Why This Policy Exists
I use AI tools daily to manage Google Ads accounts, analyse data, and streamline my work. This policy sets out how I use AI responsibly, how I protect client data, and what safeguards are in place.
If you're a client and have questions about any of this, just ask.
How I Use AI
I use AI to support my work, not replace my judgement. Specifically:
- Campaign analysis – Identifying trends, spotting anomalies, reviewing search terms
- Reporting – Generating quarterly reports and performance summaries
- Competitive intelligence – Analysing competitor messaging and positioning to find opportunities for your business
- Landing page audits – Reviewing page content, structure, and conversion elements
- Strategic decision-making – Using AI to stress-test ideas and evaluate campaign strategies from multiple expert perspectives before making recommendations
- Research – Staying current with Google Ads changes and industry best practices
- Communication – Drafting emails and documents (always reviewed before sending)
- Automation – Reducing repetitive manual tasks so I can focus on strategy
Every AI-generated output is reviewed by me before it reaches a client or goes live in an account.
Approved Tools
I only use AI tools I've reviewed and trust:
- Claude (Anthropic) – My main AI assistant, used through Claude Code on Anthropic's Team plan. This is a business plan: it comes with a data processing agreement, and my content is not used to train Anthropic's models.
- 8020Brain system – Personal knowledge management and workflow automation
- Google Ads built-in AI features – Smart bidding, responsive search ads, Performance Max (where appropriate)
- Google Analytics 4 – AI-powered insights within GA4
- Gemini (Google Workspace) – A note-taker in Google Meet that writes up client calls
- Wispr Flow – Voice dictation, so what I say is typed for me, and sometimes call notes
Claude is the only AI tool I use to do the work on client accounts. Gemini and Wispr Flow are used for call notes and dictation, and those notes can then be used with Claude, on the Team plan described above. Other AI tools may be used for general research or images, never with client information.
I don't use free, unvetted AI tools for any client work. If I adopt a new tool, I review its data handling and security first.
Client Data Protection
What I do use with AI tools
- Your business name and contact details (so I can manage your account efficiently)
- Campaign performance data (impressions, clicks, conversions, cost)
- Search term reports
- Ad copy and landing page content (this is already public-facing)
- Account structure information (campaign names, ad group names)
This information is stored in my own private system that only I can access. It helps me work faster and more accurately on your account.
What I never put into AI tools
- Passwords or login credentials
- Bank details or payment information
- Confidential business information you've shared in confidence (supplier contracts, internal strategies)
Your customers' personal details are only used when a task needs them, for example checking lead quality in a lead-tracking export such as WhatConverts. These exports can include names, phone numbers and email addresses.
The simple test: if I wouldn't be comfortable with it appearing on a public noticeboard, it doesn't go into an AI tool.
Client Brains
If I build a Client Brain for your business, it is built for you and handed over to you. Until then, I may own or belong to the organisation it is stored in.
- It is stored in a private GitHub organisation set up for your business.
- Your team use it with their own Claude accounts, so your contract for Claude is directly with Anthropic.
- When I work inside your brain, I use my own Claude Team account, described above.
- Once it has been handed over to you, you decide who has access to it.
- Your organisation is not automatically covered by GitHub's data processing agreement. That comes with GitHub's paid plans, such as Team, which is the plan my own brain uses. A client's organisation can be on GitHub's free plan, which has no data processing agreement. Upgrading is your choice.
Human Oversight
AI supports my decisions – it doesn't make them. I always:
- Review all AI outputs before sharing with clients or making account changes
- Verify data and calculations against source platforms (Google Ads, GA4, WhatConverts)
- Make the final call on budget changes, bid strategies, and campaign structure
- Write client communications in my own voice (AI may help draft, but I review and edit everything)
No automated action is taken on a client's Google Ads account without my review.
Accuracy and Honesty
- I don't present AI-generated content as verified fact without checking it
- If I'm uncertain about something, I say so
- If data is incomplete or AI couldn't verify something, I flag it clearly
- I never fabricate data, screenshots, or results
Data Storage and Security
- Client data is stored in private, password-protected repositories on GitHub. My own brain, where client work lives, sits in a GitHub organisation on the paid Team plan, which is covered by GitHub's data processing agreement (from 28 October 2026).
- Each client has a separate folder – data is never mixed between clients
- AI conversation history is not shared between clients
- I use two-factor authentication on all platforms where available
Keeping This Policy Current
I review this policy at least every six months, or sooner if:
- I adopt a new AI tool
- Google or Anthropic change their data handling policies
- A client raises a concern
- Industry best practices change
- The ICO (the UK data protection regulator) publishes new guidance on AI
Questions? Contact sandi@sandbar-solutions.co.uk
Last updated: November 2026