Practice 03
Custom software and modernisation
The systems that decide whether an operation runs are rarely the ones anyone photographs: claims intake, dispatch, reconciliation, document control, approvals.
At a glance
We build and rebuild the operational software an organisation depends on, without a three-year rewrite nobody survives.
Tools we use most
- Java / Spring Boot
- Python / FastAPI
- .NET
- Node.js
- React
- Next.js
- PostgreSQL
- Temporal
- Redis
- RabbitMQ
- Kafka
- Docker
- Kubernetes
- Playwright
Related work
- A vendor portal that replaced 11,000 emails
- Fleet update software for imaging consoles
- Routing and SLA escalation for a public-sector grievance system
What we get called in for
- A system built by someone who left, running on a framework that stopped receiving patches.
- Six teams, six integrations, no contract between any of them.
- A spreadsheet that quietly became a business-critical application.
- A vendor product that fits 70% of the process, and the other 30% is where the margin lives.
- An audit finding that requires you to prove who changed what, and when.
- Approvals and escalations still get tracked by reply-all, because the workflow tool that is already licensed does not fit how the team actually works.
How we build it
Trace the real process first
We follow one real case end to end, including the WhatsApp message and the exception nobody documented. The gap between the process diagram and reality is the project.
Strangle, do not rewrite
New capability goes in a new service behind a facade; the legacy path keeps running until traffic has moved. Big-bang cutovers are how modernisation programmes end up in board minutes.
Contract the interfaces
Two teams building against each other's systems agree the shape of the exchange — an OpenAPI or event schema — before either writes code, and tests enforce it automatically from then on. Integrations stop being a coordination meeting and start being something a build either passes or fails.
Design for the exception
Any operational system spends most of its life handling the 5% that does not fit. We build the override, the audit trail and the reversal path as first-class features.
Leave it maintainable
Readable code, decisions recorded, environments reproducible, and a handover your engineers can accept without a hostage negotiation.
What you get
- Process trace and gap analysis against the system as built
- Target architecture with a migration sequence and rollback points
- The software: services, integrations, admin tooling and user interfaces
- Automated test suite, CI pipeline and reproducible environments
- Audit trail, role model and data retention design
- Workflow and approval orchestration: routing, escalation, SLA tracking and a full audit trail — a BPM engine or a purpose-built state machine, chosen on the requirement, not the acronym
- Architecture decision records and a handover pack
Further reading
Next step
Tell us what you're trying to ship.
Send the brief, the RFP, or three messy sentences about the problem. You get a written point of view from an architect within two working days — not a sales deck.