Marrow vs boost.ai | AI Insurance Distribution Compared
Marrow and boost.ai get compared more often than they should be, because both involve AI and both involve insurers. But they solve different problems. boost.ai gives an insurer the tooling to build its own virtual agent for its own website, app and contact centre. Marrow gets an insurer's products showing up, quoted and bound compliantly, inside ChatGPT, Claude and Gemini, platforms the insurer doesn't own or control. Most insurers who have one will eventually need the other.
What does boost.ai do?
boost.ai is a Norwegian conversational AI platform, founded in 2016 and backed by Nordic Capital, that lets enterprises, including banks, telecoms and insurers, build and deploy their own virtual agents. Its approach is intent-driven and high-control: the insurer's team defines what the agent can and cannot say, which is a deliberate design choice that appeals to risk-conscious carriers, and the platform now orchestrates that structured approach alongside large language models rather than replacing it outright. boost.ai holds SOC 2 Type II certification and has a strong European data-handling focus. Reported insurance customers include Tryg.
What boost.ai does not do is get an insurer into ChatGPT's, Claude's or Gemini's own app ecosystems. Its virtual agents live on channels the insurer already owns, its website, its app, its contact centre, not inside a third-party AI platform's app store.
What does Marrow do?
Marrow is compliance middleware built specifically for that gap: it connects an insurer's existing pricing, underwriting and policy admin systems to ChatGPT, Claude and Gemini, so products can be quoted and bound inside a conversation happening on someone else's platform. It grounds every quote in real pricing, enforces conduct rules as the conversation happens, blocks hallucinated values before they reach the insurer, keeps a full audit trail, and does not store personal data. No re-platforming required; see the features page for the full breakdown.
Marrow is not an insurer or broker; every product distributed through it is underwritten by a regulated insurer, currently including Aviva, with new Tier 1 and Tier 2 insurers onboarding now. Its founding team came directly out of insurance and actuarial work: Tim Graham (CEO), Head of Product at Ki Insurance after the same role at D2C insurtech Kudo, and co-founder Adam Mesout, lead algorithmic and analytics actuary at Ki alongside Tim, and senior actuary at Hadron Insurance.
Why aren't Marrow and boost.ai really competitors?
Because they operate on different sides of a channel boundary. boost.ai automates conversations that happen on channels the insurer owns and controls end to end. Marrow handles conversations that happen on channels the insurer doesn't own, ChatGPT, Claude, Gemini, where the AI platform, not the insurer, decides how the conversation is structured and what the customer sees first.
This is exactly the gap a lot of insurers don't realise they have. An insurer with a mature boost.ai deployment has solved for "customer lands on our website and wants to self-serve." It has not solved for "customer asks ChatGPT directly, without ever visiting our website." Those are genuinely separate integration problems, with separate compliance requirements, because the insurer has far less control over the second one by default, which is precisely why a dedicated compliance layer like Marrow exists for it.
Marrow vs boost.ai: the comparison table
| Marrow | boost.ai | |
|---|---|---|
| Category | Compliance middleware for AI insurance distribution | In-house virtual agent / conversational AI platform |
| Where it operates | Third-party AI platforms: ChatGPT, Claude, Gemini | The insurer's own website, app and contact centre |
| Who buys it | Insurers and brokers | Enterprises across financial services, insurance, telecoms and other regulated sectors |
| Market | UK-born, onboarding insurers globally | Norway-founded, customers across Europe and North America |
| Status | Live with Aviva; onboarding new Tier 1 and Tier 2 insurers now | 600+ deployed virtual agents reported across industries; insurance customers reported include Tryg |
| Founding team background | Tim Graham (CEO): Head of Product at Kudo, then Head of Product at Ki Insurance. Adam Mesout (co-founder): lead algorithmic/analytics actuary at Ki Insurance, senior actuary at Hadron Insurance | Founded 2016 by Lars Ropeid Selsås, Hadle Ropeid Selsås and Henry Vaage Iversen |
| Regulatory model | Not an insurer or broker; products underwritten by regulated carriers | Software vendor; not an insurer. SOC 2 Type II certified |
| AI platforms supported | ChatGPT, Claude, Gemini | None; deploys on the insurer's own channels |
| Underlying approach | Grounds every quote in live pricing data; hallucination firewall validates every field before it reaches the insurer | Intent-driven, high-control dialogue design, now orchestrated alongside LLMs |
| Data handling | Does not store personal data | Not independently confirmed; strong European data-handling focus reported |
FAQ
Do we need Marrow if we already use boost.ai?
Most likely, yes, if you want to show up inside ChatGPT, Claude or Gemini specifically. boost.ai solves for your own website and contact centre. It doesn't put your products inside a third-party AI platform's app ecosystem, and the compliance requirements for the two are genuinely different because you control one channel end to end and don't control the other.
Could boost.ai eventually do what Marrow does?
Possibly, in principle, if it built and maintained integrations into each AI platform's own app framework, plus the insurance-specific compliance layer for disclosures, suitability and conduct across jurisdictions. That's a different, insurance-specific build to what boost.ai is designed for today, which is a horizontal platform across banking, telecoms and other regulated sectors, not an insurance-only compliance rail.
Is boost.ai's AI the same technology as ChatGPT or Claude?
No, not originally. boost.ai's platform was built around intent-driven, rule-based dialogue design that gives the insurer tight control over what the virtual agent says, and it now layers large language models on top of that structured approach rather than being a pure LLM product from the ground up. Marrow, by contrast, is built specifically to sit in front of ChatGPT, Claude and Gemini as they are.
This comparison reflects publicly available information as of July 2026. Statements characterising boost.ai's products or business model reflect Marrow's own understanding of public sources, are provided for informational purposes only, and are not a substitute for verifying directly with boost.ai. boost.ai is a trademark of its respective owner. Details change; if something here looks out of date, let us know.