Insurance is structurally suited to AI enablement — and structurally resistant to it
Insurance has every characteristic that should make AI enablement easy. Decision-rich workflows. High-volume, well-defined processes. Decades of historical data. A workforce that already thinks in models, probabilities, and risk-adjusted returns. Regulatory frameworks that explicitly contemplate model-driven decisions. The ingredients are unusually favourable.
And yet, of the insurance leaders we work with, almost all of them describe the same picture: extensive model use in pricing and reserving, bolt-on automation in claims and underwriting, and an operating model that is fundamentally unchanged from a decade ago. Combined ratios that move slowly. Operational expense ratios that resist meaningful improvement. AI initiatives that produce demos and pilots and rarely produce structural advantage.
The reason is that insurance has its own version of the structural lag — and the carriers that solve it first will compound advantage that competitors cannot close.
Is this you?
- You have actuarial models that are technically excellent and an operating model that does not act on what they say in real time.
- Your underwriting team uses an AI-assisted triage tool and the cycle time per case has barely changed.
- Your claims function has invested in straight-through processing and 70%+ of claims still touch a human at multiple points.
- You have a customer-facing chatbot in market and a customer satisfaction score that has not moved.
- Your regulatory reporting (Solvency II, IFRS 17, prudential returns) is a quarterly programme that exhausts the team.
- You are watching insurtechs deploy AI-native pricing and underwriting in lines you used to dominate.
- Your model risk function has just seen PRA SS1/23 and the EU AI Act land and they are not yet sure what your defensible posture looks like.
If three or more of these are true, you are in the right conversation.
Where we focus in insurance
Five priority value streams account for almost all of the structural opportunity in a typical large insurer. We sequence the work based on which value stream has the highest combined cost, risk, and customer-outcome impact for your specific situation.
1. Underwriting and pricing
The decision-density is enormous and most insurers are doing some version of it with models already. The structural opportunity is in the workflow shape: the underwriter still does the case, just with more inputs. The AI-native version routes routine cases through end-to-end, surfaces the genuinely ambiguous ones to underwriters with full context, and turns the underwriter's decisions back into structured training signal. Cycle time drops, capacity expands, and the model gets measurably better quarter over quarter.
Regulatory frame: EU AI Act high-risk classification potential for insurance pricing where it materially affects access. PRA SS1/23 model risk core territory. FCA SYSC. Consumer Duty for retail lines. Solvency II for the actuarial models that feed downstream.
2. Claims operations
The largest single cost line in most insurers and the workflow with the most legacy human-in-the-loop assumptions. The redesigned version segments claims by complexity at first notice of loss, handles the routine 60–70% end-to-end (subject to defined human review at material thresholds), and concentrates claims handlers on the complex cases, fraud edge cases, and customer-facing situations where their judgment actually matters. Combined ratio improves; customer experience improves; the team shrinks but each role becomes more demanding and more valuable.
Regulatory frame: FCA SYSC, Consumer Duty (claims handling is one of the most scrutinised areas), PRA SS1/23 for the triage models, DORA for any third-party claims platforms.
3. Distribution and intermediary servicing
Insurance distribution is data-rich (broker submissions, customer interactions, intermediary performance, market signals) and most carriers are reactive about it. The redesigned version is continuous: brokers and agents see real-time recommendations, declined business is analysed for missed opportunity, retention models trigger proactive outreach, and the relationship with intermediaries becomes data-driven rather than relationship-driven only.
Regulatory frame: IDD and equivalents, FCA SYSC, Consumer Duty.
4. Regulatory and prudential reporting
Solvency II, IFRS 17, prudential returns, ORSA, and the increasing weight of climate-related disclosure. Most insurers run this as a quarterly sprint with substantial manual data wrangling. The redesigned version maintains a normalised reporting layer fed continuously from source systems, generates anomaly detection on the underlying numbers, and compresses cycle time from weeks to days. The team's role shifts from preparation to attestation and judgment.
Regulatory frame: Solvency II directive, IFRS 17, BCBS 239 principles applied to insurance, EIOPA technical standards, PRA SS1/23 for the anomaly and reporting models.
