AI SDLC

Improve software delivery with coding agents

I help engineering teams define where coding agents fit, improve the context and checks around them, and measure delivery time, review effort, rework, and defects.

The problem

Coding agents can increase the amount of code a team produces without improving delivery. Specifications may be incomplete, repository context may be scattered, and review capacity may not match the volume of changes. Teams also need evidence that shows whether coding agents change delivery time, review effort, rework, or defects.

How I work

I assess how changes move through the client engineering team: how work is specified, what context an agent receives, what it may change, which checks run, and how a reviewer confirms the result. My team and I then improve the parts that limit safe delivery and measure the agreed outcomes over a defined period.

How this connects to my work

I lead engineering at KnackLabs, where our engineers use Symphony Forge as our software factory process. That experience informs my work with client teams; the engagement is adapted to their repositories, delivery process and constraints. See the products I have built and led.

What an engagement covers

Current delivery process

Repository, ticket, review, and release data used to identify where coding agents help and where they add review or rework.

Specification and context

Clear task briefs, relevant repository decisions, and codebase-specific instructions that give an agent enough information to work within scope.

Review and tests

Automated checks for important behavior and a review process that compares the approved task with the resulting change.

Permissions

Explicit rules for the commands, files, environments, and external actions an agent may use, including actions that require a person.

Task evaluation

A small set of representative client tasks used to compare models, settings, cost, and completed results in the client's own environment.

Measurement

A baseline and defined observation period for delivery time, review effort, rework, and defects.

Team practice

Documented working practices and training that the client engineering team can maintain as tools and models change.

Related writing

• The cheapest model can cost you more

  A cheaper model can need more attempts to finish the same task. Compare the full cost, including thinking effort, caching, and review time.