The principles below come from more than forty years inside the industry across our leadership team — broker side, insurer side and MGA side, underwriting risks, running delegated authorities, and working with a good many of the systems already in this market. They are what you learn from watching platforms hold up, and from watching them quietly fill with spreadsheets and workarounds. This is how we build, and the scope is the part most people underestimate.
Most "underwriting platforms" are a quote engine with the rest left as an exercise. The gap between quoting a risk and running a book is where operations quietly fill up with spreadsheets and email.
A broker-facing portal for quoting, binding, documents and mid-term changes. A separate underwriting portal for referrals, overrides, portfolio views and administration. Brokers never see internal underwriting reasoning; underwriters never work blind.
Several distinct classes of commercial business, each with its own question set, rating basis, appetite and wordings — all running on one engine rather than as separate systems bolted together.
Quote, refer, decline, bind, mid-term adjustment, cancellation and renewal. Each one re-rated and re-documented properly, which is precisely what tends to be missing when a platform demonstrates well but struggles in its second year.
Schedules, statements of fact, certificates, invoices and correspondence generated from the rated outcome. A cover, limit or excess appears identically on screen, in the document and in the rating — because all three read the same source.
Different sections of a single policy can sit with different carriers, with the split handled in pricing, in the documents and in the reporting that goes back to each of them.
Bordereaux and management reporting produced from the same records that priced and bound the business, in each carrier's own required shape, rather than rebuilt monthly in a spreadsheet.
An existing portfolio brought across without rekeying it, reconciled policy by policy, and behaving like natively written business from the day it lands. This is the part that stops most moves happening, so it has a section of its own below.
This is the reason most brokers and MGAs stay where they are. The platform is rarely the problem — the book is. Nobody wants to rekey several thousand live policies, mid-term, while still writing new business, and a migration that goes wrong is visible to every client you have. So the decision gets deferred, year after year.
It is a solvable problem, and it is the part of a move we take most seriously.
Existing policies are read by machine from the documents you already hold, and turned into structured risk records. What used to be months of temporary staff and rekeying becomes a supervised exercise your own team reviews rather than performs.
Every migrated policy is checked back against its own source document — premium, cover, limits, dates, insured. Anything that does not reconcile is flagged for a person rather than quietly accepted. You finish with a report you can put in front of your carrier, not an assurance that it probably worked.
A migrated policy keeps the cover and premium it was sold on. The platform does not re-rate or re-decide historic business on the way in — current rules apply at its next renewal or mid-term change, where the client expects a conversation anyway. This is the single most important reassurance we can give: moving platform does not change your book.
Migration happens in parallel with the build, in batches. You can be writing new business on the platform while the back book is still landing, so nothing stops and there is no weekend cutover with everybody holding their breath.
Once across, a migrated policy adjusts, renews, produces documents and reports to the carrier exactly like one written on the platform. No second-class records, no parallel system to keep alive, no "we still have to look that one up in the old system".
If moving your book feels like the risky part, that is the part to test first — on a sample of your real policies, before you commit to anything.
A live book is never a finished product. Rates move, appetite shifts, wordings are revised, and a new product lands beside the existing ones. The risk is never the change you are making — it is the product you are not thinking about while you make it.
So automated checks are built as first-class parts of the system, not an afterthought. Not generic tests either: checks that encode specific underwriting rulings — that a particular excess is right, that a cover charged for is a cover shown, that a document says what the rating decided.
Above those sits a set of locked reference outcomes — real quotes run end to end, with premiums and terms compared against known-good results. If a change moves a number anywhere in the book, it fails before release rather than surfacing in a broker's schedule.
Every defect found in testing becomes a permanent check of its own, so the suite grows with the book and a problem fixed once does not return a year later when somebody touches an adjacent rule.
Confidence to change quickly comes from being able to prove nothing else moved — not from changing carefully.
None of the above is specific to one class of business. The lifecycle, the document production, the reporting and the testing discipline are common to underwriting itself. Your class, appetite, rates, wordings and question sets are what make it yours — and they stay yours.
A system that holds your appetite, applies it consistently, evidences every decision, and produces the reporting your capacity provider or your board expects — without a monthly reconciliation exercise. New classes and new schemes follow in weeks rather than becoming their own projects.
The ability to quote and bind your own schemes properly, with the same audit trail and document quality as a carrier's own system, instead of a spreadsheet rater and a wording library nobody is quite sure is current.
We have described capability and approach, not mechanism. There is nothing here about rating structures, appetite thresholds, scoring behaviour, data sources, carriers or clients, and there will not be.
That restraint is the point. Anything we are told in confidence stays that way — and a supplier willing to describe someone else’s rating logic to win your business would describe yours to win the next.
Under an NDA, and with a genuine opportunity in front of us, we can go considerably deeper.
A conversation about your own book is more useful than any description — including a migration trial on a sample of your real policies. Tell us what you write and where the process slows you down. Timescales are on the delivery page; the approach to machine assessment is on AI underwriting.
Send us an enquiry →Or email hello@apdrit.com