Most Carriers Are Using AI in Claims. Only 7% Have Scaled It.

According to a new Sedgwick report on AI in property claims, 82% of carriers are now using AI tools somewhere in their operations. That’s a remarkable number. And yet only 7% have managed to scale the technology successfully.

The report states that nearly two-thirds of carriers acknowledge a gap between their AI vision and their current reality. The industry’s AI investment is expected to grow from $10 billion in 2025 to nearly $80 billion by 2032. More money flowing into a problem that’s barely been solved yet.

The gap between adoption and scale isn’t a technology shortage. The right tools exist. The budgets exist. What the report makes clear is that the carriers who reach scale approach implementation differently from the ones who stall. Here’s what that looks like in practice.

Why Most Claims AI Initiatives Stall After the Pilot

The Sedgwick report points to three patterns that show up repeatedly in operations that struggle to scale:

1. Infrastructure. Most claims systems weren’t designed for the API connectivity that modern AI requires. When AI gets layered on top of a legacy platform rather than embedded into core workflows, the result is friction: inconsistent data, duplicated effort, and performance that degrades as volume scales.
2. Data. When multiple AI tools and vendors each handle a different part of the claims process, the data they produce is often inconsistent, incomplete, or siloed. AI outputs are only as reliable as the data feeding them, and at scale that inconsistency compounds.
3. Adoption. Adjusters carrying heavy caseloads don’t have bandwidth for tools that feel like extra work. The report also flags a specific pattern worth noting: carriers that hold AI to a standard of perfection from day one, rather than measuring incremental progress against a real-world baseline. That expectation can kill momentum.

Where Claims AI Works Best

The carriers seeing real results have focused AI on the parts of claims handling where speed and consistency matter most: intake, document processing, low-severity claim handling, and administrative coordination.

The Sedgwick numbers are specific. Intake automation has cut average processing time from 10 days to 36 hours. AI-powered photo analysis has improved claim handling efficiency by up to 54%. For low-severity claims, carriers have seen 80% faster processing and 50% productivity gains in documentation. Without AI, claims handlers spend roughly 30% of their time on low-value administrative work.

These gains are real and measurable. They’re also concentrated in a specific category of work: tasks that are repetitive, rule-based, and high-volume. That’s where AI-driven automation earns its keep.

Complex losses, ambiguous coverage questions, and emotionally sensitive claims are a different matter. The report found that human-in-the-loop models, where AI supports rather than replaces human decision-making, quadruple trust in AI outputs.  

What Scaling Claims AI Projects Actually Requires

The operations that reach meaningful scale tend to share a few traits:

1. They start narrow. One workflow, well-defined success criteria, clean enough data to work with. Intake is a common first target because it’s high volume, relatively standardized, and easy to baseline. A measurable win here builds the credibility needed to expand.
2. They build for orchestration rather than collecting point solutions. An AI tool that handles one step in isolation creates new coordination problems at the seams. Scaling requires an AI layer that understands claim state across the full lifecycle: what’s happened, what’s missing, and what should happen next. When a document arrives, that means updating the claim’s state, triggering the next action, and surfacing the file to the right adjuster with context already assembled. Document extraction alone doesn’t get you there.
3. They involve adjusters early. The people who live in the workflow every day know exactly where it breaks. Bringing them in as co-designers before launch is far more effective than managing their resistance after it.
4. They assign a real owner post-launch, a project manager. Pilots have champions. Scaled deployments need operators: someone accountable for adoption, performance, and continuous improvement after the launch energy fades. Without that, even solid implementations drift.

One Thing Worth Knowing About Claims AI Infrastructure

A common assumption is that scaling AI in claims requires replacing the core claims system. It usually doesn’t.

Most of the friction that slows claims down lives in the coordination layer between systems: documents that arrive but don’t surface, tasks that are ready but nobody flags, context that disappears on reassignment. Those are coordination gaps, and they can be addressed by adding an orchestration layer on top of existing infrastructure. 

AI layers like Clive™, Five Sigma’s Multi-Agent AI Claims Solution, are built specifically for this, automating across the full claim lifecycle without requiring a system replacement.

The Path Forward

The Sedgwick report puts it plainly: strategy is the new competitive advantage in claims AI. Not technology, not investment size, not vendor selection. The carriers in the 7% started with a clear-eyed view of where time was being lost, built toward orchestration rather than isolated automation, and measured progress against real baselines rather than aspirational benchmarks.

———————————————————————————————————————————————-

Frequently Asked Questions

Where should a claims operation start with AI implementation? Start with a workflow that has a clear problem, measurable success criteria, and sufficient data quality to produce reliable outputs. Intake is a common entry point because it’s high volume, relatively standardized, and easy to baseline. A focused early win builds both leadership credibility and adjuster trust, which are prerequisites for expanding to more complex use cases.

Does scaling AI in claims require replacing existing systems? Not necessarily. Most of the friction in claims operations lives in the coordination layer between systems rather than in the systems themselves. An orchestration layer on top of existing infrastructure can address coordination gaps, trigger workflow actions, and maintain claim-state awareness without requiring a full platform replacement.

What role do adjusters play as AI scales in claims operations? Complex losses, ambiguous coverage questions, and claims where a policyholder needs a real person on the other end aren’t going away. Those moments require judgment, empathy, and experience. The practical shift is in where adjuster time goes: less administrative coordination and document chasing, more coverage decisions, negotiation, and claimant engagement. That’s a better use of experienced talent, and it tends to improve retention too.