Where the industry is now: stuck in efficiency mode #
Recent Celent CIO survey data reveals that 61% of North American life insurers are already investing in generative AI (GenAI), another 31% plan to within two years — but almost all of that spend is concentrated on efficiency use cases (submission triage, document processing, summarization, internal co-pilots). These are pure "optimization plays," not transformation — they improve throughput but don't change how insurers actually perceive or act on risk.
Why growth is different, and harder #
AI implementations that create revenue growth, on the other hand, are much harder to pull off. They require a fundamentally different capability set. They have to be able to analyze all types of data, including unstructured data like calls and chats. Based on that data, the need to be able to make good predictions. And finally, they have to be able to actually take action based on their predictions. There’s a lot that goes into that type of process—not just coding or prompting complexity but also real organization work around governance, transparency and auditability of the models, drift and many other thorny problems that need to be solved.
Why most insurers stop at efficiency #
Efficiency use cases are the low-hanging fruit for a reason: they're obvious, they're fast to build, and the return shows up almost immediately. Growth use cases require deploying a capability, testing it, and waiting to see the revenue impact play out, which takes real time and executive patience most organizations don't budget for. The result is an industry where everyone is getting a little more efficient, but only a few companies are actually winning more business because of it.
What growth actually requires: assess, predict, execute #
Moving from efficiency to growth means combining three capabilities instead of one.
- Assess: pull together the context you already have, including unstructured data like phone conversations and chat transcripts, to spot where an opportunity might exist.
- Predict: run that context through predictive modelling, the traditional advanced analytics kind, not generative AI, to identify which customer might be eligible for a specific rider or product.
- Execute: use agentic AI to act on the opportunity, drafting an email in the client's own tone or notifying the agent directly.
In practice, this looks like cross-sell and upsell opportunities flowing directly into the agent's portal instead of sitting unused in a dashboard, or smart alerts and triggers surfacing a retention risk before a client actually lapses. None of it works without solving the data problem first. If you aren't doing the underlying data governance, you don't have the raw material to assess anything in the first place.
From static decisions to dynamic interventions #
The same shift shows up in underwriting. Instead of a single point-in-time decision, underwriting starts to behave like a continuous process, driven by real-time signals and event-driven architecture. Risk changes can trigger contextual cross-selling automatically. Retention shifts from a reactive save to proactive risk prevention. Both are growth levers, not just efficiency ones, and both require real-time context, flexible decisioning and the ability to act continuously. Without that foundation, insurers stay stuck in efficiency mode, which is where most of the industry still sits today.
Improving retention is worth the investment on its own: keeping an existing client is consistently cheaper than acquiring a new one, and customer experience is increasingly seen as a direct driver of growth, not just a satisfaction metric.
How to make the shift from efficiency to growth #
A few things separate carriers that reach growth from carriers stuck in efficiency mode:
- Build an architecture that can respond to events in real time, not just process batches on a schedule.
- Invest in customer segmentation and predictive modelling so opportunities are identified before a customer or agent asks.
- Treat data governance as a prerequisite, not a side project, since unstructured data is where most growth opportunities actually live.
- Connect insights to action, so predictions reach an agent's portal or a client's inbox instead of sitting in a dashboard no one opens.
- Set expectations early that this is a capability to build continuously, not a project with a fixed end date.
Conclusion #
Cost-cutting is table stakes. Every carrier can point to an efficiency win somewhere in their operation by now. Revenue generation through data and AI is what actually separates the carriers pulling ahead from everyone else standing still. The good news is that most of the pieces, event-driven architecture, predictive modelling, and agentic AI, already exist. The work is connecting them.
For the full breakdown of what that shift looks like in practice, watch the complete conversation between Fabio Sarrico of Celent and Ghassan Karam of Equisoft.
