From tracking Everest Group PEAK Matrix rankings across four assessment cycles, the shift from process automation to decision intelligence as the defining vendor capability marks the clearest inflection point in P&C technology this decade. Process automation streamlined existing workflows. Decision intelligence changes which decisions get made, by whom, and on what evidence. That distinction reshapes vendor evaluation criteria, carrier procurement strategy, and the analytical frameworks actuaries use to assess technology-driven changes in underwriting accuracy and claims efficiency.
The Everest Group Top 50 P&C Insurance Technology Providers 2026 report, published in Q1-Q2 2026, assessed an initial universe of more than 200 technology providers. The final Top 50 recognized 38 providers across eight segments: underwriting orchestration, claims intelligence and decisioning, SaaS-delivered core platforms, pricing intelligence, intelligent workflow automation, property and location intelligence, digital experience platforms, and distribution and agency management. Separately, the PEAK Matrix assessments ranked vendors within specific segments, identifying Leaders, Major Contenders, and Aspirants based on a weighted composite of market success (30%), innovation and investment (30%), value chain coverage (20%), geographic reach (10%), and line-of-business coverage (10%).
Trade press covered individual vendor recognitions from these reports, but no outlet has synthesized the cross-segment findings into a unified view of where carrier AI spending actually lands and what it means for the vendor competitive landscape. This article does that synthesis, then translates the findings into the metrics that matter for actuaries evaluating vendor claims against production evidence.
The $16-18 Billion P&C Tech Ecosystem: What Eight Segments Reveal
Everest Group sizes the global P&C insurance technology market at $16 to $18 billion. That figure represents the first credible independent sizing of the ecosystem that goes beyond the insurtech venture funding numbers that dominate industry conference presentations. The $16-18 billion covers enterprise software, analytics platforms, data services, and workflow orchestration tools deployed by carriers, MGAs, and brokers across all lines of P&C business.
The eight-segment taxonomy itself tells a story about where value is concentrating. Five of the eight segments, underwriting orchestration, claims intelligence, SaaS cores, pricing intelligence, and intelligent workflow automation, are fundamentally about embedding AI into operational decision points. The remaining three, property and location intelligence, digital experience platforms, and distribution management, support the decision layer but do not define it. This structural emphasis explains why the report title foregrounds “decision intelligence” rather than broader categories like “digital transformation” or “modernization.”
Geographic distribution reinforces the North American concentration: 62% of Top 50 providers are headquartered in North America, with growing representation across Europe, the UK and Ireland, Asia-Pacific, and the Middle East and Africa. For carriers evaluating vendor partnerships, that distribution matters because regulatory requirements for AI governance vary sharply by jurisdiction, and a vendor’s home market shapes which compliance frameworks it builds into its platform natively versus which it bolts on.
Why Decision Intelligence Replaced Process Automation
The conceptual shift from process automation to decision intelligence is not just semantic rebranding. Process automation took existing carrier workflows and made them faster: faster data entry, faster document routing, faster status updates. The underlying decision architecture remained unchanged. A claims adjuster still decided the same things, just with less paperwork between steps.
Decision intelligence inverts that model. It treats data as live, structured intelligence rather than historical policy snapshots, enables event-driven cores that ingest continuous signals, recalculate exposure dynamically, and orchestrate rules-based and AI-driven decisioning in real time. The Everest Group report describes the shift as moving from an “inside-out, tech-first model” to an “outside-in, customer-driven operating model” in which multi-agent orchestration enables AI to move beyond isolated task automation toward coordinated, context-aware execution across the insurance value chain.
The production data supports this framing. Full AI adoption among insurers jumped from 8% to 34% year over year between 2025 and 2026. But adoption alone does not capture the shift. Celent survey data shows that 22% of insurers plan to have agentic AI in production by year-end 2026. The agentic AI insurance market is projected to grow from $5.76 billion in 2025 to $7.26 billion in 2026, a 26% growth rate that reflects carriers moving from experimental chatbots to multi-step decision agents that can execute across underwriting, claims, and pricing functions without constant human intervention.
For actuaries, the distinction matters because decision intelligence changes which variables enter the pricing and reserving models. When AI systems classify risks, triage claims, and identify fraud patterns in production, they generate structured decision logs that become inputs to actuarial analysis. Process automation never produced that kind of signal. It made existing processes faster but did not create new information. Decision intelligence creates new information at every step, and actuaries need frameworks for evaluating whether that information improves or distorts the models it feeds into.
Claims and Underwriting Dominate Live AI Production
The Everest Group report quantifies a pattern that industry conference presentations have claimed but rarely documented: 58% of live, production AI use cases across the Top 50 vendor ecosystem are concentrated in claims functions, and 46% are in underwriting. These percentages sum above 100% because some use cases span both functions. Pricing and distribution trail substantially, occupying smaller shares of production deployment despite receiving significant vendor marketing attention.
