Everything you need to know about Marrow and agentic AI insurance distribution

Marrow is a compliant AI middleware layer for UK insurance distribution. It connects AI agents including ChatGPT, Claude and Gemini to insurer pricing, underwriting and policy admin systems, translating unstructured conversational requests into validated insurance transactions in real time without storing personal data. Marrow is operated by Project Malcolm Limited.

No. Marrow is not an insurer or a broker. Marrow is a technology infrastructure provider. All products quoted and bound through Marrow are underwritten by FCA-authorised insurers. Marrow provides the middleware that makes compliant insurance distribution possible through AI channels such as ChatGPT.

Marrow is designed to work with any major AI channel that supports tool use, function calling or MCP-style API access, including ChatGPT, Claude and Gemini. Which channels we activate for a given customer depends on their distribution priorities. New platforms are added as they cross consumer-adoption thresholds.

Marrow enforces mandatory disclosures, advice boundaries and per-jurisdiction regulatory rules in real time on every AI interaction. Compliance checks run before any quote is surfaced to a customer.

Marrow is designed for FCA-regulated markets but does not itself hold FCA authorisation. Marrow enforces mandatory disclosure requirements, keeps AI agents within permitted advice boundaries and generates a full audit trail for every customer interaction. Insurers using Marrow remain responsible for their own regulatory obligations under the FCA Handbook and Consumer Duty.

Marrow strips personal data in-flight and does not store PII at any point in the quote or bind process. The system is built to ISO 27001 standards and aligned with the EU AI Act. Data minimisation is enforced at the platform layer, so insurer data never crosses into the LLM context window in identifiable form.

Marrow grounds every AI response in insurer-approved content via a FAQ and knowledge base tool. LLMs pull answers directly from the insurer's own approved knowledge base rather than generating responses from training data. Every quote is validated against live insurer product rules before being returned. Hallucination and refusal rates are monitored in real time through the Marrow analytics dashboard.

Yes. With Marrow, insurers can offer real, validated quotes inside ChatGPT and route customers to a pre-filled, insurer-branded checkout link to complete the bind. Marrow powered the UK's first such deployment by a major insurer.

Marrow supports AI chat through ChatGPT, website chat widgets and embedded insurance at checkout, with WhatsApp support rolling out next. All channels share a single compliance layer, audit trail and analytics dashboard, so insurers can monitor every AI-mediated interaction in one place.

Marrow connects directly to existing policy admin, pricing and underwriting systems via legacy connectors and canonical schema mapping. No re-platforming is required. Marrow normalises fields and translates modern unstructured payloads into formats existing insurer endpoints already understand, including SOAP and XML for legacy core systems.

Most Marrow integrations are live within weeks. Marrow uses a canonical schema mapping that connects to existing insurer API fields, which avoids the need to rebuild core systems. The actual timeline depends on the complexity of existing endpoints and the number of products in scope. A single-product MVP can be live in three to four weeks; multi-product rollouts take longer.

Yes. Marrow exposes insurer products to AI agents via the Model Context Protocol (MCP), the emerging standard for connecting LLMs to external tools and data. Each insurer integration becomes an MCP-compatible endpoint that any supported AI agent can call, with compliance, validation and audit handled by Marrow at the platform layer.

A GEO (Generative Engine Optimisation) score measures how an insurance brand performs when consumers ask AI platforms for insurance recommendations. Marrow's quarterly UK Insurer AI Leaderboard blends four dimensions into a single 0 to 100 index: visibility (whether the brand appeared), sentiment (whether surrounding context was positive), accuracy (whether AI claims about the brand were factually correct) and recommendation rate (whether the brand was explicitly recommended).

AEO (Answer Engine Optimisation) is the practice of structuring content and data so that AI answer engines return your brand, products or pages when consumers ask direct questions. For insurers, AEO determines whether your products appear when someone asks ChatGPT "what is the best home insurance in the UK". It is distinct from traditional SEO, which optimises for ranked blue links rather than direct AI-generated answers.

SEO optimises for ranked blue links in Google and Bing. AEO (Answer Engine Optimisation) optimises content to appear in direct AI answers from ChatGPT, Claude, Gemini and Perplexity. GEO (Generative Engine Optimisation) optimises brand visibility and recommendation quality across generative AI platforms. AIO (AI Optimisation) is the broader infrastructure discipline covering how systems, data and APIs are structured to let AI agents transact on a brand's behalf. Marrow operates at the AIO layer.

Insurers can improve their GEO score across four areas. For visibility, ensure products and brand are referenced in structured, crawlable content that AI platforms can retrieve, including an llms.txt file, FAQPage schema on key pages and a clean subdirectory architecture. For accuracy, audit all public-facing product information and correct any factual errors that AI models may have indexed. For sentiment, build a consistent presence in third-party editorial and comparison content. For recommendation rate, the highest-leverage step is enabling compliant AI distribution directly: insurers that can be quoted and bound through AI conversations are significantly more likely to be recommended. Marrow's AI Distribution Playbook covers each of these in detail.

Most UK insurers are invisible in ChatGPT answers because their websites are not optimised for AI retrieval. Common gaps include missing FAQPage schema, no llms.txt file, no Bing indexing (ChatGPT retrieves heavily from Bing), thin product pages, marketing-heavy copy that does not answer direct questions, and no transactional capability through AI agents. Without these foundations, AI models default to citing comparison sites, news outlets or competitors that have them.

Marrow's pricing has both a recurring platform component and outcome-based elements. We tailor the mix to each deployment. Get in touch to talk specifics.

Marrow works with UK insurers and brokers looking to distribute through AI channels, B2B platforms embedding insurance into their products, and AI agent developers who need access to validated insurance capacity with built-in regulatory guardrails.

Marrow is focused on the UK market, but available wherever our customers are.

The Marrow UK Insurer AI Leaderboard is a quarterly ranking of major UK insurers scored on AI visibility, sentiment, accuracy and recommendation rate across ChatGPT, Claude and Gemini. It is published free of charge. Rankings are not sponsored and insurers cannot pay to be listed, removed or re-ranked. The methodology is published alongside each release.

Insurers can get started with Marrow by booking a 20-minute call at onmarrow.com/contact or by requesting the AI Distribution Playbook, a practical framework covering AEO, GEO and AIO for insurers at any stage of AI distribution readiness.