5. Fraud detection and SIU operations
Insurance fraud is an extraordinarily well-suited AI use case (high volume, labelled outcomes available eventually, high cost per false negative) and most carriers are doing some version of it. The structural opportunity is the data flywheel: the model that learns continuously from SIU outcomes will be substantially better than the off-the-shelf vendor product two years from now. Few carriers have built the flywheel.
Regulatory frame: PRA SS1/23, FCA SYSC, GDPR for the data inputs, Insurance Fraud Bureau coordination where relevant.
What we actually do in an insurance engagement
Our work spans the same five enablement pillars as our flagship AI Enablement service, but tailored to insurance realities:
- Production function redesign — workflow rebuilds in BPMN 2.0, anchored to one priority value stream and sequenced from there
- Action-data layer architecture — built around policyholder wide-rows, claims event streams, and observable lineage from policy admin systems through to the model
- Decision systems and feedback loops — structured override capture from underwriters and claims handlers, decision logs queryable for any individual decision, continuous retraining
- Operating model and roles — first-line accountability, system supervisor roles for the model owners, exception handler career paths for underwriters and claims professionals
- Embedded governance — three-lines-of-defence for AI integrated with the actuarial function, evidence as a by-product of build, regulatory dialogue with the PRA and FCA built into the cadence
The difference in insurance is the long tail. Decisions made today have consequences that play out over years (claims, reserving, customer lifetime value). The data flywheel takes longer to spin up than in fraud or KYC. The redesign work has to be patient about feedback loops and disciplined about not over-optimising for short-term signal.
How a typical insurance engagement runs
Phase 1 — Diagnostic (Weeks 1–6)
We map your existing AI portfolio against the compounding test, triage your use cases against PRA SS1/23 and EU AI Act, run an honest current-state mapping of one priority value stream (usually underwriting or claims because the ROI is large and the data layer is partially understood), and produce a defensibility memo against your highest-risk in-production models.
Outputs: AI portfolio audit, regulatory triage, current-state map of priority workflow, defensibility memo, board-ready strategic narrative.
Phase 2 — Strategy & Blueprint (Weeks 7–14)
We design the future-state operating model for your priority value stream, including the action-data layer architecture (typically a customer-and-policy wide row joined to claims events), decision rights matrix, governance machinery, and the operating model implications. We embed second-line risk in the design phase and align early with the actuarial function.
Outputs: Operating model blueprint, redesigned workflow specification, data architecture, decision rights matrix, governance framework, sequenced implementation roadmap.
Phase 3 — Activation & Delivery (Months 4–18)
We embed alongside your operations, technology, actuarial, and risk teams to lead the rebuild. Data layer first, then workflow, then governance instrumentation, then the role design changes that hold it all together.
Outputs: Live redesigned workflow with measurable outcomes, action-data platform reusable across adjacent workflows, embedded governance machinery, named first-line owners, retrained underwriters and claims professionals in the new role design.
Engagement models
Insurance AI Diagnostic — £45k–£75k, 6 weeks Focused diagnostic on one priority value stream with portfolio audit, regulatory triage, and a board-ready strategic narrative.
Insurance AI Enablement Blueprint — £120k–£220k, 12–14 weeks The full Phase 1 + Phase 2 engagement. Operating model blueprint, redesigned priority workflow, data architecture, governance framework, and sequenced 18-month roadmap.
Insurance AI-Native Transformation Programme — £350k+, 9–24 months Strategy plus hands-on delivery across one or more priority value streams. We embed alongside your teams, lead the workflow rebuilds, oversee data layer implementation, and run the change programme.
Executive Advisory Retainer — £10k–£25k / month Senior advisory access for insurers already executing on an enablement strategy. Monthly working sessions, ad-hoc reviews, and direct support for the executive sponsor.
Why this work is different in insurance
A few honest observations:
The actuarial function is your most important ally — and your hardest internal stakeholder. Actuaries already think in models and probabilities. They have the technical fluency to be excellent partners on AI enablement work. They also have decades of investment in their own model risk frameworks and a justifiable suspicion of "AI" that does not respect those frameworks. The right move is to engage actuarial leadership early, treat their model risk discipline as the foundation rather than the obstacle, and build the AI work on top of it rather than around it.