That concentration makes sense from a carrier economics perspective. Claims is where carriers spend the most money: loss adjustment expenses, indemnity payments, fraud investigations, and subrogation recovery. Any technology that compresses cycle time, reduces manual handling, or improves fraud detection has a direct and measurable impact on the combined ratio. Underwriting is where carriers make their money: risk selection, pricing accuracy, and portfolio construction determine whether the book of business generates underwriting profit or loss.
The production metrics from carriers that have deployed AI in these functions are specific enough to anchor actuarial analysis. Industry data shows STP (straight-through processing) rates for claims jumping from 10-15% to 70-90% where AI triage is fully deployed. Average claims resolution time has compressed from 30 days to 7.5 days in documented implementations. Cost per standard claim has dropped from $40-60 to $25-36, a 30-40% reduction. Manual document handling, which once consumed 80% of adjuster time, has fallen to 20% in AI-augmented workflows.
On the underwriting side, processing time for submissions has compressed from three days to three minutes in leading implementations, with 20% improvements in risk assessment accuracy and loss ratio improvements of 3-5 percentage points for commercial P&C carriers using ML-enhanced pricing. Quote-to-bind ratios have improved by 60-99% in documented deployments.
| Function | Share of Live AI Use Cases | Key Production Metric | Source |
|---|---|---|---|
| Claims | 58% | STP rates: 10-15% to 70-90% | Everest Group Top 50 2026 |
| Underwriting | 46% | Submission processing: 3 days to 3 min | Everest Group Top 50 2026 |
| Pricing | Trailing | Loss ratio improvement: 3-5 pts | Carrier earnings disclosures |
| Distribution | Trailing | Quote-to-bind: 60-99% improvement | Vendor case studies |
The gap between claims/underwriting and pricing/distribution deployment is meaningful for vendor evaluation. Vendors that claim AI capabilities across all four functions but can only demonstrate production evidence in claims should be evaluated accordingly. Carrier AI deployment failures frequently stem from this gap between vendor capability claims and production reality.
Underwriting Orchestration: The PEAK Matrix Leaders
The Everest Group PEAK Matrix for Underwriting Orchestration in P&C Insurance evaluated 24 leading technology providers enabling orchestration across real-time intake, AI-driven risk evaluation, and integration with rating, pricing, and compliance systems. The global underwriting technology market is estimated at $5 to $10 billion in 2025 with an expected CAGR of 12.5%, making it one of the fastest-growing segments within the broader P&C tech ecosystem.
Three vendors earned Leader designations. Insurity, announced in January 2026 via BusinessWire, was recognized for consolidating workflows, configurable risk scoring, ecosystem integrations, and rapid time-to-value. Insurity serves 22 of the top 25 P&C carriers and 7 of the top 10 MGAs in the US, with over 400 cloud-based deployments. Sylvester Mathis, Chief Insurance Officer and Chief Revenue Officer at Insurity, characterized the recognition as reflecting “the unparalleled power of our platform and the transformative impact it delivers.”
Duck Creek Technologies earned Leader status with an overall enterprise satisfaction score of 7.1, surpassing the market average of 6.94. Aurindum Mukherjee, Practice Director at Everest Group, noted that “Duck Creek delivers tightly integrated underwriting workflows across policy, rating, and data modules, enabling carriers to manage the full lifecycle from submission intake through quote, bind, and servicing within a single ecosystem.” That lifecycle integration is the kind of capability that distinguishes decision intelligence from point-solution automation: the platform does not just speed up one step but orchestrates the entire sequence.
Appian was recognized as a Leader in AI-enabled underwriting, with differentiation in ecosystem fit, rapid deployment, business-led transformation, and transparent AI governance. Appian’s average deployment time of 11 weeks for new underwriting modules leads the market in time-to-value, a metric that matters to carriers calculating the payback period on technology investments. Appian also differentiates through its partnership with InsureMO, combining its Connected Underwriting Workbench with InsureMO’s Excel Rater for evaluate-and-price workflows in a single platform.
For actuaries tracking core system vendor performance, the underwriting orchestration PEAK Matrix provides a structured framework for evaluating whether a vendor’s AI claims translate into measurable underwriting outcomes. The criteria that Everest Group uses, product capabilities, vision and strategy, market impact, and overall value delivered, map reasonably well to the kinds of due diligence questions actuaries should ask when a carrier switches underwriting platforms and the change shows up in loss ratio trends.
