What we build

Software shaped like your business, not like a template.

Four kinds of work, all of them bespoke. Every engagement ends the same way: the system runs in your accounts, the code is yours, and your team knows how it works.

01 / Custom internal systems

The tool no vendor sells you

Every business has a few processes that are genuinely its own. Those are exactly the ones off-the-shelf software handles worst — so they end up in a spreadsheet with a person guarding it.

We build the system that process deserves: your terminology, your rules, your edge cases, and only the features you will actually use. It is easier to learn than a configured platform precisely because it does not contain everything somebody else's business needed.

Typically

  • Admin consoles and back-office tools
  • Client and supplier portals
  • Quoting, booking, scheduling and job tracking
  • Inventory, asset and case management
  • Operational dashboards showing the numbers you manage by

02 / Workflow systems

The path work takes through your company

Intake, approvals, handoffs, chasing, reporting. The work that eats a day a week and that nobody owns, because it lives between the tools rather than inside one.

We map how it actually runs — including the steps that only exist because of a decision nobody remembers making — design the shortest version that still satisfies your real constraints, then build it. Exceptions escalate to a person with context attached, and monitoring tells you when something did not happen. That last part is what most automation skips.

Process mapping Intake & triage Approvals Handoffs & SLAs Scheduled jobs Exception handling Monitoring & alerting

Most companies already own the pieces — accounting, CRM, e-commerce, payroll, a warehouse system, three Google Sheets. The problem is that none of them agree, so somebody reconciles them by hand every month and the numbers are always a few days old.

We connect what you already pay for, decide which system is authoritative for each fact, and build the plumbing that keeps them consistent — so reporting comes out of the system instead of out of an afternoon's work.

03 / Integrations & data

One version of the truth

Not a rip-and-replace. Making the tools you have behave like one system.

System integration

APIs, webhooks and scheduled syncs between the tools you use, with retries and a log you can actually read when something disagrees.

Data pipelines

Moving and reshaping data on a schedule, with validation at the edges so bad records are caught on the way in rather than found in a report.

Reporting layer

A single place the numbers come from, so two people asking the same question get the same answer.

Migrations

Getting years of history out of a system you are leaving — including the parts the vendor's export quietly drops.

04 / Systems consultancy

What to build, what to buy, what to stop

A sequenced plan with honest numbers — including the parts we think you should not build.

Most companies do not need a transformation programme. They need to know which two or three systems are worth changing this year, in what order, and roughly what each costs to build and to run.

Sometimes the answer is a product you can buy for a fraction of a custom build. We would rather tell you that and keep the relationship than bill for the wrong project.

What you get

  • A systems audit: what you run on today and where it hurts
  • Build-versus-buy calls with the reasoning shown
  • A sequenced roadmap with dependencies made explicit
  • Running-cost estimates, not just build estimates
  • A short list of things to deliberately not do

A component, not the product

Where AI actually fits

We are glad to use it. We are not going to pretend it is the interesting part of your system.

Language models are genuinely good at the jobs where rules run out: reading messy documents, triaging free-text requests, summarising, drafting a first pass for a person to approve. Where a system needs that kind of judgement, we build it in — with guardrails, an evaluation set you can re-run, a cost ceiling and an audit trail.

Everywhere else, a rule, a form or a database constraint is cheaper, faster and does not drift. Most of what makes custom software valuable is the unglamorous part: the right data model, sane permissions, validation, and error handling that still works at 3am. We build that first.

Document understanding Free-text triage Drafting with human approval Retrieval over your data Guardrails & evals Cost controls

How it goes

The same three steps every time

01

Look at the actual work

We sit with the people doing the job and write down what really happens — including the spreadsheet nobody mentions in meetings.

02

Build the smallest useful thing

One workflow, end to end, in production, on real data. Something you can judge with your own eyes in weeks rather than a slide deck.

03

Widen it, then hand it over

Extend into the neighbouring processes, document it, train your team — and transfer the code and infrastructure into your own accounts.

Which of these is your problem?

Describe the process that is costing you the most. We will tell you which of the four applies — or that you should buy something instead.