Atlas/ AI-Assisted TDD Workshop Playbook/ Facilitation & Failure Modes

Workshop Ops Scoreboard

90min session

3roles

10failure modes

15sources

Facilitation Roles — leads must never operate platform controls [7]

LEAD Lead

Delivers content & runs exercises
Timeboxes discussions, calls breaks
Reads the room — adjusts energy
Never touches Zoom controls

Task-switching to Zoom during content kills pacing

PRODUCER Producer

Manages Zoom polls & breakouts
Watches chat; voices questions aloud to Lead
Handles recording & countdown timer
Posts exercise text to chat (verbal-only missed by stragglers)

Answers logistics in chat so Lead is never interrupted

HELPER(S) Helper(s)

1 per breakout room of 4–6
Unsticks participants; demos on own screen only
Joins room first 3 min, then cycles
Full material knowledge not required

Never takes over a participant's keyboard

Session Timeline

Check-in

Context

Exercise 1 (35 min)

↗

Exercise 2 (30 min)

Debrief

00:00

00:10

00:20

00:55

01:00

01:30

01:40

Hands-on exercises (65 min · 59%)

Facilitated discussion

Break (non-negotiable)

Pre-Mortem Failure Matrix — 10 predictable modes

Impact	Failure Mode	Prevention
CRITICAL	Env setup in live session Loses 20–30 min; derails all subsequent timings	Codespaces prebuild + mandatory pre-check 24 hrs before ^[6]
CRITICAL	AI API key failure on day Blocks all exercises; kills workshop credibility	Per-participant keys with expiry; day-before `curl` test required to claim ^[9]
HIGH	Demo-heavy, hands-on-light Expert disengagement within 15 min	Hard rule: ≤7 min explanation before participants touch code; enforce in run-of-show ^[2]
HIGH	Dominant expert hijacking Others disengage; session follows one rabbit hole	Parking lot + timebox; round-robin debrief; silent brainstorm before open floor ^[5]
HIGH	Exercise too ambiguous Participants stuck; helpers overwhelmed; pacing collapses	Test every exercise solo end-to-end before session; embed "if stuck" hints as stub comments ^[10]
HIGH	No helper in breakout rooms Stuck participants wait silently; frustration builds	1 helper per room of 4–6; briefed on exercise goals; arrives in room first 3 min ^[7]
HIGH	Expert resistance to AI tooling Overt scepticism infects room culture	Address AI limits explicitly in context frame; peer-champion framing; concrete first win ^[4]
MEDIUM	Tool sprawl Cognitive overload; participants lose their place	One primary tool per task; introduce sequentially; avoid simultaneous Zoom + Miro + Slack + IDE ^[3]
MEDIUM	Overrun debrief, no synthesis time Participants leave with open loops	Hard 10-min closing slot in run-of-show; parking lot absorbs overflow; written recap within 24 hrs ^[2]
MEDIUM	No break in session Focus degrades in last 30 min; diminishing returns	5-min break at midpoint, non-negotiable even under time pressure ^[10]

Expert Resistance Patterns — peer-driven adoption is 22% more effective than mandate [4]

⚡

Resistance #1

Trust / Reliability

"AI makes subtle errors on complex code"

Demo AI failures in Exercise 2; make catching bugs the learning objective

🔑

Resistance #2

Identity Threat

"This diminishes craft and intellectual satisfaction"

Frame as augmentation not replacement; use peer-credibility over mandate

⚙

Resistance #3

Complexity Gap

"It doesn't understand our architecture"

Explicit sandbox safety: workshop is a protected experiment; failures are data

⏱

Resistance #4

Time Cost

"I don't have cycles to learn another tool"

Concrete first win in <5 min — passing test before scepticism can calcify

Works

"Here's what I tried" framing over "this is the standard". Peer-driven adoption 22% more effective. ^[4] Silent brainstorm before group discussion prevents groupthink. ^[5]

Fails

"This is the new standard" triggers identity resistance. Skipping the AI-limits discussion — experts find gaps mid-exercise and derail the group if not pre-empted. ^[4]

Grouping

Moderate cognitive diversity — inverted-U efficiency curve. Avoid all-sceptics or all-enthusiasts rooms. ^[15] Pre-assign; random assignment feels infantilising to expert audiences. ^[14]

Pre-Session Gate — send 72 hrs before; treat as a hard gate, not optional prep

Codespace open — GitHub account; workshop link opens; VS Code loads with test file visible

AI extension authenticated — generate a "hello world" function and accept it

API key claimed — run the provided curl snippet; confirm a valid response

Zoom desktop client installed — browser Zoom breaks screen sharing in breakout rooms

Also in this playbook

survey AI-assisted TDD techniques & evidence (2026) survey Workshop exercises & scaffolding survey Tooling stack for the workshop

Sources