If you want a team to use AI well, you cannot leave the learning to chance. People pick up habits from whatever tool they happened to try first, the vocabulary fractures, and conversations about AI turn into people talking past each other. This play fixes that with a deliberate, tiered training program: a baseline everyone shares, plus a deeper track for the people building production systems.

When to use this play#

Stand this up once your organization has settled on an AI approach and an approved set of tools — it is the enablement half of the three-phase rollout. The goal is simple: everyone should leave with a similar understanding of what AI is, what it is not, and how to apply it responsibly for clients.

How to run it#

Use two tiers. A general AI tier that every employee completes, and an AI engineering tier that only engineers complete. Everyone gets the shared foundation; engineers go deeper into building with it.

Keep sessions small and focused. Roughly one-hour sessions, groups of three to four people, cameras on, and no doing other work during the session. Break each tier into a couple of sessions rather than one long marathon — attention and retention both fall off a cliff past an hour.

Fold it into onboarding. Existing staff complete the training in a first pass; after that, new hires go through it as part of their onboarding so the shared foundation never erodes as the team grows.

What each tier covers#

The general tier (everyone).

• A shallow but real grounding in AI and machine learning: the types of tools, the core terminology, who the major model providers are, and what an MCP server is.
• An introduction to your organization's AI approach, your internal gateway, the basics of effective prompting, and where to keep learning.

The engineering tier (engineers).

• AI-assisted development in practice: effective prompt engineering, how to configure AI tools safely, and a library of example engineering prompts.
• A deeper dive into the Model Context Protocol — its architecture and real use cases.
• A live demonstration of AI rapid prototyping.
• An overview of what it actually takes to build a production-grade, scalable AI solution, not just a demo.

Common traps#

• One giant session. Long sessions feel efficient and teach almost nothing; short and repeated wins.
• Skipping the baseline for non-engineers. When only engineers learn the vocabulary, the rest of the org can't participate in decisions about AI.
• Training once and never again. Without an onboarding hook, the shared foundation decays with every hire.
• Teaching tools instead of judgment. Tools change; the durable content is terminology, the responsible-use approach, and how to think about where AI fits.

Signals it's working#

People across the organization use the same words for the same concepts, engineers can configure and prompt tools without reinventing standards each time, new hires arrive already fluent in your approach, and conversations about AI move forward instead of restarting from definitions.