Marrow vs WaniWani | Insurance-Only vs Multi-Vertical AI Distribution
Marrow and WaniWani both put insurance products inside ChatGPT, Claude and Gemini. The difference is scope. Marrow builds for insurance and nothing else, connects to insurer core systems, and stewards an open standard the whole market can use. WaniWani builds an AI storefront product across four verticals, insurance, budgets, HR and travel, on a proprietary SDK. Both are credible. They are built for different jobs.
What is the core difference between Marrow and WaniWani?
Marrow is a vertical specialist. Insurance is the only market it serves, agent-mediated insurance is the only problem it solves, and every part of the platform, the canonical schema, the conduct rules, the legacy connectors, the audit trail, exists because insurance is a regulated product with underwriting, disclosure and suitability duties attached.
WaniWani is a horizontal platform with an insurance lead. Its own site lists four product verticals: insurance, budgets, HR and travel. Insurance is listed first, its customer logos are mostly insurers, and its published research is insurance app testing. So the insurance focus is real. But the product is built to generalise across verticals, and a platform designed to serve HR and travel alongside insurance is making a different architectural bet than one that only ever has to serve insurance.
Neither approach is automatically better. Horizontal platforms reach more customers and amortise engineering across markets. Vertical specialists go deeper into one. Which one an insurer wants depends on how much of the insurance-specific problem they need solved.
Does vertical specialisation actually matter for AI insurance distribution?
It matters where the product hits the regulated parts of insurance.
A general AI distribution layer can handle discovery, Q&A and lead capture across almost any product category, and there's no reason a travel product and an insurance product need different plumbing for that. The moment the conversation touches a real quote, from a real rating engine, with real disclosures attached, bound as a real policy, the requirements stop generalising. Rating factors, underwriting rules, IPIDs, cooling-off periods, advice boundaries, complaints procedures, per-jurisdiction conduct duties. None of that has an analogue in an HR or travel product.
That's the boundary Marrow is built on the far side of.
How deep does each go into insurer systems?
This is the most concrete difference and the one that matters most for large insurers.
WaniWani's product is an AI storefront: it describes real-time personalised pricing and lead capture, delivered without requiring an app or a form, plus monitoring and analytics around how AI platforms present the product. That's a well-defined product with a clear job, getting a product discoverable and quotable in front of AI users, and it's a strong fit for a digital-native insurer whose systems are already modern and API-first.
Marrow is built to reach into the systems a Tier 1 insurer actually runs, including the old ones. It maps existing insurer API fields to a canonical schema in the portal, no re-platforming. It ships legacy connectors for SOAP and XML core systems. It validates every field an AI agent sends against insurer-defined product rules at runtime, catching hallucinated values before they touch the insurer. And it ends in a pre-filled, insurer-branded checkout so the policy is bound on the insurer's own site, under its own brand, with no third-party checkout and no personal data stored by Marrow.
The shorthand: WaniWani gets you a storefront. Marrow gets you from an AI conversation to a bound policy against the core system you already have. See the carriers page for how the integration works.
Open standard or proprietary SDK?
Marrow publishes and stewards the AMI (Agent-Mediated Insurance) Standards: an open specification for how an AI agent requests, compares, discloses and binds a regulated policy across motor, home, SMB and travel, with the shared conduct rules underneath, consent and authority, disclosure, suitability, audit. It defines the interface, not the underwriting. Any insurer, broker or AI platform can read it, build on it, and challenge it, including WaniWani. Marrow maintains the document; it doesn't control who uses it.
WaniWani's SDK is its own. Insurers build against WaniWani's spec, on WaniWani's platform.
For an insurer, the question is what happens in five years. A proprietary integration is a bet on one vendor. An open standard is a bet on the market, with the current best implementation of it as the starting point. Marrow expects to compete on execution, not on lock-in, which is the entire reason for publishing the standard rather than keeping it.
Who built each company?
Compliance middleware is only as good as its authors' understanding of what they're enforcing.
Marrow's founders are two insurance operators, one of them an actuary. Tim Graham (CEO) was Head of Product at Kudo, a D2C motor insurtech, then Head of Product at Ki Insurance, the Lloyd's of London algorithmic underwriting platform built with Google Cloud and UCL. Adam Mesout (co-founder) was lead algorithmic and analytics actuary at Ki alongside Tim, and senior actuary at Hadron Insurance, an A- rated specialty carrier. Between them: pricing, algorithmic underwriting, D2C distribution and specialty carrier actuarial work.
