Running on rails, not just vibes: orchestrating AI in underwriting and Send’s next chapter with Duck Creek

Andy Moss is back on the InsTech Podcast – this time as General Manager, Underwriting at Duck Creek, following the recent acquisition of Send. Talking to host Matthew Grant, he explains why the deal closed in just 120 days, why "vibe coded" AI never survives contact with production, and the one hiring principle he'd pass on to any founder selling a business.

From workbench to orchestration engine

Andy was one of the first people in the market to talk about the "underwriting workbench" – pulling the four or five applications underwriters juggle into one place. Today, he says, the whole industry is converging on a bigger idea: the orchestration engine.

"Most successful AI projects in enterprise are based on AI running on rails, not just some app that someone's vibe coded and managed to stand up in a production environment. You need the infrastructure; you need the governance around it."

That shift matters because proofs of concept are easy. Getting an AI agent to survive in a live underwriting workflow, with real governance and controls, is the hard part – and it's where most enterprise AI projects stall.

An industry-first tagline, unpacked

The deal was announced with a deliberately punchy line: the combined business is "the industry's only agentic underwriting to core platform." Andy explains what's behind the words – pairing Duck Creek's two decades of core policy, billing and claims technology with the best part of a decade Send has spent purely on underwriting. Two hard problems, solved separately by two companies, now under one roof.

For Send's customers, that means a closer connection between underwriting, policy administration and billing, lowering implementation costs as products defined in Duck Creek's core sync straight into Send. For Duck Creek, it means Send's deep London Market expertise, opening the door to a market where Duck Creek has had limited presence to date but wants to significantly invest in.

The knowledge walking out the door

The conversation's most striking theme isn't AI at all – it's people. A significant share of the underwriting workforce is set to retire within 15 years, taking decades of judgement and process knowledge with them, just as the job itself gets more complex.

Andy's viewpoint: codify that knowledge onto a platform rather than leaving it in people's heads and inboxes. It's why, he says, customers increasingly want their whole portfolio running through Send – not because every team works identically, but because every team shares the same cross-cutting jobs: a sanctions check, a finance report, a handoff that has to happen the same way every time.

Buy the commodity, Build the differentiators

Asked how insurers should think about build versus buy now that AI has made software cheaper to write, Andy's principle is simple and – he admits – self-serving for a software vendor to say out loud:

"Buy the commodity and build the differentiators."

The value, he argues, is in the internal builds that do something genuinely unique with data or process – not in insurers quietly reinventing core systems that already exist. Nobody's building their own Office 365.

The founder's lesson

Asked what advice he'd give someone else selling a business, Andy doesn't point to a deal tactic. He points to who he hired: co-founders and colleagues who cared as much about doing right by customers as they did about building something ambitious. Functional skills are testable, he says. That kind of alignment isn't – and it's the harder thing to get right.

To hear more about Send’s next chapter following the acquisition by Duck Creek and the future of AI in underwriting listen to the full episode of the InsTech Podcast here.

Send and InsTech also recently published a report here on orchestration in underwriting, referenced in the episode.