At work, AI is leverage — a fast junior collaborator. It drafts, outlines, and pressures your thinking. It does not own your standards, your reputation, or your name on the file.

Operating rules

You ship, it assists. Anything client-facing, legal, medical, or financial gets your review.
Prompt for the artifact you need — bullets, options, a critique — not “make it good.”
Protect confidential data. Don’t paste secrets, private HR notes, or restricted documents into tools you don’t control.

Builder move: use AI to expand options and catch blind spots. Keep decisions, voice, and verification.

Your turn — pick the builder move

For each workplace situation, choose the sharper approach.

0 of 5 completed

Further

Prompt Engineering GuideStructured patterns for clearer asks at work.
Google AI EssentialsShort, practical course if you want a fuller path.

Trail stop complete

You practiced using AI as leverage without handing over the wheel.

Next: Trail 05 — Hallucinations & citations

Stay in the loop:

← Back to Trail Builders home

Trail stop 05

Hallucinations & citations

Models can produce fluent answers that never happened — including fake quotes, studies, and URLs. That’s not “lying with intent.” It’s pattern completion without a guaranteed fact-check.

What to remember

Confidence is a style. It is not evidence.
Invented sources are a hard fail — never ask a model to fabricate papers, quotes, or citations.
Verification is part of the workflow — especially before you publish, file, or advise.

Better ask: “Help me find what to verify and how” beats “Give me three papers that prove X” when X may not be real.

Your turn

Sort the claims, then pick the safer prompt when it matters.

0 of 6 completed

Further

Factuality risks (Prompting Guide)Why models invent details — and how to reduce the risk.
Hallucination (AI) overviewClean reference definition if you want the term locked in.

Trail stop complete

You practiced treating fluency as a draft, not a document of record.

Next: Trail 06 — Tools & workflows

Stay in the loop:

← Back to Trail Builders home

Trail stop 06

Tools & workflows

Chat is one surface. Builders match the job to the channel: draft in chat, verify with search, calculate in a sheet, escalate to a human when stakes are high.

Four lanes

Chat — draft, outline, rephrase, brainstorm options.
Search — check facts, find primary sources, confirm what exists.
Specialist tool — spreadsheet, design app, IDE, analytics — when precision matters.
Human — judgment, relationships, regulated advice, final accountability.

Default: draft with chat → verify with search or primary data → finish in the right tool → human signs off.

Your turn — place each task

Drag a task into a lane — or tap a task, then tap the lane. You’ll get immediate feedback.

0 of 8 placed

Chat

Search

Specialist tool

Human

Further — model examples

Frontier (hosted)

GrokxAI — reason, rewrite, pressure-test.
ChatGPTOpenAI — everyday drafting.
ClaudeAnthropic — long docs, careful edits.
GeminiGoogle — same prompt, different take.

Open source / open-weight

LlamaMeta — open weights, widely used.
MistralStrong open models, lighter footprint.
DeepSeekOpen weights, strong on reasoning.
Hugging FaceHub to browse and run open models.

Trail stop complete

You practiced matching work to the right surface — the quiet skill behind speed.

Next: Continue to Advanced — Scout →

Stay in the loop:

← Back to Trail Builders home