The actuarial half of that matters more than it might look. A quote is a pricing artefact before it's a conversational one. Knowing why a rating factor exists, and what breaks when an AI agent guesses at one, is a different kind of knowledge from knowing how to ship software.
WaniWani's co-founder Raphael Vullierme has said publicly that he spent close to a decade running an insurer before founding the company, so there's genuine operator experience there too. The difference is depth of the bench in one vertical, not presence versus absence of it.
Marrow vs WaniWani: the comparison table
| Marrow | WaniWani | |
|---|---|---|
| Verticals served | Insurance only | Insurance, budgets, HR, travel (per WaniWani's own site, insurance listed first) |
| Standard | Open AMI Standards, published, free to build on, stewarded not owned | Proprietary SDK |
| Depth of integration | Connects to insurer core systems, including legacy SOAP/XML. Quote to bound policy. | AI storefront: personalised pricing and lead capture, no app or form required |
| Founding team | Two insurance operators, one a qualified actuary. Tim Graham: Head of Product at Kudo, then Ki Insurance (Lloyd's algorithmic underwriter). Adam Mesout: lead algorithmic/analytics actuary at Ki, senior actuary at Hadron | Co-founder Raphael Vullierme: close to a decade running an insurer, per his own public comments |
| Regulatory home | UK, designed for FCA regulation, onboarding insurers globally | US-founded, multi-jurisdiction from the outset |
| Status | Live with Aviva; onboarding new Tier 1 and Tier 2 insurers now | Live with Tuio (Spain), first insurer-built app approved on ChatGPT; insurer logos on its site include AXA, Nationwide, Progressive, Oscar Health |
| Business model | Platform fee plus outcome-based pricing; engagement models from a low-commitment agent build to pay-on-bound-policy commission with zero software maintenance cost | Open-source core SDK, paid compliance, analytics, pricing optimisation and anti-scraping layer |
| Validation approach | Runtime validation of every field against insurer-defined product rules; hallucination firewall before data reaches the insurer | Synthetic buyers test each model; automatic regression testing on model updates |
| Checkout and data | Pre-filled, insurer-branded checkout on the insurer's own site. Marrow stores no personal data | Lead capture in-conversation, no app or form required |
| AI platforms | ChatGPT, Claude, Gemini | ChatGPT, Claude, Gemini, WhatsApp |
| Partners | Founders Factory | Deloitte, WTW |
FAQ
Is Marrow only for insurance?
Yes. Insurance is the only vertical Marrow serves and agent-mediated insurance is the only problem it solves. That's a deliberate constraint. Every design decision, the canonical schema, the conduct rules, the legacy connectors, assumes a regulated product with underwriting and disclosure duties attached, because that's the only kind of product Marrow will ever carry.
Is WaniWani an insurance company?
No, and it doesn't claim to be. It's an AI distribution platform whose largest vertical is insurance, alongside budgets, HR and travel. Its insurance traction is real: its customer logos are mostly insurers and its published research tests insurance apps on ChatGPT.
Which is better for a Tier 1 insurer with legacy core systems?
Marrow was built for that case specifically: legacy SOAP and XML connectors, field mapping to a canonical schema with no re-platforming, and binding against the systems already in place. An insurer running modern, API-first infrastructure has more options open to it, WaniWani included.
Can an insurer use both?
Technically yes, for a multinational running separate propositions per market. In practice most insurers pick one distribution partner per market to avoid duplicating compliance and integration work.
Which one is cheaper?
Neither publishes flat pricing. Both use platform-fee-plus-outcome models, and Marrow offers commission-only engagement with zero software maintenance cost. It depends on volume and which capabilities are needed. Book a demo to talk specifics.
This comparison reflects publicly available information as of July 2026, including WaniWani's own published materials. Statements characterising WaniWani's products, funding or business model reflect Marrow's understanding of public sources, are provided for informational purposes only, and are not a substitute for verifying directly with WaniWani. WaniWani is a trademark of its respective owner. Details change; if something here looks out of date, let us know.