Home
{
explore
}
what we do
AI Services
Playbooks
Industries
{
Company
}
How we work
Case studIEs
insights
Book a call
{
or email us
}
[email protected]
{
Public Sector
}

Public sector systems rebuilt to hold up under scrutiny

We help government and public bodies replace ageing systems with digital services that actually work, and that clear security review, audit and procurement along the way.

Book a call
{
challenges
}

Ageing systems, thin documentation, no margin for error

Public services often run on applications that have been in place for many years. They are expensive to maintain and lightly documented, and many of the people who built them have moved on, so every change is slow and cautious. Developers are already writing code with AI tools, but code produced outside a controlled path arrives with no record of how it was made. Anything new still has to clear security scanning, audit and public scrutiny.

{
what we do
}

Modernisation that clears review

Legacy migration with agents

Agents recover the business logic in existing code and rebuild it on a modern stack. In one proof run, a 350,000-line Java application reached a running TypeScript build in five days.

A controlled path for AI-written code

One route for AI-assisted code into your repositories, with security scanning on the way in and a record of how every change was made.

Specification before code

Every build starts from a written specification, so auditors and reviewers can trace any change back to the requirement it serves.

A Centre of Enablement

Shared standards, tooling and practices across your product teams, so an approach that works for one team reaches the others.

Training your own people

Cohort training for developers and product managers on working with agents and writing specifications, so your teams can carry the work on without us.

How ATOM applies

ATOM is our operating model for agentic engineering teams. In the public sector, it gives agencies a way to modernise ageing systems with small teams, working to standards that satisfy security review and audit.

The unit is typically a pod of four. A senior engineer owns the specification and the architecture, and two juniors orchestrate the agents. A business expert from the agency owns the acceptance criteria, which keeps policy and service knowledge with your own people.

We are running this model with a large public sector organisation in Singapore. As a proof run, one engineer with up to 20 sub-agents rebuilt a 350,000-line Java and JSP application in TypeScript, reaching a running development build in five days. Our estimate for doing the same work by hand was three engineers and close to a year. A pod smaller than a traditional delivery team is now migrating one of the organisation’s live applications.

ATOM covers the people side as well. Around 60 developers and product managers there have been trained so far, in cohorts of about 30, on working with agents and writing specifications. A Centre of Enablement sits above five product teams and keeps one set of standards across them, so each new migration starts from what the last one learned.

We start with a two to four week ATOM Maturity Assessment, and the first pod is usually live by week six.

Frequently asked questions

How can government agencies modernise legacy systems with AI safely?

Agents recover the business logic in existing code and rebuild it on a modern stack, working from a written specification that auditors and reviewers can trace. All AI-assisted code enters your repositories through one controlled path, with security scanning on the way in.

How do you make AI-written code auditable?

Every change has a record of how it was made and which requirement it serves. Code produced outside that path has no history, which is why we route AI-assisted work through a single governed pipeline before it reaches your repositories.

How fast can a legacy application be migrated with agents?

In a proof run for a large public sector organisation in Singapore, one engineer with up to 20 sub-agents rebuilt a 350,000-line Java and JSP application in TypeScript, reaching a running development build in five days. Our estimate for the same work by hand was three engineers and close to a year.

Do you train our own staff to work with AI agents?

Yes. At one public sector organisation, around 60 developers and product managers have been trained so far, in cohorts of about 30, on working with agents and writing specifications, so your teams can carry the work on without us.

How does a public sector engagement start?

With a two to four week ATOM Maturity Assessment. The first pod is usually live by week six, and a Centre of Enablement keeps one set of standards across product teams as more migrations follow.