For software companies

Agentic SDLC. Move your engineering team to AI agents, starting with what's already there.

Agents speed up whatever a team already has, including its debt. We read the real system first, say what's ready, and work with your team and delivery partners on the foundation, so agent-written code can ship under human review.

Start a conversation See the companies

TL;DR

An agentic SDLC (software development lifecycle) has AI agents write, test and document code changes. People set direction, review the work and approve what ships. We start with a read-only audit of your code, tickets and pipelines, build the foundation agents need, then ship one reviewed change before anything scales.

How we work

Read the real system
A read-only audit of your code, tickets and pipelines, with no changes to your code.
Say what's ready
A plain scorecard of what will hold up under agents and what won't.
Build the foundation first
Tests, tracing from ticket to code, real review and pipelines kept as code.
Ship one governed change
One agent-written change on real code, through tests and human review, before anything scales.
Move the team
New ways of working, clear owners and support for the people changing how they work.

Where this is real today

Aprio, under way

Aprio makes board portal software. We've worked with them since 2024, first on go-to-market and sales. By late 2025 the bigger question was engineering: could the team move to building with AI agents?

We started in spring 2026 with workshops, then changed course and audited the real system instead: a read-only look at Aprio's code, tickets and pipelines. We found sound architecture with practices that weren't ready for agents. In June we put the readiness plan into Aprio's own ticket system, and agreed with their team to start with the back end: every ticket linked to a pull request, then tests, quality gates and continuous integration. In July we set out a modernization blueprint and what a delivery partner would need to do.

Aprio then chose a delivery partner, whose 16-week project began in October 2026. The plan is to document, test and gate the code before agents write any of it. That team has been writing documentation and tests for review and setting up the agent workspace, and the plan is now moving toward building the new delivery setup alongside the current service. The first agent-written change on Aprio's real code is still ahead. We're still working with Aprio's leadership while it happens.

Our own work

We run our own software this way. Agents write the change, separate checks test it, and a person approves what ships.

Golden Rule

We built AI skills and data connectors that Golden Rule's team runs from its own accounts.

Read the story

Common questions.

What is an agentic SDLC?

A software development lifecycle where AI agents write, test and document changes, and people set direction, review the work and approve what ships.

How is this different from AI-assisted coding?

With AI-assisted coding, an engineer directs each step and the tool suggests the next lines. In an agentic SDLC, the engineer hands agents an outcome, and they plan, write, test and document the change for a person to review.

Where should a team start?

With a read-only look at the code, tickets and pipelines it already has. Agents speed up whatever is there, so tests, tracing and review come first.

How do you keep agents under control?

Agents get only the access a job needs, and their work leaves a record. Tests come first, separate checks review each change, and a person approves what ships.

Do you do the engineering?

We build agentic systems on our own work and for clients, and on a large migration we may work alongside your team and a delivery partner.

What results should I expect?

That depends on your starting point, so we don't quote a result. We start by measuring what you have today.

Would your codebase hold up if agents started writing it tomorrow?

A short conversation. We'll start with a read-only look at your code, tickets and pipelines, and tell you plainly what's ready.