Where Successful Claims Automation Starts

Ask most insurers where their automation opportunity lies, and the conversation goes straight to AI: which model, which vendor, which use case to pilot first. That’s the wrong starting question. The right one is: what problem are we solving, and what’s the most effective way to solve it?

Automation isn’t just technology investment. It’s the elimination of unnecessary effort – and that effort hides at three levels, roughly in order of complexity and speed to value.

Most insurers jump straight to level three. That’s not because it’s wrong – it’s often necessary – but because levels one and two get skipped entirely, which means the AI initiative ends up automating a broken process instead of a clean one.

Level 1: Business process changes. No development required, and usually the fastest win in the building. Are all approval steps actually necessary? Are we asking customers for more documents than we need? Are communications standardised, or does every handler improvise their own version? Are customers being pushed toward digital channels, or defaulting to phone and post? How much of the current process exists only because "we’ve always done it this way"?

Level 2: System usage optimisation. Configuration, not code –is often the most underrated layer. Most insurers significantly underuse the claims platforms they’ve already paid for. Are handlers using the system consistently, or has everyone built their own workarounds? Are the capabilities that were switched on at go-live still switched on? Do the KPIs point to a specific bottleneck – assignment, reserving, routing – that a configuration change would fix? Are manual workarounds quietly replacing standard functionality that already exists in the core system?

Level 3: System changes and AI. New development, new integrations, workflow orchestration, advanced analytics, self-settlement models – this is where the transformative capability lives, but it should be the third question, not the first. It’s most powerful once levels one and two have already removed the effort that didn’t need a platform to fix.

The framing of the business case predicts the outcome. Automation initiatives that start with “we want to implement AI” tend to stall – the technology becomes the deliverable, and success is measured by whether the pilot is shipped, not whether anything got cheaper or faster. Initiatives that start with “we want to reduce claim handling costs by 20%” or “we want to cut average settlement time from 14 days to 7” tend to succeed, because the business outcome is primary and the method – process redesign, better system usage, or new technology – is whatever gets there fastest.

That reframing also changes where AI belongs. Positioned correctly, it’s a highly valuable accelerator, not a universal answer: email triage, FNOL classification, document extraction, medical documentation review, fraud detection support, claim summarisation. All of these work because they’re applied to effort that’s already been identified and understood. Applied before that identification happens, AI risks automating activity that should have been simplified or eliminated in the first place – making a wasteful process faster instead of making it disappear.

Claims transformation at one of the leading UK insurers shows how proper approach to automation looks like at scale. Facing rising claims handing costs, long settlement times and overloaded claims teams, the programme deliberately resisted the instinct to lead with AI. Instead it asked where unnecessary effort was being created across the claims journey – and worked through all three levels: redesigning the total loss journey and reducing handoffs (Level 1), getting substantially more value out of claims core platform through better workflows and queue management (Level 2), and only then adding targeted integrations and analytics (including MID and Verify), as well as AI proof of concept for one of the processes (Level 3), with Sollers supporting the core system development, integrations and test automation behind it.

Programme benefits in cost savings were 10 and 50 times higher (combined first-year and three-year benefits, respectively) than the programme expense. A substantial cut in FNOL handling time and a projected reduction in recovery lifecycle by 40% belong to the outcomes from the programme.

Claim transformation didn’t succeed because it deployed AI across claims. It succeeded because the foundations were laid first, and technology was introduced once the business case for it was strong.

In claims automation, the biggest mistake isn’t choosing the wrong technology. It’s focusing on the technology in the first place.