The long tail of decisions matters. Underwriting decisions made today have consequences that play out over years. Claims patterns shift slowly. Customer lifetime value emerges over decades. The data flywheel takes longer to spin up and the temptation to over-optimise for short-term signal is dangerous.
Distribution complexity is real. Most insurers operate through multiple distribution channels with multiple intermediary structures. The data layer rebuild has to handle that complexity rather than pretending it isn't there. Customer wide-rows that work for direct retail will not work for broker-distributed commercial without modification.
Regulatory fragmentation is the hardest part of multi-jurisdiction work. A large insurer operating across the UK, EU, and US has to satisfy different supervisors with different expectations on AI governance. The framework we use — embedded governance, evidence as by-product, single decision log, regulator-specific reporting overlays — handles this without producing four parallel compliance streams.
Who this is for
We work best with insurers that meet at least three of the following:
- Gross written premium of £500M+ or comparable scale in other metrics — the structural opportunity is largest where the legacy operating model is mature
- Executive sponsor at COO, CRO, CTO, or Chief Transformation Officer level
- A real (not theoretical) AI ambition in the value streams that determine combined ratio and customer experience
- Regulatory exposure to PRA, FCA, EIOPA, or equivalent that makes governance non-negotiable
- Some existing AI portfolio to triage — we can work greenfield, but most of our value comes from being honest about what is compounding and what is not
Frequently asked questions
How is this different from your flagship AI Enablement service?
The flagship AI Enablement service is sector-agnostic. This is the same engagement structure with an insurance-specific lens: the value streams (underwriting, claims, distribution, regulatory reporting, fraud), the regulatory frame (Solvency II, IFRS 17, PRA SS1/23, EU AI Act, FCA Consumer Duty), and the sector-specific failure modes we know to avoid.
How does this work with our existing actuarial function?
Closely. The actuarial function is the most natural internal partner for AI enablement work in insurance. We treat their model risk discipline as foundational rather than parallel — the AI work builds on it, not around it. In most engagements, the actuarial leadership becomes one of the most important sponsors of the work.
Do you work with both life and P&C carriers?
Yes. The workflow patterns are different (claims dominates P&C; reserving and policyholder lifetime dominates life) but the structural framework is the same. We tailor the engagement to your line mix.
What about Lloyd's syndicates and the specialty market?
Yes. The specialty and Lloyd's market has its own complexity (broker-led distribution, complex underwriting authority, syndicate structures) and we have run engagements in that environment. The value streams that matter most differ from retail P&C — underwriting decision support, broker servicing, and regulatory reporting are usually higher priority than claims automation.
How does this fit with PRA SS1/23 and the EU AI Act?
Both are treated as the supervisory baseline rather than retrofitted compliance. By the end of Phase 1 you should be able to walk the PRA or your home regulator through the redesigned workflow with full evidence.
What if we are mid-acquisition or post-merger integration?
This work fits well with M&A integration. The AI enablement design becomes the target operating model for the integrated entity, and the data layer rebuild becomes the integration platform. Multiple of our most successful insurance engagements have been anchored to post-merger integration programmes.
What this looks like in practice
For an anonymised example of this engagement structure in a real insurance environment, see our case study on rebuilding claims operations at a specialty insurer. It walks through the 16-month engagement: the starting position (cycle times double the market average, NPS stuck for three years despite a chatbot rollout), the diagnostic findings, what we redesigned across the five enablement pillars (with the actuarial function as a key partner), and the outcomes that landed (58% of claims handled end-to-end, 11-day cycle time vs 28, +22 NPS points on closed claims, 100% of decisions reconstructable on demand).
Start here
The first step is an executive working session — 90 minutes, no deck, no pitch. We use the time to understand your current operating model, your AI portfolio, the regulatory environment you operate in, and the value streams where the structural opportunity is largest. If we are a fit, we scope the diagnostic. If we are not, we say so.
For supporting depth, see the pillar essay on what AI enablement actually means, the FS Sector Playbook, and the AI Enablement Maturity Diagnostic.