Your client’s risk data does not live in one place. It sits in your agency management system, in the MGA’s underwriting platform, in the carrier’s policy admin, in bordereaux that arrive as spreadsheets and claims feeds that arrive as something else again. Every handoff between those systems is a re-keying, a reformatting, a small chance for the record of the risk to drift from the risk itself.
The London market is trying to fix exactly that, and the attempt is worth your attention even if your book never touches Lloyd’s — because the shape of the fix tells you where placement data is heading, and what carriers will soon be able to ask of yours.
A London firm called StrataLens describes its product in one sentence: “One governed record between MGAs, carriers and brokers, built from the data you already have, with each party’s own rules in charge. Nothing you run is replaced.” Your data lives in a dozen systems, their homepage says. We make it one.
How it works, in plain terms
The platform installs as an app inside your Microsoft 365 tenant — no rip-and-replace, live in days rather than months, on their account. A submission flows the way it already flows, “as each sender already sends it.” StrataLens routes it from the sender’s platform and transforms it into a single governed record aligned to MRC v3 — the Market Reform Contract standard, the London market’s shared contract format. Then comes the part that makes the design interesting: each receiver’s own rules are applied to that record. Clearing, rating, pricing and quoting still happen through the receiver’s own services. What surfaces to a human underwriter is only the exceptions — “only what the rules could not decide,” in their words. Binding writes back into the receiver’s own system.
The inputs are the market’s real plumbing: bordereaux, submissions, policy-admin exports and claims feeds, taken by API or as a plain file drop. Their published position today is CSV, with Excel, PDF and email ingestion listed as next. Questions about the data can be asked in plain language inside Microsoft Teams, and — this is the claim to hold onto — every figure traces back to its source field.
Two of their stated positions deserve quoting because they are the opposite of how integration usually gets sold. “Vendor neutral” — the layer is built to work across Guidewire, Duck Creek, Sapiens “and the spreadsheet nobody will retire.” And: “People decide. The machine does the transcription, not the underwriting.”
Why a shared record is not a shared system
The distinction in that design is the whole ballgame, and brokers should learn to see it, because vendors will blur it on purpose.
A shared system asks every party to work the same way — one platform, one workflow, one set of rules, and whoever owns the platform owns the process. A shared record asks only that the parties agree on what the risk’s data is: one governed description of the risk, in a standard format, that each side then works under its own rules. StrataLens’s phrase is “each party’s own rules in charge,” and their flow diagram is honest about the boundary — some steps are “handled by StrataLens,” others are “your people, or the other party.”
That is a federation, not a takeover. The carrier keeps its pricing. The MGA keeps its underwriting rules. The broker keeps its own book and its own systems. What changes is that nobody has to re-transcribe anybody else’s data before they can start — and that the number in front of the underwriter can be traced, field by field, back to where it came from.
The honest limits, from their own page
Read the vendor’s fine print like you read a policy, and this one’s own page supplies the caveats.
The transformation is “one governed record,” but the decision steps remain with the parties — which means the record can be perfect and the price can still be wrong. A governed pipe does not validate what flows through it; a wrong figure, faithfully transcribed and provably traced, is still a wrong figure. Lineage proves where a number came from. It does not prove the number was right at the source.
The reach is also narrower than the headline. CSV is what works today on their own telling; the messier formats most real submissions actually arrive in are roadmap, not product. The platform is London-shaped — MRC v3 is a London market standard — so for a US broker this is a direction-of-travel story, not a buying option. And some of the page’s numbers are the team’s history, not the product’s: “11,000+ market users served by platforms our team has built and run” and “750+ brokers, carriers and MGAs our team has worked with” describe the founders’ past work — twenty years of London market operating experience is genuinely relevant, but it is not a deployment count for this platform. The one market statistic on the page, that 46% of London market firms cite legacy system integration as a key barrier to AI adoption, is attributed to an LMA AI survey; treat it as their citation, not as a figure we have verified independently.
One more, because it matters to how you file this: StrataLens publishes no pricing, and its two public legal documents — a terms of service and a privacy policy, both dated 29 September 2026 — govern the website, not the platform. The terms say so in as many words: platform services run “under a separate written agreement with each client.” The privacy policy is just as plain: for data clients process through the platform, “the client is the controller, we act as its processor, and the terms are set out in the contract with that client.” So who owns the record, contractually, is a question only their sales process can answer.
What is public is their stated policy, on a trust page that dates every claim and separates what is held from what is in progress from what is still planned — a habit more vendors should copy. On that page: data stays in Microsoft Azure’s UK South region; your data is “not used to train any model”; AI only proposes and a person approves; every configuration change, approval and denied access goes into an append-only audit log; and on termination your data is exported if you ask, then deleted within 30 days, with retention periods agreed per customer in the data processing agreement. That posture points the right way — processor, not owner; no training; export-then-delete on exit. But policy is not contract, and the difference is the whole lesson of reading vendors. The trust page also records, in the company’s own words, what is not yet true: the independent penetration test is planned before the first paid pilot, SOC 2 is a target for early 2027, ISO 27001 is planned for 2027. For a product whose entire pitch is governance, the first question a broker asks is whether the contract says what the trust page says — and that question can only be asked inside the process.
What it means at your desk
None of this requires you to buy anything, and most of it is not aimed at your market yet. What it does is put three things on the record.
First, the standard is the story. MRC v3 alignment means the London market is converging on one machine-readable description of a placed risk. When a market agrees on the record, the record starts travelling — into analytics, into AI tools, into questions asked in plain language over the top of it. The brokers who benefit are the ones whose own data is clean enough to become that record without a rescue operation.
Second, exceptions-based underwriting is the shape of what’s coming. If the routine flow is transcribed, routed and rule-checked by a layer, the human underwriter’s day becomes the exception queue — the submissions the rules could not decide. That changes what a good submission is. A risk that arrives complete, consistent and machine-readable clears rules. A risk that arrives as a narrative and a prayer becomes somebody’s exception, priced with somebody’s caution.
Third, lineage is becoming a deliverable. “Every figure traces back to its source” is a promise carriers will increasingly make in both directions — and one they will increasingly demand. When a number on your submission is challenged, “where did this figure come from, field by field” is about to be a normal question. Shops that can answer it will move faster than shops that cannot.
The practice
This week, take the next submission you send to any market — London or domestic — and do the trace test on it yourself. Pick the five figures that matter most to the price: the values, the loss figures, the key dates. For each one, write down where it came from — which system, which document, whose entry. If any of the five ends at “someone sent it in an email” or “it’s what we had last year,” you have found the seam a governed record would close, and the seam an underwriter’s exception queue will find if you don’t. You do not need a London platform to close it. You need to know it is there.