Software that fits your business like a glove.

Forward-deployed engineers for AI-assisted software builds, Palantir Foundry & AIP, Databricks and the Foundation reinsurance pricing platform. Bermuda-based, working worldwide.

AI has changed the build-vs-buy math for good. Our forward-deployed engineers embed with your team, run the workshop, prove the concept and ship the system — built around how you work, not the other way round.

Scope your first buildSee how an engagement runs
First call · 00:12:404 participants
TFTreefrog · Senior engineer
OLYou · Head of Operations
CLClaims lead
FDFinance director

Two ways to get AI out of the strategy deck and into production.

Forward-deployed AI engineering

A senior engineer embeds with your subject-matter experts, remotely and as deeply as the problem demands, until we understand the work as well as they do. In a workshop your people describe the system they would have if nobody told them no. Then we build that system — all of it — proving the AI parts on real data along the way.

  • Workshops where your team designs the system of their dreams
  • Living Requirements Document: that system, clickable, within days
  • Proofs of concept for the AI pieces, judged on real results
  • AI-powered builds that deliver the whole thing, not a prioritised slice
How forward deployment works →

Palantir Foundry & AIP builds

When Palantir is the right platform, the same engineers deliver on it. Right-sized Foundry and AIP engagements for firms that want the platform working without a programme-scale commitment.

  • Ontology, pipelines, Workshop apps and AIP agents
  • Fixed-scope first builds, typically weeks not quarters
  • A way of working that makes AIP builds dramatically faster
  • Extension of existing Foundry estates

Platforms we build on

The engineers are the practice. The platform is whichever one fits.

Databricks for the data, Power BI for the reporting, Palantir when the problem calls for it, Foundation for (re)insurance pricing, or the stack you already own. The choice is made for the problem, not for our margins.

Databricks

Nightly ingestion from SFTP, SharePoint, Azure file shares, websites and APIs; cleaning, validation and alerts; NAV calculations; one consolidated dataset powering reports and workbenches.

Data engineering on Databricks →

Power BI

Reports built with the latest AI agents, row-level security from day one, and Power BI integrated with source control so releases are repeatable.

Power BI done properly →

Palantir Foundry & AIP

Ontology, pipelines, Workshop apps and AIP agents as right-sized, fixed-scope builds in your tenancy.

The Palantir practice →

Foundation

Our own deal pricing and portfolio roll-up platform for reinsurance and ILS, built to run headless under your screens.

About Foundation →

AI models & agents

Anthropic, OpenAI and open models on private tenancy: assistants, agents and evaluation wired into the systems above, with the guardrails a regulated business needs.

Forward-deployed AI engineering →

Your own stack

Azure, .NET, SQL Server, Python, low-code — if that is what your team already runs, that is what we build on. No platform tax, nothing to re-learn.

How we work on your stack →

The build-vs-buy math has changed. Permanently.

For twenty years, building your own software was the expensive, risky option, so businesses bought off the shelf and bent themselves around it. AI-accelerated engineering has flipped that equation. Bespoke is now the fast, affordable choice, and the advantage goes to whoever acts first.

Buying off the shelf

Your business conforms to the software.

  • Workarounds, spreadsheets and re-keying fill the gaps the vendor never built
  • Your process is now the same as every competitor on the same platform
  • Licence fees climb every year; the roadmap is somebody else's
  • Change requests queue behind a thousand other customers
Building with Treefrog

The software conforms to your business.

  • Every screen, rule and report is shaped around how your people actually work
  • AI-accelerated builds land in weeks, at a fraction of what bespoke used to cost
  • You own it outright: no per-seat fees, no vendor roadmap, no lock-in
  • It becomes a competitive asset nobody else can buy

Weeks,
not years.

What used to be an eighteen-month programme is now a workshop, a proof of concept and a build. The cost of trying has collapsed, so the cost of waiting is what matters now.

Call us before
the RFP.

The best decision a business makes in this cycle is to talk to an engineer early. A few days of forward-deployed work gives you a clickable baseline to judge every vendor against, or to skip the vendor entirely.

Leapfrog,
don't catch up.

Catching up means buying what your competitors already have. Leapfrogging means building the system they will be trying to copy in two years. The window is open now.

Already bought the platform? Good. We'll make it fit.

Build vs buy is rarely all or nothing. The best platforms today run headless: the engine, the data and the analytics are there through an API, and the screens are whatever you want them to be. Your team describes the tool they wish they had; we build that tool on top of the platform you already pay for.

Our own Foundation platform works exactly this way: deal pricing, loss models and portfolio roll-up underneath, your underwriters' own screens on top. Every mode below lands in a fraction of the time a traditional software project takes.

Underwriting team whiteboard · what we actually want
One screen per deal. Layers, terms and price side by side.
Re-price in seconds when a layer changes.
Portfolio view the CUO will actually open.
"Price" ← big and obvious
Don't make us learn another vendor's screens.

