Become the insurer you set out to be
We strategize, build and operate custom insurance software. We meet you where you are, and build until the numbers move.
What your systems do to you
Eight symptoms we find in almost every insurance estate. Each one is caused by how the systems are arranged, not by how the business is run.
Root cause 01
The rules live in the wrong place
Business logic is compiled into systems, so the people who own it cannot change it.
- A new program needs a release →
- Rating changes go through engineering →
- Twenty years of customisation nobody will touch
Root cause 02
People are the integration
Wherever two systems meet, a person sits between them, retyping.
- Submissions arrive as documents and leave as typing →
- Bordereaux are assembled by hand →
- Every distribution partner wants something different
Root cause 03
There is no single record
Each system keeps its own copy of the truth, and the copies disagree.
- Six systems, three answers on premium
- Nobody can answer who changed a bound policy
All eight, one root: your systems cannot talk to each other.
- Received
- 06:12, broker email + 3 attachments
- Risk read
- 142 locations, 4 years loss history
- Appetite
- In appetite for this binder
- Indicative rate
- Produced, 3 options
- Needs you
- Flood zone A on 2 locations, outside authorityneeds a person
One decision, not an afternoon of assembly. Accept, decline or refer, and the reason is recorded against the submission.
Illustrative surface · product name and figures are examples, not a case study
On the screens people use all day · an underwriter
The submission is a decision by the time you open it. Ingested, validated against the binder and rated before anyone opens it. What reaches the underwriter is one judgment call.
Moves the numberCost per policy bound
All eight have one cause, so they have one answer
Every one is what happens when systems cannot talk to each other, so the business rules, the integration logic and the reporting end up duplicated inside each of them.
The answer is one connecting layer, custom-built for your organization: the layer that makes your systems talk to each other, shaped around your business constraints — delegated authority, approval limits, audit, residency. It does not replace what you run and it does not lock you into ours: everything is reachable through the same door on the same terms: what we build, what you license, and what you have run for twenty years.
Every capability is an API and an MCP tool before it is a screen. So a system we have never seen can be integrated by your own team, and what you already run stays a first-class part of the estate.
For a carrier
Modernise around a core you cannot replace. The policy system stays where it is, and everything you build around it is integrated and governed the same way.
For an MGA
Launch a program without building the plumbing again. The ontology and the gateway are already standing, so the second program costs materially less than the first.
For a broker
One book that reconciles across a panel of carriers who each want something different, with placement and commissions running off the same record.
It should all fit together
Today it does not: the portals, the platforms, the spreadsheets and the twenty-year-old core each speak their own language, and your people translate between them. This is your operation when they finally speak one.
Custom portals and AI agents
built for how you work — or bring your own
One connecting layer
custom-built for your organization
Built around your business constraints
Headless building blocks
API-first · on-prem or your cloud
or something big you already run
InsureEdge blocks · yours as source
your data layer · custom-built
The blueprint in full →
On the screens people use all day · a program manager
The month-end file ties before it is sent. Derived from the policy record and reconciled against the ledger as it is produced. Exceptions come to you, not from the carrier.
Moves the numberBordereaux accuracy and timeliness
- Policies
- 1,284 in period
- Written vs billed
- Reconciled
- Ceded premium
- Reconciled
- Exceptions
- 3 raised: 2 mid-term endorsements, 1 currencyneeds a person
- Carrier format
- Applied, ready to send
The errors a carrier finds are the ones that cost you credibility at renewal. These get found here instead.
Illustrative surface · product name and figures are examples, not a case study
Your organization, before and after
Same systems on both sides. The difference is one layer, and where the people sit.
Where you are
Point-to-point connections, and people retyping in between.
Policy core, 2004, Guidewire, Rating spreadsheets, Broker emails, Reporting, by hand — each connected point to point.
Where we take you
The same systems, connected once. Software assembles; your people decide.
One connecting layer — built around your constraints
- Submission 4471
- Quoted, awaiting your client
- Submission 4489
- With underwriting, flood referral
- Outstanding
- Signed proposal form on 4489needs a person
- Your book
- Bound YTD, commission statement
Status questions are a large share of inbound servicing calls. This removes the reason to make one.
Illustrative surface · product name and figures are examples, not a case study
On the screens people use all day · a broker, and your servicing desk
The status question answers itself. They see it themselves, scoped by entitlement to the programs they are appointed on. The call does not happen.
Moves the numberGross written premium per employee
Pick one number, and the next one gets cheaper
Fix the one that hurts most. Each one you fix makes the next cheaper, because the domain model and the gateway are already in place by the time you reach it.
Step 01
Bordereaux accuracy and timeliness
__ %
One process, one output, and a carrier who tells you immediately whether it worked.
Then makes possible
Forces the canonical model to be correct, which everything upstream then inherits.Step 02
Cost per policy bound
__ per policy
With the record authoritative, the bind path can be shortened without downstream risk.
Then makes possible
Makes the case for touching intake, the biggest single cost in the chain.Step 03
Time to launch a new program
__ weeks
A new program becomes configuration and composition rather than a release.
Then makes possible
Changes what the business is willing to attempt commercially.Step 04
Gross written premium per employee
__ GWP / FTE
The compounding number. It moves only after the first three.
Blank on purpose: real baselines with the first case, not illustrative figures.
We run it with you, so custom never means headcount
The reason custom software gets vetoed is not the build — it is the team you would have to hire to keep it alive. That team is us. You invest in underwriting and distribution; the platform headcount is the part of the deal we carry.
