Train — Workshops, Hackathons, Prompts
The AI COEE's most undervalued mandate. The thing that separates the orgs where AI actually changes behavior from the ones where it just changes the toolbar.
McKinsey's data is direct: nearly half of employees say their #1 request for AI is formal training. More than a fifth report receiving minimal to no support. The training gap is the adoption gap.
Kristina Tsys ran four foundational Copilot workshops with her team before they built a single agent. That was a deliberate slow-down. She understood that without baseline literacy, the agents wouldn't get used.
The three-layer training curriculum
The AI COEE designs and runs training in three layers, each with a distinct purpose. Treat them as a stack — you can't skip Layer 1 and expect Layer 3 to land.
Layer 1 — AI Literacy (for everyone)
What is AI, what is generative AI, what is an LLM, what are the limitations, what are the risks. Two hours, delivered live (not async video). Every employee — not just recruiting — completes this in their first 90 days.
The goal isn't technical depth. The goal is shared vocabulary and shared risk awareness. Jordann Savage's framing is right:
Layer 2 — Role-Specific Application (for recruiters & hiring managers)
Half-day, hands-on. Built around the actual tools the team will use and the actual workflows they own. For a recruiter, that means: JD drafting, candidate outreach, intake notes, sourcing briefs. For a hiring manager, that means: skills definition, structured interview design, AI-summarized candidate review.
Patrick Lindsley emphasized one principle that should anchor every Layer 2 session: prompt engineering is the hidden skill that changes everything. Teach how to ask, not just what tools exist. Teach how to assign a persona to the AI. Teach how to iterate — and teach the magic question Melissa Entzminger surfaced:
Layer 3 — AI Hackathons (for everyone willing)
Quarterly. Cross-functional. Problem-led. A hackathon is the AI COEE's most powerful capability-building tool — and the most underused. The format is simple:
- The org submits real workflow frustrations in advance
- The AI COEE picks 6–10 problems and forms cross-functional teams
- Teams have 1 day (or 2 half-days) to build a working AI solution
- Each team demos to a panel — including the CEO if you can get them
- The best solutions get AI COEE budget and a path to production
The hackathon model is what Kristina Tsys' team built into their ongoing cadence — bring your roadblocks, brainstorm together, ship something usable in a single day. It builds skills, it surfaces problems the AI COEE didn't know about, and it produces the agents the org actually needs.
Why hackathons work where mandatory training doesn't
In enterprise environments, the people who learned conversational AI fastest were the ones who built something. Watching a video doesn't change behavior. Building a working agent — even a janky one — does. The hackathon converts curious skeptics into champions in a single afternoon. Make it a quarterly fixture.
The Manager Activation track
Within Phase 6, the AI COEE runs a parallel track focused specifically on managers. Gallup's data is unambiguous: employees with managers who actively support AI are 98.7x more likely to say AI has transformed their work. The manager is the lever.
The Manager Activation Program has four components:
- AI tool certification. Every hiring manager and recruiting team lead completes structured proficiency before their teams are expected to use AI tools.
- Role-specific playbooks. Not generic AI training — specific guidance on how AI changes the workflow for that manager's recruiting context (high-volume, executive, technical, hospitality, etc.).
- Team-level adoption check-ins. Managers hold regular, structured conversations with their teams about AI usage, questions, and concerns. Not surveillance — support.
- The AI champions network. The most engaged managers across the org get connected for peer learning. The AI COEE facilitates monthly champion calls.
What to put on the training calendar
A realistic Year 1 cadence for a 200–500 person TA org:
- Monthly — open office hours with the AI COEE (45 minutes, drop-in, anyone can ask anything)
- Every 6 weeks — Layer 1 literacy session for new hires
- Quarterly — AI hackathon
- Quarterly — role-specific Layer 2 deep dive (rotating: sourcing, screening, candidate experience, hiring manager partnership)
- Annually — Manager Activation Program refresh and recertification
Don't outsource the workshops — at least not the first round
Kristina Tsys partnered directly with Microsoft for her foundational sessions, but she co-led them. That's the right model. External experts bring depth and credibility. Internal facilitators bring context. The combination is what makes the training land. Vendor-only training tends to be generic. Vendor-plus-AI COEE training tends to stick.
Training is the AI COEE's most renewable mandate — and the one that has the largest compounding effect on every other phase. When training works, governance gets easier (people don't bypass tools), scaling gets faster (champions emerge organically), and the next round of pilots gets sharper (the org is asking better questions).
Speaking of governance — that's where Phase 7 takes us.