such wow doge
AI WOW
Published on

How to support AI adoption on your team

Authors

Epistemic status: early notes, written quickly. This post will get a major update in November.

I think it's time for teams to invest aggresively, and with urgency.

Some general suggestions below...

1. Assign explicit responsibilities

2. Establish a pattern of continuous learning for all staff

  • People often say they have "no time" to experiment with AI. This is one of the strongest headwinds to adoption.
  • So: schedule weekly or fortnightly time blocks for AI experiments and learning.
  • Make this a mandatory calendar event, like a team meeting. Video call or in-person.
  • Example format:
    • 1-2 mins: Introduction
    • 60 mins: Experiments and learning (individuals or small groups)
    • 15 mins: Show-and-tell (whole team)
    • 10 mins: Consolidation and takeaways (individuals)
    • Facilitation is easy. Your main job is to create space.
  • I'll probably write a more detailed playbook in November. Email me if you'd like a copy.

3. Make a "#using-ai" Slack channel

  • Make a Slack channel for sharing learnings, tips, links, products, questions, etc.
  • Frame it so the bar for posting is low.
    • Think: "things that worked for you (or didn't)", "share your experiments", "good blog posts", "products that might be of interest".
    • Do not frame it as "recommendations for best practice" (but welcome effort-posts if people share them).
  • Use a separate channel to discuss formal initiatives e.g. a project to automate a workflow.

4. Encourage internal champions

  • Identify a couple of "AI adoption champions" in your team.
  • Add this to their role description.
  • Give them a weekly time budget for experimentation and informally supporting others.
  • Reward them.
  • Further reading: OpenAI Academy; GitHub Resources.

5. Make it easy

  • All staff should have access to ChatGPT, Claude and Gemini (or your internal equivalents).
  • Create affordances for common use cases.
  • Everyone should know who to contact if they have questions.
  • Make it easy to get API keys for vibe-coding.
  • Make your "what you're allowed to do/use" policies clear.
  • Evaluate requests to whitelist new tools within 2 days.

Appendix 1. Further reading

Appendix 2. Ambitious internal tooling

Some teams invest heavily in internal tooling. For example, Claude reports:

Stripe built an internal web application called LLM Explorer with a ChatGPT-like interface, rolled it out company-wide with proper security controls, and watched over 33% of employees adopt it within days. The tool now powers over 60 internal LLM applications across the company, with hundreds of reusable interaction patterns documented and shared.

The most popular feature is the prompt sharing and discovery system. Employees create and share prompts like the "Stripe Style Guide," which transforms any text to match company tone for emails, website copy, and presentations. This single prompt is used by account executives, marketers, and speakers across functions. The backend API service supports over a dozen models with automatic model selection based on context size, logging for auditing, and auto back-offs for rate limits.