(Re)thinking insurance

The episode has a handful of genuinely useful practitioner insights - particularly the distinction between generative AI as a suggestion layer and agentic AI as an executor, the 'governance agents policing other agents' idea, and the concern about future expert training pipelines - but these are diluted by a long host intro, repetitive regulatory caveats, and generic 'change management is the real barrier' statements.

“it's agentic AI which is the force multiplier. So that is when all these useful tools that suggest things actually link up interconnect and still with an appropriate human in the loop...will actually act and actually do things um, around the whole control cycle”

“I can see a world in which you actually train up a number of um, agents that are actually governance agents, policeman agents, so sort of uh, design them to be particularly fastidious things looking for issues to question. And you know, you pit one set of agents against another almost to, to check that what's been done is okay”

The episode recycles several standard AI-in-enterprise talking points ('technology is ready, change management is the barrier', 'augments not replaces') but earns some credit for the insurance-specific argument that LLMs must not sit at the deterministic algorithmic core and for the underexplored expert-pipeline degradation concern; the borrowed 'at the table or on the menu' quote exemplifies the recycled-take problem.

“at the heart of a lot of what we do in pricing, reserving, capital modelling there's a very specific hardcore calculation um regulators are very unhappy if that changes for no apparent reason. So I think at the heart of a lot of what we do we will need a deterministic um algorithmic core”

“with AI you're either at the table or you're on the menu”

Duncan Anderson is a genuine 35-year insurance practitioner and actuary who now runs a large specialist software business, giving him credible cross-functional perspective; however, as a vendor executive he frames much of the conversation around WTW's own tooling, and the episode lacks an insurer-side operator who has actually deployed agentic systems in production.

“My current role, and, um, actually has been for seven or eight years, is to lead our insurance technology business. But I'm not actually a technology professional. I'm not a software engineer. I'm an actuary by profession”

“we have some patented algorithms that we have, the actual execution of those in terms of, of coding them up, um, some of the uh, coding AI tools are really quite amazing now”

The episode gestures at concrete specifics - daily/weekly data ingestion cadences, a '50 people to 5 people plus 50 agents' vision, and a mandated 45-minutes-a-day AI training programme - but the companies, tools, and outcomes remain unnamed throughout, and the host's '~two thirds of insurers using generative AI' statistic is asserted without a source.

“for the last year they mandated 45 minute lunch and learn sessions on AI every day for all staff”

“one of our software offerings we ah, introduced a year or two ago, um, is a portfolio management monitoring tool. It automatically ingests emerging data on a sort of daily, weekly basis”

The host structures the conversation logically and asks a few targeted follow-ups (notably pushing on what 'accountability changing in nature' actually means and whether off-the-shelf AI is adequate), but questions are frequently leading or multi-part, no claim is ever challenged, and the extended host monologue at the close adds nothing new and reads like a prepared summary rather than earned dialogue.

“So expertise gets enhanced. But I think the accountability piece is really important. So what do you mean by I guess how will accountability change?”

“Off the shelf AI though I guess my question is that sufficient or appropriate or do we need something I guess more domain specific and explainable?”