Workshop Pack - Paired Engineering with AI

Usage note

This workshop pack, Practitioner Workshop Deck Outline - Paired Engineering with AI, and Practitioner Workshop Deck Slide Copy - Paired Engineering with AI now form the accepted locked markdown baseline for the practitioner workshop package.

Use this note as the facilitator and exercise layer unless a substantive audience or content gap appears.

Purpose

This draft turns the practitioner playbook into a teachable session for pilot teams.

Audience

Developers, QA, SDET, architects, and product owners in the first pilot wave.

Workshop goals

• understand the paired-engineering stance
• distinguish learning mode from delivery mode
• apply capability-aware rules to realistic work
• practice verification habits on AI-assisted tasks

Suggested format

• 15 min framing and evidence summary
• 15 min capability model walkthrough
• 20 min live paired example
• 30 min role-based exercises
• 10 min debrief and team commitments

Facilitator notes

• emphasize that oversight readiness is not job title alone
• reinforce that explanation is not verification
• keep examples grounded in real delivery work
• ask participants to name where verification is easy and where it is hard
• avoid learner-style labels; vary examples and supports without pretending each person needs a unique style match
• reference the supporting learning-design notes when formalizing this workshop further

Opening script

AI can help us move faster, but speed is not the only outcome that matters.

In this pilot, we are using AI as a paired-engineering aid. That means we want better understanding, better debugging, better review, and safer acceleration. We are not treating AI as a substitute for engineering judgment.

Exercise 1. Mode selection

Prompt:

Given a task, decide whether it belongs primarily in learning mode or delivery mode, then explain why.

Scenarios:

• new internal framework, first feature in the codebase
• familiar service, localized unit test updates
• production incident root-cause analysis with incomplete evidence
• rewriting acceptance criteria for a known backlog pattern

Debrief questions:

• what made the task unfamiliar
• what made the task hard to verify
• what would have made you overconfident

Exercise 2. Verification check

Prompt:

Participants review an AI answer and list what would count as real verification before use.

Debrief questions:

• what evidence would you trust
• what would still worry you
• what would require peer review or stakeholder review

Role-specific exercises

Developers

Scenario:

Use AI to help debug a bounded defect in a familiar service. Then decide what you would still need to inspect manually.

Expected focus:

• explanation of failure mechanism
• test-based verification
• checking hidden assumptions

QA and SDET

Scenario:

Use AI to propose additional test cases for a flaky workflow. Then separate good candidates from shallow or redundant ones.

Expected focus:

• edge cases
• data sensitivity
• whether the proposed cases would actually catch failures

Architects

Scenario:

Use AI to generate two architecture options for a new integration. Then critique the missing tradeoffs and unknowns.

Expected focus:

• unsupported claims
• operational consequences
• where more evidence is needed before decision

Product owners

Scenario:

Use AI to improve acceptance criteria and stakeholder questions for a story with ambiguous behavior. Then identify what the model cannot know.

Expected focus:

• hidden business assumptions
• semantic correctness
• unresolved stakeholder decisions

Debrief prompts

• where did AI genuinely help
• where did it sound good without being trustworthy
• what verification step caught the biggest issue
• what would an engineer with lower oversight readiness likely miss here

Team commitments

Ask each participant to write down:

• one workflow they will pilot
• one anti-pattern they will avoid
• one verification habit they will adopt

Follow-up materials

• Practitioner Playbook - Paired Engineering with AI
• Exercise Library - Paired Engineering with AI
• Capability Model - Oversight Readiness x Task Familiarity x Risk
• Rollout Matrix - Role-Specific Workflows by Phase
• Rollout Matrix - Capability Rules by Phase

Deeper exercise extension

Use Exercise Library - Paired Engineering with AI when the base workshop needs a stronger progression path for junior, intermediate, senior, or Staff practitioners.

That library:

• stays mostly language-agnostic
• rates exercises by size and complexity
• emphasizes verification and judgment rather than syntax drills
• gives facilitators a reusable progression beyond the first workshop session

Use Exercise Worksheet Pack - Paired Engineering with AI when the facilitator needs ready-to-run worksheet structure, timing, traps, and debrief guidance.