The industry spent 2023–2025 deploying generative AI as a productivity layer — drafting policy wordings, summarising claims notes, powering customer service chatbots. The ROIC has been real but modest: 10–20% efficiency gains within functions, no structural change to the combined ratio.
The strategic discontinuity arriving between 2026 and 2030 is categorically different. Autonomous agent systems — capable of perceiving context, making sequential decisions, using tools, and coordinating with other agents — do not improve insurance workflows. They replace them.
The Core Thesis
The question for insurance leadership is no longer whether to deploy agents, but how fast, how deep, and under what governance architecture. Carriers that architect toward Optimal Autonomy will structurally reduce combined ratios by 8–15 percentage points by 2030. The window to set the architecture is now.
Why this window matters
~2.3%
Real premium growth forecast 2026–27
Swiss Re Sigma. In a softening market, profitability cannot be earned through rate alone.
$100bn+
Global insured NatCat losses
Five consecutive years. Real-time data-driven underwriting becomes structurally critical.
Aug 2026
EU AI Act enforcement deadline
High-risk AI systems — including insurance pricing and risk scoring — face full compliance obligations.
"CEOs who defer architectural decisions in the expectation of greater market clarity are not managing risk — they are accepting structural disadvantage."
Defining Optimal Autonomy
Maximum autonomy — 100% automated decision-making — is neither achievable nor strategically correct within this planning horizon. Novel risk classes, systemic catastrophe events, litigation exposure, and regulatory requirements for human oversight create hard limits.
Definition
Risk-aware automation characterised by a calibrated balance of human and automated judgment, ensuring that decision-making is optimised for both speed and efficiency and long-term loss-ratio integrity. Autonomy is expanded precisely to the boundary where human judgment adds more value than it costs — and no further.
The most useful strategic frame is the shift from Human-in-the-Loop — where people are embedded in every decision — to Human-on-the-Loop, where people govern, audit, and intervene, but do not execute routine decisions. Across a hypothetical £5bn GWP P&C carrier, this shift implies a structural combined ratio improvement of 8–14 percentage points by 2030.
Three horizons to Optimal Autonomy
Horizon 1 · 2026
Task-Specific Agents
Efficiency and ROIC
CR: 92–97% → 88–94%
Purpose-built agents automate individual tasks within existing workflows. Claims intake and triage, underwriting data extraction, fraud signal detection, personal lines straight-through processing. Eliminates the 60–70% of processing time that is data assembly rather than judgment.
Target KPIs
Expense ratio –3 to –5pts
STP rate personal lines >80%
Motor claims cycle: 14 days → <3 days
ROIC on H1 investment >35% within 18 months
Horizon 2 · 2027–28
Multi-Agent Orchestration
Process Transformation
CR: 88–94% → <92%
Specialised agents communicate, delegate, and coordinate across functions without human intermediation. An orchestrator layer governs task decomposition and escalates to humans when confidence falls below threshold. The claims lifecycle, underwriting workflow, and distribution journey reconstitute as agent-native processes.
Target KPIs
Loss ratio –4 to –7pts via leakage reduction
Zero-touch claims >60% of volume
Customer retention +5–8pts
Combined ratio target <92%
Horizon 3 · 2029–30
The Autonomous Carrier
Business Model Reinvention
CR: <88% → <85%
A new operating model built around agent capabilities rather than adapted from legacy processes. Parametric products settle within minutes of event confirmation. Embedded distribution eliminates broker intermediation for vanilla products. Continuous usage-based underwriting reprices coverage in near-real-time.
Target KPIs
Parametric GWP 15–25% of portfolio
Zero-touch settlement >80%
Embedded channel 20–30% personal lines GWP
Autonomous carrier CR target <85%
Peer analysis — where the leaders stand
The following assessment draws on full-year 2025 results and strategy disclosures from AXA, Allianz, Zurich, Chubb, and Aviva. The Evident AI Insurance Index 2025 independently ranks AXA and Allianz as the only two carriers in the top five across all four pillars — against an industry average score of 35.5.
