By Rakibul Hasan, Moonlab
If you use AI or automation tools with personal data about people in the UK or EU, GDPR or UK GDPR applies to you. The practical checklist is short: have a lawful basis, sign a data processing agreement with each vendor, collect only what you need, know where data goes, and delete it on a schedule.
This guide is a general checklist, not legal advice. Rules differ by country and situation, and regulators publish guidance that changes. Check with a qualified professional and the official regulator pages, such as the ICO in the UK, before relying on anything here.
Why GDPR matters for automation and AI
An automation that moves customer details between tools, or an AI step that reads an email, is processing personal data. The law cares about what you do with that data, not how clever the tool is. Your business is usually the controller, and the tool vendor is a processor acting on your instructions.
Small businesses are not exempt. The good news is that a few habits cover most of the risk. Treat the points below as a checklist to work through before each new tool or workflow goes live.
1. Know what personal data you use
List what the workflow touches: names, emails, phone numbers, addresses, order history, messages. Mark anything sensitive, such as health, biometric or financial details, since these get stricter treatment. Note where the data comes from and where it goes.
A simple table with columns for data, source, tool, purpose and retention is enough. It also helps you answer customer questions later.
2. Have a lawful basis and be transparent
You need a lawful basis for each purpose, such as contract, legitimate interests or consent. Which fits depends on the case, and marketing has specific rules, so get advice for anything unclear.
Tell people what you do. Your privacy notice should mention that you use service providers and, where relevant, automated tools or AI. If a decision is made solely by automated means and has legal or similarly significant effects on a person, extra rules apply, so keep a human in those decisions.
3. Sign data processing agreements
Each vendor that handles personal data for you should have a data processing agreement (DPA). Most major automation and AI vendors publish a standard DPA you can accept or sign. Find it in their legal or trust pages, and keep a copy.
- Check whether the vendor uses your data to train its models, and whether you can opt out.
- Ask who its sub-processors are.
- Check how a data breach is reported to you.
- Business and API plans often have different data terms from free consumer plans, so read the current terms for the plan you use.
4. Minimise what you send
Send only the data needed for the task. If an AI step is classifying emails, it may not need the sender's full address history. Remove or mask names, numbers and identifiers where you can, and never send card numbers, passwords or special category data to a tool that is not designed for it.
Minimisation lowers your risk and your exposure if something goes wrong. It also reduces what you have to disclose and delete later.
5. Check where data is processed
Ask each vendor where data is stored and processed. If personal data moves from the UK or EU to other countries, such as the US, a legal transfer mechanism must be in place. Many large vendors rely on approved routes, and some offer regional hosting options. Confirm what applies to your plan, and ask a professional if it is unclear.
Self-hosting a tool on a server in your region can simplify this for the automation layer, but any AI provider you call still has its own processing location. See moonlab.space/guides/self-hosted-n8n-guide.
6. Set retention, access and security
Decide how long each type of data is kept and delete on schedule. Automation logs often store full message content, so set them to purge or avoid logging sensitive fields.
- Use strong, unique passwords and two-factor authentication on every tool.
- Give each person and each connection the minimum access it needs.
- Remove access when staff leave.
- Keep a record of your tools, DPAs and workflows.
- Have a plan to respond to access and deletion requests, including data held in automation tools.
A worked example: checking a new AI step before launch
Say a small estate agency in Manchester wants an AI step that reads incoming viewing requests and drafts replies. The owner walks through the checklist before launch. Data used: name, email, phone number and the property they asked about. Nothing sensitive. Lawful basis: the person asked about a property, so replying is part of dealing with their request. Privacy notice: updated to say that service providers and automated tools help to process enquiries.
Vendors: the automation platform and the AI provider each have a published DPA, which the owner accepts and saves. Data location: both vendors state where they process data and the owner records it, asking an adviser about transfers because one is outside the UK. Minimisation: the AI step receives the message text and property name only, not the full contact record. Retention: automation logs are set to delete after a short period, and the CRM follows the agency's existing retention policy. Finally, the owner writes a one page record listing the tools, purpose, data and who is responsible.
- Mistake: assuming a free consumer account has the same data terms as a business plan.
- Mistake: leaving logs on forever. Automation tools often store full message bodies.
- Mistake: forgetting deletion requests. Know every place a person's data lives, including automation history.
- Mistake: sending full records to an AI step when two fields would do.
- Mistake: no process for staff leaving. Remove their access and rotate shared keys.
- Mistake: relying on this checklist alone. For anything high risk, take professional advice.
7. Consider a risk assessment and get advice
For higher-risk uses, such as large-scale profiling, sensitive data or automated decisions about people, a data protection impact assessment may be required. A short assessment is good practice for any new AI use.
Marketing messages have their own rules, for example PECR in the UK. See moonlab.space/guides/cold-email-deliverability for the basics. We are not lawyers, and this guide is not legal advice. If you would like help mapping your workflows and tools before you talk to an adviser, the free audit at moonlab.space/contact is a place to start.
