Workshop: Design Your Quality Gates for AI-Assisted Delivery

By Michał (Mike) Buczko, quality and delivery leader at Carrier/Viessmann.

A quality gate that was designed for code written by people does not automatically hold once AI writes part of that code. In this half-day workshop you bring one pipeline from your own work, map where AI touches it and redesign each gate. You test the new design against failure cases from a real AI-built product before you leave.

What the workshop covers

• Mapping: you bring one pipeline and mark every point where AI touches it.
• Redesign, gate by gate: what evidence the gate needs, who or what reviews it, and what happens on failure.
• Stress test: you run your design against failure cases from a real AI-built product and see which gates would have caught them.
• Output: a one-page gate map and a release-evidence checklist for your pipeline.

Agenda

The full version runs 3.5 hours:

• 0:00 Welcome, groups, pick a pipeline
• 0:20 Exercise 1: map where AI touches the pipeline
• 1:05 Exercise 2: audit the gates you have today
• 1:40 Break
• 1:55 Exercise 3: redesign the gates, then a gallery walk
• 2:55 Exercise 4: stress test the gate map against failure cases
• 3:15 Checklist, commitments and close

Each exercise block is a five-minute input, group work at the tables and a short debrief. A three-hour version is possible.

Who it is for

Up to 24 people who own a delivery pipeline or can change one: quality leads, test managers, engineering managers and the engineers who maintain the pipeline. It works best when participants from the same company sit together. Each group needs one delivery pipeline that someone at the table knows well enough to draw from memory.

What you take away

• A one-page gate map of your own pipeline, showing where AI touches it and what each gate checks.
• A release-evidence checklist you can take back to your team.
• A way to test a gate design against known failure cases before you depend on it.

Where I have given it

New for 2027 — available for conferences and company events.

The failure cases come from the build I presented at CAST 2026. For the thinking behind the workshop, read open season as a quality gate. If you want the leadership argument without the hands-on part, see Who Signs the Release When AI Writes the Code?.

Slides and materials

This workshop is new, so there is no recording of it yet. Participants keep their gate map, their gate cards and the release-evidence checklist. The facilitation deck is available to organisers on request.

Format and practical details

• Length: half day, 3–4 hours.
• Group size: up to 24 people.
• Language: English or Polish. Polish title: "Zaprojektuj bramki jakości dla zespołu pracującego z AI"
• Level: intermediate. It suits teams that already use AI somewhere in delivery, or are about to.
• What to bring: one pipeline you know well enough to draw. No laptop is needed; the work is done on paper.
• Room and tech: tables for five or six small groups, flipcharts and a projector.
• Formats: conference workshop or in-house session.

Related talks

• Who Signs the Release When AI Writes the Code?
• The Speed Trap: Reporting AI Productivity to the Board Without Fooling Anyone
• From Idea to Working Product: How AI-DLC Unlocks Human Creativity (CAST 2026)

Book this workshop

To book this workshop, start at the Speaking page, book a 30-minute call or message me on LinkedIn.