learn: write Train your AI Employee and Build a Custom AI Employee steps

[Shivapriya92]

What this is

Adds two lessons to the "Hire your first AI Employee" learning path at /learn/ai-workforce/: Train your AI Employee (step 4) and Build a Custom AI Employee (step 5, lab). Closes out the 360Learning-course-to-Learn-tab work tracked in ET-690.

Files

docusaurus/training/ai-workforce/train-your-employee.mdx — was a stub, now written
docusaurus/training/ai-workforce/custom-employee-lab.mdx — was a stub, now written
learn-revamp/drafts/OUTLINE-ai-workforce-train-and-custom-lab.md — the approved outline and decision record behind both steps, kept for review history

What each step covers

Train your AI Employee: re-running the ten-question audit on a live employee, fixing a knowledge source instead of guessing, adding a guardrail capability so pricing answers never get invented, reading Explanation to see why an answer happened, and the ongoing coach-and-correct habit.
Build a Custom AI Employee (lab): starting from an outcome statement, creating a blank employee (profile, autonomy level, role prompt), connecting knowledge and built-in capabilities, adding one custom capability with a simple tool, writing the capability's instructions, and a pass/fail test before calling it done.

Source material: the two 360Learning courses for ET-690 (AI Receptionist Setup & Configuration, Meet Your First AI Employee), cross-checked against businessapp-docs and this repo's own /docs/ai/ tree, plus existing partner-call evidence already in learn-revamp/partner-call-insights.md.

Decisions baked in (see the outline doc for full reasoning)

Both steps use the full lab pattern from PROPOSAL-lab-pattern.md (Lab callouts, LabChecklist, "What you now have").
Step 4 is tagged Beginner, step 5 Intermediate.
The AI Employee autonomy level (Level 3: Autonomous default) is included and confirmed directly in-product — it isn't documented anywhere in businessapp-docs yet, which is worth flagging to that docs team separately.
Step 5 stops at one simple tool and defers anything with chained tools/webhooks to the Builder track, to stay out of Cal's territory.
A "reset to default" safety net is described by behavior only, without naming a specific button, since that UI wasn't independently verified.

Not in this PR: steps 3 (Teach it to book), 6 (Autopilot), and 7 (Sell and manage) are still stubs — they need their own outlines and aren't sourced from the uploaded courses.

Testing: checked both files by hand against the working reference implementation (run-your-first-snapshot.mdx) and confirmed all JSX/markdown syntax is balanced. Could not run a full npm run build in the sandbox this was drafted in — recommend running that locally before merge.

[CalCooper]

Really nice work, Shiva. I went through both lessons against the /docs/ai tree and the navigation holds up all the way through — the create path, the Professional-plan requirement, Custom Capabilities, the Purpose field, all of it. Links resolve, the lab pattern is clean, and the knowledge checks are sharp. Approving.

Three things to fold in, none blocking the merge:

1. The autonomy claim is the one to tighten. "Level 3, Autonomous, is the default… books on its own" isn't in the docs anywhere, so it rests entirely on the in-product check. It's also the highest-stakes line in the lesson — autonomy is how much the AI runs without review — so it's the one place I don't want us leaning on a single unverified source.
2. Two asks: soften the wording so it survives a UI change (something like "Autonomous is the highest level and the usual starting default"), and drop a first-person verification trace on it (your own note/screenshot, per the tagging tweak on #782). And let's file the docs gap to the businessapp-docs team so it graduates to a real citation.
3. Partner Center vs Business App drift — the lessons say Partner Center throughout, but the linked Creating Custom Capabilities guide says "Business App dashboard." Not yours to fix, but worth flagging to whoever owns those docs so a learner clicking through doesn't hit mismatched framing.
4. Custom-lab Required line promises "use the example in this step," but the step only shows CheckOrderStatus as an illustration, not a runnable API. Either add a real sample endpoint or soften the wording.

One more thing — please run npm run build locally before merge (you flagged it wasn't run; I confirmed the links resolve by hand, but the build is the real check). But thi smight be because the content is still hidden?

And when it's live: send me the link! I want to walk it the way a partner would — see how it looks and how the flow feels end to end, not just how it reads in the diff. That's the real test.

Nice work Shiva!!!!!