IntegrateWrap AI around what you own

Connect your off-the-shelf system to the rest of the business, automate the gaps it never covered, and put an AI assistant on top of it. The vendor's roadmap stops being your ceiling.

Typically 2–6 weeks
ExtendBuild the pieces that make you different

Keep the commodity core, then build bespoke modules beside it for the workflows that are actually your edge: pricing, triage, reporting, client service.

Typically 4–10 weeks
BuildReplace it with something that fits

When the platform is the problem, build the system your business actually needs. AI-accelerated engineering makes bespoke the fast option, not the slow one.

Typically 8–16 weeks, not 18 months

Walk through an engagement

From first call to live system, stage by stage.

Click a stage or step through. Each one shows what happens, what you walk away with, and what exists on screen by then.

Day 0 · 30 minutes

The first call

You talk to a senior engineer, not a salesperson. We listen for the part of your business that runs on heroics, spreadsheets and workarounds, and tell you honestly whether it's a build, an integration, or not worth touching yet.

You walk away with: a scoped first build and a straight read on build vs buy, whether or not you hire us.

Your time: 30 minutes

Week 1 · Remote working sessions with your team

The workshop

Our engineer works with the people who do the work, not the org chart, in remote sessions that go as deep as needed. We map the real process, then ask the question that changes the project: what would this system do if nobody told you no? We write down everything, because we intend to build everything.

You walk away with: the complete wish list, a first sketch of the screens, and a plan to build all of it.

Your time: a few hours, on video

Weeks 2–4 · Something to click

Prove it

Within days your users are clicking through a Living Requirements Document: a working prototype of the system, with your data in it. The parts that depend on AI run as a proof of concept against your real cases and are judged on results, not demos. The system you asked for is not up for debate; how each AI piece earns its place is.

You walk away with: a prototype your team has already used, a POC with measured results, and a baseline to hold any vendor to.

Your time: 2–3 hours a week

Weeks 5–12 · Weekly demos

The build

The winners become production systems. Integrate around what you already own, extend it with bespoke modules, or replace it outright, on your stack, ours or Palantir. You see working software every week, and change your mind while it's still cheap to.

You walk away with: a system shaped around how your people work, tested by them along the way.

Your time: one demo a week

Week 12 and beyond · Yours

Live, and yours

Documented, handed over, owned outright. No per-seat licence, no vendor roadmap. Your team runs it; we're on call through FlexTech if you want us. And because it fits like a glove, the next improvement is a conversation, not a change request.

You walk away with: a competitive asset nobody else can buy, roughly a quarter after we first spoke.

Elapsed: about three months

Scope your first build
First call · 00:12:404 participants
TFTreefrog · Senior engineer
OLYou · Head of Operations
CLClaims lead
FDFinance director
Use case #1: claims triage. Needs AI summary on arrival.
Adjusters want one queue, not three.
Workshop sketch · day 3
Living Requirements Document · Claims triage · v0.4
Triage queueAdjuster viewReportsPoliciesSettings
Triage queue+ New claim
Open38
Awaiting docs12
Avg triage11 min
CLM-2041 · Property · HamiltonAI summary ready
CLM-2042 · Marine · DockyardNeeds documents
CLM-2043 · Liability · PagetRouted
Policy adminVendor system● Synced
Triage queueAdjuster viewReportsPolicies
Triage queueBuild · week 7
Open
38
Awaiting
12
Avg triage
11 min
CLM-2041 · PropertyLive data
CLM-2042 · MarineLive data
CLM-2043 · LiabilityLoading…
SSO · 2 min agoDeploy #14 ✓
Claims databaseSQL Server● Live feed
Document storeSharePoint● Synced
AI modelPrivate tenancy● Connected
Claims Triage · claims.yourcompany.com14 online
Triage queueAdjuster viewReportsPoliciesAudit logv2.3 · yours
TodayThursday · 09:41
Triaged today127
Avg triage4 min
Backlog0
CLM-3310 · Property · WarwickApproved
CLM-3311 · Motor · PembrokeAI summary ready
New claim routed to M. Simmons
What exists on screen at this stage
Stage 1 of 5
Featured article · LinkedIn

Is Software as a Service Dead? How AI is changing the build vs buy decision

An operations team built the system they would once have bought. What made it work, and where experienced engineering still decided the outcome.

Read on LinkedIn →
AnnouncementTreefrog now delivers Palantir Foundry & AIP builds The argumentWhy the build-vs-buy math has changed for good

The rest of the bench

Twenty-plus years of (re)insurance and ILS software, from Bermuda. All services →

Foundation PlatformCustom SoftwareFlexTech TeamVirtual Insurance CTOExcel TamingSystem IntegrationBusiness AnalysisOffice Automation

The sooner we talk, the further ahead you finish.

Thirty minutes with a senior engineer. You leave with a scoped first build and a clear read on build vs buy, whether or not you hire us.

Book a build call