Home
{
explore
}
what we do
AI Services
Playbooks
Industries
{
Company
}
How we work
Case studIEs
insights
Case study
Book a call
{
or email us
}
[email protected]
Book a call
{+} MISSION+ Case study · Public sector organisation

→ Engagement spotlight / Centre of Enablement

Agents migrated a Singapore public sector legacy application in five days, as part of a Centre of Enablement we are building together. The codebase was 350k lines

Yardstick

This was a proof run on a test project, to check agentic migration holds at that scale. One engineer ran it with up to 20 sub-agents per stage. Our estimate for doing it by hand was three engineers and close to a year.

Migration is one of several workstreams we run with them under a Centre of Enablement, alongside training and tooling. In each one we listen first, then recommend what we think is right for them. The goal is to leave their own teams more capable than we found them.

ClientA large public sector organisation in Singapore
EngagementCentre of Enablement covering secure AI tooling, legacy migration and training. Ongoing
MethodAgent Centric Development Cycle, specification before code
Year2026

→ The situation

Ageing systems to modernise, and five product teams each finding their own way with AI.

The client is a large public sector organisation in Singapore. It already wanted to change how it builds software, but needed to see the approach work on its own systems before committing.

The first concern was its legacy estate. Several internal applications have been in service for many years. They are expensive to maintain and thinly documented, and many of the people who built them have moved on, so every change is slow and cautious.

Another was how developers were using AI. Many had started writing code with it on whatever platform was open, and the output generally held up. But getting that code into the organisation's repositories was harder, because everything has to clear security scanning within a tight window, and code assembled outside a controlled path arrives with no record of how it was made.

A third concern came up once we started working with the product teams. Each of the five worked in its own way, down to keeping five separate brand guides, so their applications looked different and anything one team learned about working with agents tended to stay with that team.

By the time we met, the organisation had sat through plenty of AI strategy. It wanted a partner that would listen, give an honest recommendation, and do the work alongside its own people.

Paraphrased from the client's account of why they chose MISSION+

Just bring M+ Tech Leaders into the room and let them talk. They show you can build AI and, more importantly, build capability.

→ The partnership / how one piece of work led to the next

How the engagement grew from one conversation.

01

REX opened the door

We built part of REX, our innovation agent, for the client. The interest it created gave us room to talk about how they build software more broadly.

02

ACDC and training came next

Those conversations turned to ACDC, our agent-centric delivery cycle. Developers and product managers then trained in cohorts of around 30, first on working with agents and then on QA and writing specifications.

03

Migration followed from there

Having seen ACDC in practice, the client asked us to help migrate its legacy systems. A pod smaller than a traditional delivery team is doing that work now. In a proof run, it rebuilt a 350,000-line Java and JSP application in TypeScript in five days.

04

The client shaped the target

Midway through, the product managers we had trained redesigned the application themselves and asked to build their new design in place of migrating the old one. We advised migrating the existing system as it stands first, then making their improvements once it runs on the new stack.

05

A plan for what comes next, built together

The next stage is being planned with the client, piece by piece, around the capability its own teams need. The Centre of Enablement is where that plan lives.

Inside the training / the lifecycle, mint marks a human decision

Prototype→ Specify→ Ticket→ Build→ Validate→ Verify→ Release

↻ Loop When validation fails, the work goes back to the specification before anyone touches the code. The sessions are designed to make that a habit.

→ The work so far

Where the engagement stands.

Proof run: a 350,000-line Java and JSP application rebuilt in TypeScript, to a running development build

Elapsed time

Agentic runOne engineer, up to 20 sub-agents per stage

5 days

Hand-written estimateThree engineers, our own projection

close to a year
03 mths6 mths1 year
60

Developers and product managers trained

Across two sessions in cohorts of around 30, with a session planned every quarter from here.

5

Product teams in scope

Each goes through the same training and works to the same standards under the Centre of Enablement.

1

Team co-locating with us

We co-locate with the team and work to cross-pollinate ideas.

The five-day run was a proof

We carried it out for this client on a test project to show the approach holds at real size. It reached a running development build. Production release was outside its scope.

The training is at an early stage

Two sessions is a starting point. The client intends to grow it into an internal AI academy over the next six months.

One set of standards across the teams

The Centre of Enablement sits above the five product teams and keeps a common approach to process and tooling, so whatever works for one team reaches the others.

The tooling is developer-first for now

Our contained agent generation engine ships as a standard kit with practices and agents already installed, so every new developer starts from the same place. A version for product managers is planned.

→ Why it matters

A Centre of Enablement, with migration as one part of it.

Each piece of work has led to the next. Interest in REX started the conversation, and the ACDC training grew out of it. The migration came after that, because by then the client's teams had seen how we work.

Meanwhile the training is building the specification and QA habits the client's people need to carry on without us. The product managers who redesigned the migration target had been through two sessions at that point.

We are now planning the next stage with the client, centred on capability. The Centre of Enablement keeps common standards across the five product teams, so each new migration or team starts from what the last one learned.

MISSION+
Your product will define your business, build it  with the leaders who've done it before.
{
Subscribe
}
Insights delivered to your inbox
Please enter a valid email address.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
{
what we do
}
{
01
}
AI Services
{
02
}
Playbooks
{
03
}
Industries
{
company
}
{
01
}
How we work
{
02
}
Case studIES
{
03
}
INSIGHTS
{
connect
}
{
01
}
contact us
{
02
}
linkedin
{
Singapore
}
45 Cantonment Road, Singapore 089748
{
uae
}
Innovation One, Dubai International Financial Centre, Dubai, UAE
{
vietnam
}
47 Street 23, Giao Luu Urban Area, Dong Ngac Ward, Hanoi City
© 2026 MISSION+. All rights reserved.
Privacy policy
Terms of use
Terms and conditions