The Cloud Ecosystem: AWS Leads, Azure Accelerates
The Everest Group Top 50 data reveals a clear hyperscaler hierarchy for P&C insurance. AWS maintains the broadest ecosystem pull among Top 50 providers. Microsoft Azure is accelerating across InsurTechs, gaining share particularly among newer entrants that build on Azure’s AI services stack. Google Cloud holds a steady presence but has not expanded its share at the same rate.
This tracks with the broader cloud market, where AWS holds 30% of global infrastructure market share, Azure holds 25%, and Google Cloud holds 13%. Together the three hyperscalers control 68% of total enterprise cloud spending. Hyperscaler capital expenditure is exceeding $600 billion in 2026, a 36% increase over 2025, with most of that incremental spending directed toward AI compute infrastructure.
For carriers, 81% now use cloud for claims management, with 25% using cloud exclusively for that function. The industry is pivoting from “cloud first” to what EIS Group terms “cloud smart”: hybrid-by-design strategies that keep sensitive core transactional data on private clouds while using public clouds for AI compute and analytics workloads. Cloud services in insurance are expanding at a CAGR of 14.5% through 2034.
The cloud ecosystem data has direct implications for actuarial teams evaluating vendor concentration risk. If a carrier’s claims AI, underwriting orchestration, and pricing intelligence all run on the same hyperscaler, a service disruption affects the entire decision chain simultaneously. The legacy architecture bottleneck analysis from Insurtech Insights USA 2026 identified this single-cloud dependency as an underappreciated operational risk.
The Agentic AI Gap: Vision Versus Production Reality
The Camunda State of Agentic Orchestration and Automation 2026 report provides a critical counterweight to vendor optimism. While 71% of organizations report using AI agents, only 11% of agentic AI use cases reached production in the past year. The gap is stark: 81% of deployed AI agents function as chatbots and assistants handling summarization and Q&A rather than mission-critical decisioning, and 48% operate in isolated silos rather than integrated end-to-end processes.
For insurance specifically, 65% of organizations acknowledge a significant gap between their agentic AI vision and operational reality. Trust and control barriers dominate: 84% cite business risk when IT lacks appropriate AI controls, 80% are concerned about transparency in AI usage within business processes, and 66% cite compliance concerns. The most telling finding: 82% of organizations have not achieved the process maturity necessary for agentic orchestration.
This gap between adoption rhetoric and production evidence is precisely where actuarial skepticism is most valuable. When a vendor claims “agentic AI for claims decisioning,” the Camunda data suggests the actual deployment is more likely a chatbot that summarizes claim documents than an autonomous agent that adjudicates liability. The 11% production rate means that for every ten vendor demonstrations an actuary sits through, fewer than two reflect systems running in production on live carrier data.
UiPath’s State of Automation in Insurance 2026 report reinforces the point by identifying what it calls “table stakes” for production-grade insurance AI: speed to value, auditability, and model flexibility. UiPath argues that AI assistants and chat interfaces are not the end state, point use cases will not move the needle on cost or cycle time, and building custom agent platforms usually creates brittle, opaque systems. The real opportunity lies in applying agentic AI end-to-end across the workflows that absorb the most cost, risk, and human effort: underwriting workups, claims triage, document orchestration, bordereaux and delegated authority processing, and routine service operations.
UiPath projects that insurers can move from pilots to production in 12 to 18 months without locking into fragile architectures, provided they choose platforms designed for orchestration rather than isolated automation.
Synthetic Data: The Next Deployment Wave
The Everest Group report identifies synthetic data for risk simulation, fraud detection, and new business acquisition as the emerging deployment wave following claims and underwriting AI. The rationale is straightforward: insurance data is simultaneously highly sensitive and insufficiently diverse. Fraud events are rare relative to normal transactions, catastrophic scenarios are infrequent by definition, and regulatory constraints on using production data for model training grow tighter with every GDPR, CPRA, and ISO 27001 update.
MAPFRE provides the most documented carrier case study. Using DataCebo’s Synthetic Data Vault technology, MAPFRE augmented synthetic data to real homeowner fraud detection data and increased recall by 31% while simultaneously increasing precision by 0.85%, reducing false positive investigation time. Property fraud is both more costly and rarer than auto fraud, creating an insufficient training sample that synthetic augmentation directly addresses.
For actuaries, synthetic data introduces a methodological question that existing ASOPs do not fully address: when a pricing or reserving model is trained on a blend of real and synthetic observations, how should the actuary document the synthetic component in rate filings and appointed actuary opinions? The domain-trained AI analysis explored similar questions about model provenance in the context of fine-tuned versus general-purpose LLMs. Synthetic data adds another layer: the training data itself may be partially artificial, and the regulatory framework for disclosing that fact is still developing.
