Application Development & Maintenance
Enterprise software, delivered by small AI pods.
We have delivered mission-critical software for enterprise clients since 2006. Today we also deliver through small pods of senior people working with our Harnesses, where AI-led delivery fits the work and your governance allows it.
How it works
The work organised around named approvals.
A different way to organise the work
A pod of two senior people, a Forward Deployed Engineer and a Product Engineer, runs three Harnesses. The steps between approvals are done by the Harnesses, and the people stay in charge of each decision.
Harnesses do the volume. People hold the gates.
The Harnesses draft requirements, plans and code. Named people on your side approve each stage.
How a pod works
Two people, three Harnesses, every change approved.
The Harnesses
Three Harnesses, and where each one stands today.
Together they cover requirements, engineering and quality, with requirements live today. Where AI-led delivery doesn't fit, we deliver conventionally.
Requirements Harness
Elaborates requirements and user stories for your people to review and approve.
Engineering Harness
Plans and builds each story and runs the quality gates, leaving the change ready for a person to review and merge. Built, and going into its first client use.
Quality Harness
Generates and runs test cases from the story. A QA agent is being trialled.
Approvals
Four named approvals before anything is built.
Each gate has a named reviewer and approver on your side. After G4, the workflow checks that a story is ready for build, so readiness is a rule in the tool and not a meeting.
G1: Requirement
A named person on your side approves the fresh requirement before the Requirements Harness elaborates it.
G2: Requirement approval
The elaborated requirement is reviewed and approved by named people on your side.
G3: Design approval
The design is approved before any user stories are written.
G4: Story approval
The user stories are approved. Only then can a story reach build.
Rules the Harnesses work under
What the Harnesses are not allowed to do.
Merge their own work
A person merges every change. An agent never merges its own work.
Write to your systems of record
The model has no write path to your work-tracking tool or any other system of record. Writes follow a person's confirmation.
Outrun what was approved
An approval binds to the exact plan approved. Edit the plan afterwards and the approval is void.
Mark their own homework
Evidence comes from the pipeline that ran, not from an agent saying the tests passed.
Flow, where it fits
Hybrid delivery, run the way the work needs.
AI-led engagements run on flow, not sprints
Two boards, Planning and Live, share one status, Ready for Development. Reporting is about release readiness: where each release and its UAT stand against plan.
Conventional delivery stays available
Where Scrum or another established method suits the work or your organisation, we deliver that, with the same senior engineers and the same standards.
Evidence from a pilot
A two-week pilot in August 2026 with a North American education-technology client.
taken to opened merge requests by a two-person pod, in a two-week pilot in August 2026 with a North American education-technology client.
elapsed, human approval gates included. Self-reported from the pilot's run logs.
went on access and configuration before the first story ran; about two days were spent running. The merge requests were opened, not yet merged, and quality outcomes are not yet measured.
Case studies
Software we have built for clients since 2006.
Four systems from our delivery record, each on its own page, described without client names, logos or data.
One sign-in to every learning app in a school district
A portal that brings a district's applications together behind single sign-on, with an administration console and a native iOS app.
Supplier invoices into the accounts system, without retyping
A pipeline that reads each PDF invoice into structured data as it arrives and posts it to any accounting package with a web service.
Road reports that help a superintendent decide whether to close schools
A progressive web app for reporters on the road, with a live map and push notifications for the person who makes the call.
A shop's news and offers on the customer's phone, in the shop
A hybrid mobile app in which customers check in at a shop, then see its news, offers and codes to redeem.
Read more
How we measure, and how we are governed.
What a two-person pod did in two days, and what it cost to get there
What we did and didn't measure, and what changed.
The rules our delivery agents work under
The standing rules, written for a CTO and their security reviewer.
Where delivery time goes: nearly half of it is waiting
In 587 timed stories (of 1,244 in one client's work-tracking history, snapshot 24 September 2026), about 47% of story time was spent waiting: statuses where a story sits between people, such as ready, blocked, or finished one stage and waiting for the next. A further 36% of stories (a share of stories, not of time) went back to an earlier status.
Maintained after go-live
We look after what we build, so it keeps working long after launch.
Have something to build?
Tell us what it needs to do. We will tell you how a pod would deliver it.