AXA
H2 Leading2025 CR: ~93%Lead use cases
400+ use cases; SecureGPT; real-time IoT risk scoring; claims triage
Strategic moat
900 data scientists; Group CDAIO role; Stanford HAI partner
Gap / risk
Agentic orchestration still fragmented across 50 markets — governance inconsistency risk
Allianz
H2 Leading2025 CR: 92.2%Lead use cases
900+ use cases; Anthropic partnership; claims automation; dynamic pricing pilots
Strategic moat
Record €17.4bn profit; €2tr AUM data asset; new digital board exec
Gap / risk
Explicitly cautious on autonomy pace; regulatory complexity in Germany/EU
Chubb
H2 → H32025 CR: ~93%Lead use cases
85% UW/claims automation target; Chubb Studio 250+ embedded partners; $1.4bn digital GWP
Strategic moat
3,500 engineers globally; Chubb Studio moat; 20% workforce restructuring underway
Gap / risk
Execution risk in 20% headcount reduction; cultural change management
Zurich
H1 → H22025 CR: ~94%Lead use cases
160+ AI solutions; Agentic Hyper Challenge (200 prototypes); 13-min claims resolution (UK)
Strategic moat
AI Lab (ETH Zurich / St Gallen); new Chief AI Officer Oct 2025
Gap / risk
Moving from pilots to scale; not yet committed to explicit CR targets from AI
Aviva
H12025 CR: 94.6%Lead use cases
12 AI fraud models; £90m+ claims savings; claims summarisation; medical UW tools
Strategic moat
25m customer data asset; Direct Line acquisition adds scale
Gap / risk
End-to-end transformation still aspirational; behind leaders on architecture
Three structural moats in the agentic era
Raw access to AI models is not a durable competitive advantage — foundational model capabilities are available to all carriers. Three categories of structural moat will determine which carriers emerge as durable leaders.
01
Proprietary data
Long-tail claims loss triangles (10–20 years of internally generated data not purchasable on the market), real-time IoT and telematics streams, and behavioural data training retention agents. Aviva's 25m customers and Allianz's €2 trillion AUM are data assets in the actuarial sense. Data acquisition should be treated as a capital allocation priority on a par with technology infrastructure.
02
Governance as competitive licence
Counterintuitively, the deepest moat may not be the most aggressive automation — it will be the most trusted. Explainability infrastructure, model cards, and a formally documented agentic authority framework become differentiators with sophisticated commercial clients. Carriers that build governance before they need it will expand autonomy faster than those who retrofit it under regulatory pressure.
03
Speed to settlement as brand promise
Carriers achieving verifiable zero-touch settlement for standard claims can make this a client-facing performance commitment — differentiating on service quality rather than price. Zurich's 13-minute UK contents claim resolution is an early proof point. At scale, this becomes a structural retention advantage that a competitor cannot replicate without matching the underlying data and agent architecture.
Call to action — the next 18 months
By the end of 2027, the agentic architecture decisions made today will be embedded in systems, data pipelines, vendor contracts, and organisational structures that are costly and slow to reverse.
Architecture: build or procure?
The most consequential choice is whether to build a proprietary agentic architecture or deploy on third-party platforms. Proprietary is slower and more expensive in Year 1. It is the source of durable differentiation by Year 3. This is a CEO-level capital allocation decision — not a CTO matter.
Governance before scale
Build the governance architecture — explainability tooling, autonomy authority frameworks, model monitoring, regulatory engagement protocols — before expanding agent deployment beyond Horizon 1. The August 2026 EU AI Act deadline is an immediate compliance obligation for any carrier with European operations.
Reskilling, not retrenchment
The primary failure mode in insurance transformation is not technical — it is organisational. Underwriters whose role shifts from decision-maker to model governor require structured reskilling, not redundancy. Carriers that invest in this transition retain institutional knowledge that cannot be recovered once lost.
The 18-Month Imperative
The Autonomy Frontier is being drawn now. The technology is available. The macro tailwinds are present. The regulatory framework, whilst demanding, is knowable and manageable. What separates the leaders from the followers is not capability. It is conviction.
This Strategic Research Note has been prepared by Eudaimon Consulting, March 2026. Market impact estimates are analytical projections derived from publicly available industry benchmarks and published 2025 annual reports. Carrier assessments are based on public 2025 disclosures only. This note is for general information purposes only and does not constitute investment, legal, or regulatory advice. © 2026 Eudaimon Consulting. All rights reserved.