Guides

The front door: Chat vs Cowork vs Claude Code, when to use each, what it costs, and which model to reach for. Start here.

The lowest-friction start: no install, just type. Profiles, Projects, real accountant use cases, and the data rule before you upload anything.

Claude working on your actual files, from a normal app window, no terminal. Point it at a folder, reconcile, organize, draft, with folder choice as your data decision.

The deep end, made gentle: install Claude Code, build a workspace, write your CLAUDE.md, and wire in your client-data guardrails from the first session.

Build a profile Claude reads on every chat plus one reusable skill, so you stop re-explaining yourself. Tiered from Claude Desktop to Claude Code.

Plain-English definitions for the AI terms that show up across the Lab.

When putting taxpayer data into AI triggers a written-consent requirement, and when an exception applies.

A short check for whether an automated categorization tool keeps you inside §7216.

When the bar is "nothing leaves this machine," a local model is the answer. How to run one.

What the Lab's skills are and how to run them in your own assistant.

The firm-governance set: AI use policy, vendor checklist, WISP addendum, and disclosure.

A worked vendor-vetting verdict on a community redaction tool.

The step-by-step sourcing workflow, public data only.

Ready-made ideal-taxpayer profiles to start from.

Where to look, and where not to, for public prospecting data.

Reliable answers without exposing the client, and every cite verified to primary source.

Research-backed Approve / Conditional / Reject verdicts tied to a specific use.

Start every project with the data posture and §7216 check wired in first.

An ITP-driven prospecting procedure that stays on public data.

A public-data-only approach to finding the clients you actually want.

A starting AUP you can adapt to your firm.

A consent-to-disclose template for AI-assisted tax work.

A documented research-memo workpaper structure.

Capture the facts cleanly before you research.

Disclose your firm's AI use in the engagement letter.

What to confirm before client data touches a vendor.

Add AI tooling to your written information security plan.

A full worked example, including the verification step catching a fabricated cite.

Drafting a client email with AI while you stay the author.

What IRS OPR's first AI guidance means for practice.

The honest answer to "is this safe?", and how to make it so.

The deep track: a 12-lesson self-study series on §7216, the Safeguards Rule, AICPA standards, Circular 230, and state law.

The map: how the rules fit together.

The core disclosure/use rule for return information.

When you can proceed without separate consent.

What a valid consent must say, and when you need one.

The security obligation and your written plan.

The professional-conduct confidentiality duty.

The tax-standards update, including reliance on tools.

Competence, diligence, and supervision duties.

State privacy and AI-governance statutes on top of federal rules.

The compliance shortcut, and its limits.

Synthesizing the stack into a workable posture.

The one-page summary to keep at hand.