- Overnight
- Bordereaux runs for 3 carriers · completed 04:10
- Exceptions
- 2 routed to your program manager, with reasons attachedneeds a person
- Change
- Rating table v41 published by your underwriting team
- AI watch
- Claims-triage evals passed · no drift against baseline
- Estate
- All blocks healthy · audit log complete
A week of running it, at a glance. Your team decided twice; nobody you hired maintained anything.
Illustrative surface · product name and figures are examples, not a case study
Agents carry the routine work. Intake, triage, reconciliation and reporting land as drafts and exceptions, not as a backlog. Every judgment still goes to a named person on your side.
The people who built it run it. Releases, monitoring, model evaluation, the audit trail — operated by the team that wrote the code, inside your constraints.
Operated against the number you picked. The KPI ladder is the operating contract, not a slide. If the number stops moving, that is our problem before it is your project.
How you would start, and why it compounds
Nobody replaces an estate in one programme. Four stages, each standing on the last.
01
One workflow
An agent over a core you do not touch
Intake, or the bordereaux run. The domain model is defined for that slice only, the gateway is stood up, and nothing underneath is replaced.
02
One domain
The surfaces and systems around that workflow
The workspace the team works in, the portal the partner sees, and the two systems either side. The domain model extends to cover them rather than being rebuilt.
03
One program, end to end
A composed core for a single binder
Usually a new program, because there is nothing to preserve. This is where time-to-launch becomes the number.
04
The book
Every program on one stack
Each program after the first costs less, because it lands on a domain model and a gateway that already exist.
Nothing is thrown away between stages, and no stage assumes the next one. If you stop after the first, what you have still works and still pays.
Who we sit with before any of it gets built
The pain point comes from the person doing the work, and it is usually not what IT was asked to fix.
- We start with the people doing the work
- Underwriters, program managers, the person who assembles the bordereau. The pain point comes from them, and it is often not what IT was asked to fix.
- We prototype before anyone funds anything
- You see the surface and correct it while correcting it is still cheap.
- We run it, and it keeps changing
- Operating it is how we learn what to build next, which is why the number keeps moving after go-live.
Headless means a workflow takes only the blocks it needs
Headless means every block is an API before it is a screen, and none of them assumes the rest. So a workflow takes only the blocks it needs, they meet through the connecting layer rather than point to point, and the second composition is cheaper because the first one already taught the layer your business.
The shared tiles are the dependency story: claims cannot price a loss without the policy record, and most insurance work arrives as a document. Blocks need each other — they just meet in the connecting layer, so composing the second workflow rewires nothing in the first.
The building blocks, counted
An accelerator is a working system that fills one slot in the blueprint. Take it as it is, change it to match how you work, or leave it out and we integrate what you already have. You get the source, so changing it later is a code change in your repository rather than a request to us.
Systems of record
the ones the business is written into
2 built
Policy administration
Quote, bind, issue, endorse, renew, cancel. Configured per program rather than rebuilt per program.
or keep yours
CRM and customer 360
Client and policyholder onboarding, and one view of the customer the rest of it attaches to.
or keep yours
Business systems
the ones that decide and settle
7 built
Rating
Configurable product rating, versioned by the people who own the rates.
or keep yours
Eligibility and appetite
The binder's rules in one place, applied at the point of entry.
or keep yours
Claims
FNOL intake, adjudication, reserves and payments.
or keep yours
Billing and receivables
Invoicing, receipting and payment processing that becomes true at bind.
or keep yours
Insurance accounting
Insurance-specific operational accounting, so written, billed and ceded reconcile.
or keep yours
Commissions
Calculation and payout from the same record the policy was bound into.
or keep yours
Reinsurance and bordereaux
Bordereau generation and ceded premium tracking, structured per carrier.
or keep yours
Shared services
the ones everything else leans on
4 built
Intelligent document processing
Extraction and validation from unstructured submissions.
or keep yours
Document management
Storage, versioning and retrieval, next to the decision made from it.
or keep yours
Notifications
Email, SMS and push, on the events that matter.
or keep yours
Reporting and BI
Operational reports and the analytics layer, at program and book level.
or keep yours
Audit logging and the integration hub are deliberately missing from that list. They are not pieces you pick: they are how everything else is governed and reached, so they arrive with the gateway in every composition.
On buy versus build
If you already run Guidewire, Duck Creek, Sapiens or something like them, none of this asks you to replace it. Those platforms are the right answer to a question a lot of carriers have, and they sit under the gateway like any other system of record.
| A licensed platform | An InsureEdge accelerator | |
|---|---|---|
| What you get | A product with a roadmap, a version and a configuration surface. | A system we have built and run, given to you as source you can shape. |
| How it changes | Configuration inside what the vendor anticipated. Beyond that, a request. | Changed the way your own software is changed, because that is what it is. |
| What it costs to hold | Licence per seat or per policy, plus implementation. | Priced as a service: the composition, and running it, against agreed numbers. |
| Where it stops | At the edges of the product, which is where integration work begins. | It does not: the same gateway reaches your other systems, licensed ones included. |
| Who it suits | A carrier standardising a large, stable book on a proven platform. | An MGA or a program business that needs the shape to follow the deal. |
The honest version: if your book is large and stable and your processes are close to the market standard, a licensed platform is often the cheaper answer and we will tell you so. If your business changes shape every time you sign a binder, a product with a roadmap is the thing that will hurt.
Which of those eight is costing you the most?
We start from what you already run.