Insurance fraud costs the US industry approximately $80 billion annually. If synthetic data augmentation can improve detection recall by even half of what MAPFRE demonstrated, the reserving and pricing implications are material. Fraud teams are increasingly turning to tokenization, masking, synthetic data enrichment, and federated learning as privacy-preserving AI techniques that allow sensitive signals to be analyzed without exposing raw policyholder information.
Evaluating Vendor Decision Intelligence Claims: An Actuarial Framework
The Everest Group report provides the market map. The Camunda and UiPath reports provide the reality check. For actuaries and pricing teams tasked with evaluating vendor decision intelligence claims, the gap between these two perspectives defines the evaluation framework.
Ask for production metrics, not pilot results. With only 11% of agentic AI use cases reaching production, any vendor claim based on pilot data should be discounted heavily. Request loss ratio impact data from carriers running the system on live production books, not sandbox demonstrations. The 3-5 percentage point loss ratio improvement cited for commercial P&C underwriting AI is a production benchmark; demand equivalent evidence before crediting vendor claims in expense ratio or loss ratio assumptions.
Distinguish decisioning from routing. Many systems marketed as “decision intelligence” actually perform workflow routing: they send documents to the right queue faster but do not change the decisions made at each queue. True decision intelligence changes the decision itself, such as an underwriting system that calculates risk scores from ingested data and recommends pricing adjustments without human intermediate steps. The Everest Group evaluation criteria of “ecosystem fit, ease of deployment, business-led change enablement, transparent/governed AI capabilities” provide a useful lens, but actuaries should add a specific question: does the system change which decisions are made, or only how fast existing decisions execute?
Map vendor claims to the eight-segment taxonomy. A vendor that claims broad decision intelligence capabilities but is recognized in only one or two Everest Group segments likely has depth in those segments and marketing in the others. The eight-segment structure, underwriting orchestration, claims intelligence, SaaS cores, pricing intelligence, intelligent workflow automation, property and location intelligence, digital experience, and distribution management, provides a structured way to test where vendor capability is real and where it is aspirational.
Evaluate cloud concentration risk. As carriers consolidate on AWS, Azure, or Google Cloud, the vendor ecosystem follows. An actuarial team evaluating a vendor switch should ask whether the new vendor runs on the same cloud as the carrier’s existing core system, claims platform, and data warehouse. Single-cloud dependency creates correlated failure risk that does not appear in vendor presentations but does appear in operational risk assessments.
Require synthetic data disclosure. If a vendor’s model was trained using synthetic data augmentation, that fact should be disclosed in the model documentation provided to the carrier’s actuarial team. ASOP No. 56 (Modeling) requires actuaries to understand the data underlying models they rely on. Synthetic data augmentation is not inherently problematic, as the MAPFRE results demonstrate, but it changes the assumptions embedded in the model and those assumptions must be documented.
Why This Matters
The Everest Group Top 50 report crystallizes a market transition that has been building for three years. Process automation reached its ceiling: making existing workflows 20-30% faster yields diminishing returns once the easy friction is removed. Decision intelligence breaks through that ceiling by changing which decisions carriers make, not just how fast they make them.
For actuaries, this transition creates both opportunity and obligation. The opportunity: vendor AI systems are now generating structured decision logs at sufficient scale to become actuarial inputs. Claims triage decisions, underwriting risk classifications, fraud probability scores, and pricing recommendations all create data streams that actuaries can analyze, validate, and incorporate into reserving and pricing models. The carriers and vendors recognized in the Everest Group Top 50 are the ones generating these signals in production rather than in press releases.
The obligation: with 58% of AI use cases in claims and 46% in underwriting, the decisions AI systems make are no longer supplementary to actuarial work. They are direct inputs to it. An actuary building reserve estimates for a carrier that uses AI-driven STP for 70-90% of simple claims is working with fundamentally different data than one building reserves for a carrier with 15% STP rates. Travelers’ $1.5 billion technology budget already demonstrates what happens when AI moves from pilot to infrastructure: expense ratios improve by 300 basis points while technology spending increases.
The Camunda data showing only 11% of agentic AI reaching production is the necessary corrective. The market is moving toward decision intelligence, but it has not arrived. Actuaries who can distinguish between vendors that have production evidence and those that have only pilot demonstrations will make better vendor evaluation recommendations, build more accurate expense ratio assumptions, and avoid the trap of projecting vendor marketing into loss ratio forecasts.
The $16-18 billion P&C tech market will keep growing. The 38 vendors Everest Group recognized across eight segments will keep competing. The question for actuarial teams is not whether to engage with decision intelligence technology but how to evaluate its claims with the same rigor they apply to any other data source that enters their models.