AI agents turn a backlog into mergeable pull requests

<aside>

🇬🇧 English 🇩🇪 Deutsch

</aside>

Even with an AI coding assistant, a developer prompts, reads and corrects every change by hand. Developer attention sets the pace, and your backlog waits for it.

Agents can turn a ticket into a pull request unsupervised when it arrives as a spec they can execute, with acceptance criteria they can check themselves. Your developers review the finished pull request. Addy Osmani calls this loop engineering.

The numbers are public, with methods, baselines and limits:

The bug-fix report covers three Kimai bugs, from ticket to pull request in about 13 hours, estimated review time included. A developer and a QA engineer using AI copilots would need 148 hours. None of the three fixes adds an abstraction it doesn't need.

The regression-test report covers 350 Playwright regression test cases: 43.3 hours estimated, against 191 hours for a Playwright agency.

I set this up for your team and coach your developers to run it. Start with an audit, a pilot on one backlog item, or a workshop. Each is bounded and priced upfront.

What I delivered for an open-source project

Three Kimai bugs fixed, ticket to pull request: ~11× faster and cheaper

Regression tests, 55 of 350 test cases landed: whole backlog estimated ~4× faster, ~3× cheaper

Who I work with

Software houses and consultancies delivering client projects on headcount they can't grow fast enough. I turn requirement specs and change requests into inputs that agents can execute. I add a tracked regression suite for the integrations and permissions your clients test at acceptance. Each regression test run ends in an execution report with release gates you can hand over, such as pass rate and regressions list. The delivery commitments and estimates stay yours.
SaaS product teams with a UI-heavy product whose bug and feature backlog has outgrown the team, and where each release needs a regression check. Agents turn tickets into pull requests that your developers review. A tracked Playwright suite shows what passes before each release. The product decisions stay yours.
Mittelstand companies with an in-house software team whose backlog waits on developers they can’t hire fast enough. Agents turn tickets into pull requests that developers review. A tracked regression suite shows what passes before each release. The priorities and release decisions stay yours.

<aside> 🍸

How this ends: you own the product and the customers. I build the process that ships your decisions, which your team runs without me.

</aside>

How we can start

1️⃣ Delivery Audit

Fixed-fee assessment of how ready your team is to let agents build from your backlog.

Scored gap report on how your tickets are specified, split into tasks and tested, plus a 90-day roadmap.
Includes an Agentic Software Delivery Readiness scorecard you can hand to enterprise clients and customers.
Useful on its own, even if you decide not to continue.