back to work
case study ENG-AIRBNB airbnb · 2026
airbnb

ai enablement
at fortune 500 scale.

Airbnb's Employee Experience teams ran a multi-track AI enablement effort built around how people actually adopt new tools. I designed and built five performance-task modules for AI101, the enablement program at the center of the initiative. The work shipped in ninety days and produced 100% reported knowledge gain across the program.

01
the challenge

AI fluency at scale.
Without losing humans
in the workflow.

AI enablement inside Fortune 500s fails the same way: generic modules stapled onto existing workflows, no real measurement of adoption, no protection for the human side of the rollout.

Airbnb's Employee Experience teams needed different. AI101: five performance-task modules, ninety days kickoff to launch, designed so the person at the desk on Monday would open the tool again on Tuesday.

02
the approach

Performance task,
not lecture.

Every module is a hands-on lab. Learners build, critique, and ship real artifacts with AI inside their actual workflow. No sandbox, no seat time, no completion certificates.

Behavior change is the foundation. AI fluency is the outcome. They ship together or neither sticks.

See my Greenroom example in Labs for the type of thinking and solution I bring to this kind of work →

03
the solution

Five labs.
Designed around real
adoption moments.

Each module starts with a real adoption moment EX teams face daily and ends with an artifact the learner keeps using on Monday morning.

Action Mapping locked the design to those moments. Kirkpatrick-aligned measurement framed the evaluation. Strategy clarity jumped 25 points; every learner reported knowledge gain.

the public lab · built in parallel

the work stays internal.
the proof doesn't.

AI101 and the manager skill live inside Airbnb. Greenroom is the lab I built in parallel so the thinking has a public address: a working development copilot with real RAG, a real eval harness, and a documented path to production. Embedded below, running live.

spotlight · greenroom

greenroom.
running live.

Ask it a real talent-development question. Watch it retrieve, cite its sources, refuse the ones out of scope, and follow up. Performance over content. Behavior change over checkbox completion.

read the full lab writeup →
greenroom-strand.netlify.app open full screen ↗
live lab greenroom, live. Best experienced full screen on tablet or desktop. Tap to open the working prototype in a new tab. open the live lab
inside a module · representative re-creation

step inside one performance task.

The five AI101 modules are internal to Airbnb. So here is the pattern instead of the content: one of the most common AI tasks on EX teams, turning messy meeting notes into a decision brief, rebuilt as a click-through so you can feel how a performance task works.

The pattern is real. The content is not Airbnb's. Every name, number, and note below is invented for public view.

Step 1 of 4
step 1 · the adoption moment

monday, 9:14 am.

You just left a 45-minute stakeholder sync. Your director needs a decision brief by noon.

sync notes · 9:14 am

  • kickoff recap... timeline slipping, eng can't commit until Q3
  • 3 vendor options: Atlas / Beacon / build in-house
  • L pushed back hard on Atlas cost (~$240k/yr, "not happening")
  • pilot group = new managers, ~60 ppl?
  • IT security review for any vendor. 4 to 6 wks. J to confirm
  • DECISION NEEDED by EOM or we lose the Q3 window!!
  • actions: A sends Beacon demo. M drafts success metrics
  • parking lot: LMS integration, data residency??

Real input, real deadline. The module starts where the work starts.

step 2 · build with ai

what do you hand the model?

The first skill AI101 teaches isn't prompting. It's context. Pick your move:

step 3 · the human pass

the draft is back. now you audit.

Eight seconds later you have a one-page draft. Three spots deserve a closer look. Hover or tap each highlight to audit before you ship.

decision-brief-DRAFT generated in 8s

Recommendation: proceed with Atlas. At $240k per year it is the clear cost leader. The security review will finish within four weeks, keeping the rollout ahead of schedule.

The pilot launches with approximately 60 new managers. The Beacon demo and success metrics workstream are already in motion. what's missing here?

the audit

Hover each highlighted passage to see what deserves a second look before you ship.

Hover or tap all 3 highlights to continue

step 4 · ship the artifact

the artifact is the assessment.

You fix all three problems and send the brief at 11:42. There is no quiz. The brief is the evidence.

decision-brief-FINAL sent 11:42 am

Decision needed by end of month: select a vendor or commit to building in-house. Missing the date closes the Q3 window.

Atlas is strongest on features; its $240k per year cost drew a hard objection. Beacon demo is scheduled. Any vendor path requires an IT security review, estimated 4 to 6 weeks and not yet confirmed. Recommend a vendor decision meeting this week.

  • the pattern Every AI101 module ends with something built, fixed, and shipped inside the learner's real workflow.
  • why it works The tool earns its place on Monday, so it gets opened again on Tuesday. That is what built for adoption means.
  • the receipts 100% of learners reported knowledge gain. Strategy clarity rose 25 points. Tasks like this one did that, not slides.
  • one honest limit This re-creation runs one fixed path for everyone. The real modules met each learner where they were. The manager skill below takes that even further.
also shipped · a skill for people managers

a coach that meets managers where they are.

Beyond the modules, I built a Claude skill that coaches people managers through difficult conversations: performance concerns, role changes, hard feedback. It's grounded in the company's five decision-making best practices, and it lives wherever the manager already works. That skill stays internal to Airbnb.

So meet porchlight, a re-creation built for this page. Same pattern, stand-in content. A manager brings the conversation they're dreading, and the coach won't call them ready until every light on the prep map is on. Watch one prep play out:

cowork claude chat claude code

porchlight's five prep points are stand-ins for the internal decision practices

more from the lab · greenroom is one of several
the results

shipped in ninety days.
built for adoption.

performance-task modules
5/
kickoff to launch
90d
reported some or significant knowledge gain
100%
strategy clarity, before to after
+25pp
next field note

rolling out AI
at scale?

Send the shape of your problem. I'll send back where the adoption is going to break, and what Greenroom would do about it.

start a conversation