Bastiaan de Goei: From Intelligence Layer to Human in the Loop

AI can prepare the quotation. It cannot take ownership of the decision. In this episode of Scouting for Growth, I sit down with Bastiaan de Goei, who leads financial services marketing at OpenAI, for a conversation recorded live at ITC Vegas about what happens when AI moves beyond an impressive demonstration and into the daily work of an insurer. Having worked across AXA and Instabase, Bas understands both the promise of frontier technology and the reality of bringing it into established insurance operations. The uncomfortable truth is that automating individual tasks is not the same as designing a useful workflow. Bas points to practical progress already being made: at Travelers, he describes an OpenAI-powered tool handling 80 to 90 percent of claims intake data, while Oscar Health has accelerated claims verification by 50 percent. Those gains matter, but they are only the beginning of what becomes possible when intelligence can connect the steps between systems. We explore the intelligence layer through the working day of a commercial underwriter. Instead of manually sorting every broker submission, extracting data, checking pricing and preparing quotations, an agent could coordinate those steps and present a set of cases ready for review. The underwriter becomes the orchestrator rather than the operator, with more time for uncertainty, exceptions and decisions that genuinely require expertise. But greater automation does not transfer responsibility to the machine. Bas is clear that the final quotation remains the underwriter's decision. The technology must make its evidence visible: the steps taken, the assumptions made and the source data used. Without that transparency, human oversight risks becoming little more than a final click. Most importantly, Bas argues for what he calls 'efficiency by a thousand cuts': start with a real point of friction, prove the value and build from there. An inbox triage improvement that surfaces worthwhile submissions previously left untouched may be more valuable than an 18-month transformation programme. We also examine the economics of capacity, the need for continuing cyber safeguards and the importance of giving experienced people room to experiment. Because the objective is not to remove the expert from the workflow. It is to expand what skilled people can do, while making the decisions they remain responsible for more informed, more efficient and more trustworthy. KEY TAKEAWAYS I want to challenge the way we often think about AI adoption as a collection of isolated use cases. Improving claims intake or accelerating verification is valuable, but the bigger opportunity comes when those capabilities can work together. The intelligence layer Bas describes connects submissions, documents, pricing systems and quotation templates into a process the expert can oversee. For underwriters already overwhelmed by incoming business, the shift from operating every step to orchestrating the workflow can create capacity not just to work faster, but to consider risks that would otherwise go unreviewed. This is why I believe meaningful human oversight has to be designed into the workflow, not added at the end. Bas makes the accountability point explicit: the ultimate quotation belongs to the underwriter. That person needs to know what the agent did, which assumptions it made and what evidence informed its recommendation. An agent can prepare the work and identify uncertainty, but the human must be able to challenge it and own the decision. Trust in this environment depends on making the reasoning and source material accessible enough for judgement to remain real. For me, the practical lesson is to begin with friction and measure the business outcome rather than becoming fixated on tokens or technology costs alone. Bas's 'efficiency by a thousand cuts' could begin with something as modest as inbox triage, then compound through small improvements that save time, recover capacity and create new opportunities. Leaders should give their people room to test these tools, while designing appropriate safeguards as the technology evolves. The advantage will not come from an ambitious transformation announcement. It will come from improvements that earn trust, become part of the work and make skilled professionals more effective. BEST MOMENTS “What I found is that it is much better to basically achieve efficiency by a thousand cuts.” – Bastiaan de Goei [07:28] “You become essentially an orchestrator of the underwriting process, rather than the operator of one.” – Bastiaan de Goei [15:42] “The ultimate quotation is the underwriter's responsibility and that person is accountable for that decision.” – Bastiaan de Goei [19:52] “Now, if I'm able to identify two or three submissions that would otherwi