Agentic and Generative AI: Execution Is the Advantage
The DOCOsoft team had an opportunity to take the pulse of AI in the insurance market at the recent Agentic + Generative AI event organised by Intelligent Insurer. Our CEO Aidan O’Neill, Lead Data Scientist Bernard Cosgrave, and Product Manager Lloyd Bizaoui were among a packed crowd of insurers, reinsurers and technology leaders, all assembled to discuss the challenge of turning AI from pilot projects into sustained drivers of business advantage.
There were plenty of interesting products being demonstrated and many engaging presentations and conversations on the day. But one clear and inescapable conclusion emerged from all of this: the carriers who are achieving real progress are those who are most comfortable iterating openly.
Learning by narrowing the field
AXA Group Chief Data, AI & Innovation Officer Andreas Schertzinger spoke candidly about the scale of experimentation underway. Dozens of agentic initiatives are in development across the group, but only a small number are live in production. He framed this explicitly, as healthy discipline, not hesitation.
Speakers from Swiss Re described starting out with around 60 use cases before sharply reducing the list, finding that value began to crystallise once focus increased. The closing keynote from Matthew Richardson, Global Head of Operations and IT at Generali, summed up the direction of travel: the next decade, he said, will belong to those who execute.
Across sessions, the message was consistent and unambiguous: AI programmes rarely land perfectly first time. They are evolving. The most effective approach is to build something viable, then measure, adjust, and redeploy. The advantage sits with organisations who are best prepared to move through that cycle quickly and repeatedly.
Speed as a differentiator
Sofia Kyriakopoulou, Group Chief Data & Analytics Officer at SCOR, spoke about ‘AI at speed’ and the shift from proof of concept to enterprise value. In underwriting discussions, the point was reinforced: responsiveness shapes outcomes. A rapid, clear response influences broker behaviour and the ratio of business won.
The same logic applies for claims. When handlers have rapid access to structured information, case prioritisation, and contextual insight, claims teams can move forward with confidence. This then manifests itself in improved service and improved retention rates.
These themes align closely with our own focus at DOCOsoft and go to the heart of what execution looks like in practice. Our CMS is designed around workflow clarity and operational momentum. IMR integration reduces document friction, while AI-driven prioritisation and triage surface high-risk or time-sensitive claims early, shifting processing away from its traditional reactive stance.
When the market talks about speed as a competitive moat, it’s essentially talking about practical system design.
Fixing the mailroom first
One of the most grounded and pragmatic conversations of the day came during the underwriting panels. Reno Daigle, CIO and Head of Business Transformation at CNA Hardy, emphasised the advantages of starting with high-value, less regulation-sensitive use cases, and getting the data foundations right there.
Several speakers referenced the unhelpful operational reality that unstructured submissions continue to dominate. Email remains the main channel of intake, with underwriters and claims handlers still having to spend a significant portion of their day navigating attachments before having any opportunity to exercise their judgement.
From an operational standpoint, one consistent theme across the day was the importance of cleaning data at the point of entry, structuring it early, and routing it intelligently. This is where generative and agentic AI are already delivering impact: automated extraction from submissions, real-time classification and intelligent triage.
It’s also where DOCOsoft’s ability to operate in depth pays off. Our email claims module and IMR integration address ingestion directly. AI Email Ingestion, Document Classification, Claims Triage and Prioritisation all act on unstructured inputs the moment they land. Structuring information early enables everything downstream to function more smoothly.
The event reinforced the practical reality that sophisticated analytics depends on disciplined inputs.
From tools to orchestration
Another notable shift was in how agentic AI is being described. Andy Logani, EVP and Chief Digital and AI Officer at EXL, spoke about the journey from ‘software plus human’ toward more autonomous inferencing, combined with human oversight.
The emphasis is moving from isolated extraction tools to coordinated workflows. Agentic systems are being designed to read documents, trigger rules, escalate exceptions and update records as part of a chain.
For specialty insurance, where risks are layered, and documentation is extensive and dense, that orchestration matters. Intelligence needs to operate within the platform, interacting with rules, audit trails and human oversight in a structured way.
Governance rising up the agenda
The afternoon sessions devoted some serious airtime to issues around governance. Several speakers discussed implementing formal AI risk assessments early in development cycles, having clear accountability frameworks, and documenting mitigation steps prior to production deployment.
Data sovereignty, encryption and regulatory alignment were all discussed in commercial terms. Transparent AI governance is becoming part of procurement conversations. For technology providers, this means making sure governance is visible and embedded, not simply implied.
Building momentum
The overall impression was of a market moving forward with intent. AI spend in insurance is rising, with the carriers both building internally, where they have the scale and resource to do so, as well as partnering externally. The tone has shifted from curiosity to delivery.
As stated at the outset, the organisations gaining ground now are those who are translating AI capability into operational rhythm: making decisions, embedding intelligence where it directly affects workflow, and improving their AI capabilities with successive releases.
That is where the practical edge now lies. Not in making grand claims about having cracked AI, but in demonstrating a proven ability to deploy it purposefully, review its impact intelligently, and strengthen it over time.
From our perspective, the direction of travel is clear. The work is underway, and the opportunity sits with those prepared to press ahead with purpose and ambition.