Price Ai Disclosure Policy

---
name: price-ai-disclosure-policy
description: Writes an AI Disclosure Policy with what you use AI for and what a person always does, client data rules by tool, when and how clients are told, the client's opt-out and the legal questions marked for an adviser. Use for "run price-ai-disclosure-policy", "should I tell clients I use AI", "AI disclosure policy", "AI use statement", "how do I disclose AI to clients", "what client data can go into AI tools", "client wants no AI on their work", "AI line for my website", part of the Pricing Under AI Pressure Pack by Polar Bear.
---

# AI Disclosure Policy

## When To Use
You keep asking yourself "should I tell clients I use AI, and am I losing work to people who don't?" Your website, proposals and contracts are silent on it, and every client conversation about AI starts from scratch. This skill answers: what exactly do I use AI for, what do I tell clients, when, and what happens if a client says no?

## When Not To Use
If you need the message that sells your work alongside AI, use AI Positioning Statement; this policy is the rule behind it. If a client has already sent AI clauses to sign, use AI Contract Questions.

## Inputs
- Your AI Activity Map, or a plain list of the activities where you use AI
- Your tool list: every AI tool and subscription you use for client work
- The kinds of client data you handle (by category, never real records)
- Any client request or clause about AI you have received
If you have none of this, I start from the tools you name and the services you sell, and mark the output as a first draft.

## Approach
The principle is transparency proportionate to significance, from the OECD AI Principles (Principle 1.3, transparency and explainability): people should know when AI shapes what they receive, in proportion to how much it matters. A light touch on a spell check, a clear note where AI drafted the analysis. The EU AI Act Article 50 transparency duties apply from 2 August 2026; whether they reach your work is a legal question, so the skill writes it as one. The failure it prevents: a client finds out from a stray artifact in a deliverable, and a quiet habit now reads as concealment.

## Workflow
1. Ask at most three questions: which services the policy covers, whether any client has asked about AI or asked for none, and where you sell (the countries matter for the adviser questions).
2. AI use by activity: for each service, the activities where AI does the work and a person checks, where AI assists a person, and where only a person works. Then what a person always does: review, decide, sign off. If a step has no human check today, say so plainly; the policy cannot promise one you do not do.
3. Client data rules by category (client confidential material, personal data, financial data, your own notes) against your actual tool list: allowed, allowed with the client's consent, never. Use each tool's settings and terms as you know them; whether a tool's data terms are adequate: check with a qualified adviser.
4. When clients are told, proportionate to significance: one website line, one proposal line, a pointer clause in the statement of work, and a note on deliverables where AI shaped the substance. Light use gets a light mention; never no mention.
5. Opt-out: what a client can ask for (no AI on their data, no AI on a deliverable, human only), and what it changes in time and price, in your figures or as [placeholders]. You decide whether you accept opt-outs at all.
6. Adviser questions: does Article 50 apply to this work; do your AI claims in marketing meet advertising rules where you sell; what data protection duties apply to client data in each tool. Each ends "check with a qualified adviser".
7. Draft it in Claude Docs (beta) if you want a document to keep and share, or in the chat in the same shape.

## Output Format
```markdown
# AI Disclosure Policy
**Covers:** [services] | **Owner:** [your name] | **Last reviewed:** [date]
## How AI is used
| Service | Activity | AI does, person checks / AI assists / Person only | Who signs off |
|---|---|---|---|
| [service] | [activity] | [mode] | [role] |
## Client data rules
| Data category | [Tool 1] | [Tool 2] | Note |
|---|---|---|---|
| [category] | [allowed / with consent / never] | [same] | [condition] |
## When clients are told
- Website: [one line] | Proposal: [one line] | SOW: [pointer clause] | Deliverables: [note, when used]
## Opt-out
| What a client can ask for | What it changes in time | What it changes in price |
|---|---|---|
| [request] | [your figure] | [your figure] |
## Questions for an adviser
1. [Question with the facts above]. Check with a qualified adviser.
## Decision
[Your name] approves the policy, decides the opt-out terms and sets the date the website and proposal lines go live by [date].
```

## Done When
- Every AI tool you use appears in the data rules, and every service in the use table
- Each "person checks" row names who signs off, and matches what you actually do
- Every legal point is a question ending "check with a qualified adviser"

## Quality Bar
- Data rules are written by category; no real client records or personal data appear as examples
- The lines to clients describe your real practice, never a stricter one you do not keep
- Opt-out costs are your figures or placeholders; Claude never suggests a surcharge
- Disclosure is plain and neutral: it neither apologises for AI nor sells against it
- It never hides your use of AI; legal questions go to a qualified adviser

## Next
Run price-statement-of-work (Statement of Work) to put the AI line in the contract scope.

## About the makers

This pack is made by Polar Bear, a consultancy built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).