HR & hiringstrategy

Hackathon-wave AI rollout for non-engineers — pre-train + synchronous + repeat

Problem solved: Seminars and lectures don't move non-engineer AI adoption; multi-wave hackathons with structured prep do — but you need 2-3 waves to capture second-tier adopters, not one.

TierBMedium — Solid coverage with concrete examples
EffortM · Weeks
Forfounder · HR lead · head of enablement

What

Roll out Claude Code (or a Cowork-style hosted variant) to non-engineering staff via hackathon waves: identify 5-10 real daily pain points per role, pre-train staff ~5-10 hours on the basics, then run a 4-9 hour synchronous hackathon where each person solves their own real task with Claude assistance. Repeat 2-3 times. Multi-operator corroborated; one practitioner has run ~7 waves for clients.

Why it matters

Pure lectures don't move adoption. "Infection only happens via practice." The wave model is what catches second-tier adopters: first session converts the eager 20%; sessions 2-3 catch the cautious majority who needed to see colleagues succeed first. Pair with corporate MCPs so non-tech staff have safe, pre-wired wrappers when they sit down to work.

End-to-end

  1. Identify 5-10 real daily-rotation pain points per role. Lawyers, accountants, CS, sales — surface the tasks they grind on now. Generic exercises don't stick.
  2. Pre-train ~5-10 hours on basics. Async video or live; covers what Claude is, how to prompt, how to use Cowork or Claude Code's safe surfaces.
  3. Run the synchronous hackathon: 4-9 hours. Each person works on their own real task with Claude. Facilitator floats and unblocks. No demo project — real work only.
  4. Pair with pre-configured corporate MCPs. Non-tech staff need wrappers around CRM, analytics, ticketing — not raw API keys. See for the 8-MCP rollout pattern.
  5. Repeat 2-3 times. First wave converts eager adopters; waves 2-3 catch the cautious majority.

Gotchas

  • One wave is not enough. Operators who ran a single hackathon got eager-adopter conversion only and concluded the format doesn't work. The second and third waves are where the majority moves. Budget for all three before starting wave one.

Tools

  • Claude Cowork (or equivalent hosted Claude surface) accounts for all staff
  • Pre-configured corporate MCPs for the data sources staff actually need
  • A facilitator who's run at least one prior wave