time someone asks who's seeing ROI, more hands will go up.
Want to upskill your team for practical AI in insurance?
Explore role-based programs that align with real workflows and compliance needs: Courses by Job.
Your membership also unlocks:
