In short
- When an AI answers your enquiries, you are the controller and the AI vendor is your processor. The responsibility stays with you, and Article 28 GDPR requires a written contract with the vendor.
- Replying to an enquiry, asking qualifying questions and booking a visit usually rest on steps taken at the person's request before a contract (Article 6(1)(b)). Following up on silence or using conversations for anything else needs its own justification.
- Tell people at the first message: in the EU, the AI Act requires them to be informed that they are dealing with an AI from 2 August 2026, and the GDPR requires the privacy information at the moment you collect their data.
- Keep rejections and sensitive topics with a person. Article 22 GDPR, and in the UK the reformed rules in force since 5 February 2026, protect people from significant decisions made by software alone.
AI and GDPR work together fine for a small business, as long as you can answer four questions before an AI starts replying to your customers. Who is responsible for the data? Which lawful basis do you rely on? What do you tell people in the first message? And what has your AI vendor signed? Get those four right, and an AI that answers enquiries, asks a few questions and books a visit is ordinary data processing, not a legal gray zone.
This guide covers the EU GDPR and AI Act, with a section on the UK. It is written for service businesses that get enquiries from ads, forms, chat and social messages, from trades and clinics to real estate, schools and car dealers. It is practical guidance, not legal advice.
Who is responsible when an AI answers your customers?
The European Data Protection Board's guide for small businesses says it plainly: "A data controller determines the purposes and means of processing personal data." A processor "acts under the instructions of the controller only, by processing personal data on behalf of the controller."
When an AI replies to your enquiries, you are the controller: you decided an AI should answer them and what it should ask. The company that supplies the tool is normally your processor. If it generates replies with another provider's language model, that provider is a sub-processor. Three things follow:
- The responsibility stays with you. "The vendor takes care of GDPR" does not hold up with a regulator. The vendor handles its part. You answer for the whole.
- You need a contract. Article 28(3) of the GDPR says processing by a processor "shall be governed by a contract or other legal act." That contract is the data processing agreement, or DPA.
- Sub-processors need your approval. Under Article 28(2), a processor "shall not engage another processor without prior specific or general written authorisation of the controller." Ask your vendor for its sub-processor list and how it tells you about changes.
Whether the tool is a scripted chatbot or a real agent that decides its own next step (see what an AI agent is), the GDPR treats it the same way.
Which lawful basis covers AI replies to leads?
Every use of personal data needs a lawful basis from Article 6(1) GDPR. For an AI that handles inbound enquiries, three matter.
Steps before a contract, Article 6(1)(b). Processing is lawful when it is necessary "in order to take steps at the request of the data subject prior to entering into a contract." Someone who fills in your form or messages your page is asking you to do something: check availability, give a price range, book a visit. Replying, asking what you need to know and booking are steps at their request. This covers most of the work.
Legitimate interests, Article 6(1)(f). You can rely on your own interest when it is not "overridden by the interests or fundamental rights and freedoms of the data subject." The EDPB's Opinion 28/2024 on AI models describes a three-step test: identify the legitimate interest, check that the processing is necessary for it, and check that the person's rights do not override it. The opinion names "developing the service of a conversational agent to assist users" as an example of an interest that can be legitimate. If you use this basis, Article 21(4) requires you to point out the right to object "at the latest at the time of the first communication," clearly and separately from other information.
Consent, Article 6(1)(a). Consent covers "one or more specific purposes." You rarely need it to answer an enquiry, but an enquiry is not a blank check: your newsletter is a different purpose with its own rules.
How that usually maps onto what an enquiry AI does:
| What the AI does | Basis that usually fits | What to watch |
|---|---|---|
| Replies to the enquiry the person sent | 6(1)(b), steps at their request | Use the data only to handle that enquiry |
| Asks your qualifying questions | 6(1)(b) | Ask only what you need to help them |
| Books a call or a visit | 6(1)(b) | Confirm what was booked and with whom |
| Follows up when the person goes quiet | 6(1)(b) for a short, related follow-up, otherwise 6(1)(f) | Stop when they say stop, and mention the right to object if you rely on 6(1)(f) |
| Lets the vendor use conversations to improve its AI | 6(1)(f) after the three-step test, or not at all | This is a new purpose, so it must be in your notice |
| Adds the person to marketing lists | Not covered by the enquiry | Separate e-marketing rules apply |
GDPR and lead data: collect less, keep it shorter
Two principles in Article 5 matter most here. Data minimization means personal data must be "adequate, relevant and limited to what is necessary." Storage limitation means keeping it "for no longer than is necessary."
For lead data, write down two decisions. First, the questions the AI may ask: if you do not need a date of birth to book a roof inspection, the AI should not ask for one. Second, a retention period for enquiries that never became a job. For example, you might delete leads that have been silent for 12 months. The number is yours to justify, but you need one.
What to tell customers: a GDPR compliant AI chatbot notice
Article 13 GDPR lists what people must be told "at the time when personal data are obtained." In short:
- who you are and how to contact you
- the purposes and the legal basis, including your legitimate interests if you rely on them
- the recipients or categories of recipients, such as your AI vendor
- any transfer outside the EU, and whether an adequacy decision covers it
- how long you keep the data, or how you decide
- their rights, including withdrawing consent and complaining to a supervisory authority
- the existence of automated decision-making covered by Article 22, with "meaningful information about the logic involved"
Article 12(1) adds that it must be "concise, transparent, intelligible and easily accessible," in "clear and plain language." So use two layers: one short line in the first message, and a link to the full notice.
If a lead reaches you through someone else, such as a partner that passes enquiries on, Article 14 applies instead: inform the person within a reasonable period and "at the latest within one month."
Template: the paragraph to add to your privacy notice
How we answer enquiries. When you contact us through our website, ads or social media, an AI assistant may reply on our behalf. It uses your details to answer your questions, ask what we need to know and book a call or visit. Our team sees every conversation, and you can ask for a person at any time. The assistant is provided by [vendor], which processes your data on our behalf under a data processing agreement and uses [AI model provider] to write replies. Your data is stored in [region]. Legal basis: steps you asked us to take before a contract (GDPR Article 6(1)(b)), and our legitimate interest in following up on enquiries that went unanswered (Article 6(1)(f)). You can object to follow-ups at any time by replying STOP or writing to [email]. We delete enquiries that did not lead to a job after [period].
Fill in the brackets with facts from your DPA, not from the vendor's marketing page.
The AI Act and GDPR: say it is an AI in the first message
The GDPR covers what happens to people's data. The EU AI Act adds a separate duty: telling them they are dealing with a machine. That is the AI Act and GDPR interplay in practice, and one opening line serves both.
Article 50(1) of the AI Act requires AI systems "intended to interact directly with natural persons" to be designed so "the natural persons concerned are informed that they are interacting with an AI system," unless that is obvious to a reasonably well-informed person. Article 50(5) says the information must be given "in a clear and distinguishable manner at the latest at the time of the first interaction." It applies from 2 August 2026.
Formally the duty lands on the provider, your vendor. But you choose the greeting and the channels, so check it yourself. Do not lean on the "obvious" exception: a text from a business number looks exactly like one typed by a person.
| Channel | Where the disclosure goes | Example wording |
|---|---|---|
| Website chat | The greeting | "Hi, I'm the AI assistant for Brightside Dental. Ask me anything, or type 'person' to reach the team." |
| SMS | The first text | "Hi Sam, this is the AI assistant at Miller Roofing, replying to your request. How we use your details: [short link]" |
| First line and signature | "I'm the AI assistant at Oak Street Realty. A colleague reads this thread and can step in at any time." | |
| Facebook or Instagram messages | The first reply | "Thanks for your message! I'm the AI assistant for Harbor Travel. Want me to check dates for you?" |
The businesses are made up. Keep the line short and add the privacy link where the channel allows.
Generative AI and GDPR: the processor agreement and the training question
With generative AI, the question owners forget is what happens to the conversations afterwards. Start with the DPA: Article 28(3) lists what it must contain, which makes it a checklist for any AI GDPR tool:
- Instructions: it processes data only on your documented instructions.
- Confidentiality: its staff are bound to confidentiality.
- Security: it applies the measures of Article 32.
- Sub-processors: it follows the rules for engaging others, such as the model provider.
- Rights requests: it helps you answer requests for access, correction or deletion.
- Breaches and assessments: it helps you meet Articles 32 to 36.
- End of contract: it deletes or returns the data, at your choice.
- Proof: it gives you what you need to show compliance and allows audits.
Then ask three questions the contract may not answer plainly:
- Are our conversations used to train your models, or your model provider's? A processor acts on your instructions only. Using your customers' messages for its own product is a purpose of its own. Switch it off, or disclose and justify it.
- Where is the data stored and processed? If it leaves the EU, your notice must say so, including whether an adequacy decision applies (Article 13(1)(f)).
- Can we set our own retention period? If the tool keeps everything forever, you cannot keep the promise in your notice.
When the AI decides: automated decisions and sensitive data
Article 22(1) GDPR gives people the right "not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her."
Most enquiry handling stays well below that line: offering three time slots changes nobody's legal position. It gets closer when the AI turns people away from something that matters: an insurance quote, car financing, a place on a course. Where such decisions rest on a contract or explicit consent, Article 22(3) requires at least "the right to obtain human intervention," and the right "to express his or her point of view and to contest the decision."
The simplest design: let the AI qualify and route, and keep every "no" with a person. Check that the handover to a human is easy to reach. Konverto, for example, hands the conversation to a person when the customer asks for one or a call-back, or raises a complaint.
Sensitive data needs extra care. Article 9(1) GDPR prohibits processing "data concerning health" and other special categories unless an exception in Article 9(2) applies. For clinics and salons, the risk is an AI that asks "what are your symptoms?" and stores the answer in a message thread. Let it collect only what it needs to book, and leave medical details for the consultation.
AI and GDPR in the UK after the Data (Use and Access) Act
UK businesses work under the UK GDPR, which the Data (Use and Access) Act 2025 amended. For an enquiry AI, the change that matters most concerns automated decisions.
According to GOV.UK's summary of the changes, the Act creates "a more permissive framework" for decisions based solely on automated processing. In return, you must give people information about significant decisions made about them, let them "make representations about and to challenge such decisions," and let them "obtain human intervention." The provision behind it, section 80, came into force on 5 February 2026.
Two more notes:
- The Act adds a lawful ground called recognised legitimate interests, with examples such as crime prevention, safeguarding and emergencies. It is not a shortcut for following up sales enquiries.
- The AI Act is an EU regulation. If you also take enquiries from the EU, check whether it reaches you. Either way, "I'm an AI assistant" in the first message costs nothing.
AI and GDPR compliance checklist for a small business
Run through this before you switch on an AI for enquiries, and again once a year:
- Map the flow. Where enquiries come from (ads, forms, chat, social messages) and where the data ends up (AI tool, inbox, CRM).
- Sign the DPA. Check it against the eight points above and save the sub-processor list.
- Pick a lawful basis for each step. Use the table above and write it down.
- Write the AI's question list. Only what you need to help and book.
- Set the disclosure in every channel. The first message says it is an AI assistant and links to your notice.
- Update your privacy notice. Use the template, with facts from your DPA.
- Set the retention period in the tool, not just in the policy.
- Keep the "no" with a person. Rejections, complaints and sensitive cases go to a human.
- Keep a record. The Article 30(5) exemption for organizations with fewer than 250 employees does not apply when processing "is not occasional," and answering enquiries daily is hard to call occasional. One page covering the flow, the bases and the vendors will do.
If you use a tool such as Konverto to answer enquiries from lead ads, website forms and website chat, run it through the same list as any other vendor before the first lead arrives.
Frequently asked questions
Is it legal to use an AI chatbot under GDPR?
Yes. GDPR does not ban AI. It sets conditions: a lawful basis for each use of personal data, clear information to people when you collect it, a processor contract with the AI vendor, only the data you need, and safeguards if software alone makes significant decisions about people.
Do I need consent to let an AI reply to customer enquiries?
Usually not. When someone sends you an enquiry and the AI answers it, asks what you need to know and books a visit, those are steps taken at the person's request before a contract, which Article 6(1)(b) GDPR covers. Consent becomes relevant for other purposes, such as later marketing, which follow their own rules.
Do I have to tell customers they are talking to an AI?
In the EU, yes. Article 50 of the AI Act, which applies from 2 August 2026, requires AI systems that interact directly with people to be designed so those people are informed they are dealing with an AI, unless it is obvious, at the latest at the first interaction. Outside the EU it is still the honest default.
What should a data processing agreement with an AI vendor include?
Article 28(3) GDPR lists the minimum: processing only on your documented instructions, confidentiality for staff, security measures, rules for sub-processors, help with data subject requests and with security and breach duties, deletion or return of the data at the end, and the information you need to show compliance.
Is AI lead qualification an automated decision under GDPR?
Only if the decision is based solely on automated processing and produces legal effects or similarly significantly affects the person (Article 22). Asking questions and booking a visit rarely reaches that bar. Turning people away automatically from something important, such as a loan, an insurance policy or a place on a course, can.
How is AI and GDPR different in the UK?
The UK GDPR was amended by the Data (Use and Access) Act 2025. Its section 80, in force since 5 February 2026, replaced the general restriction on solely automated decisions with a more permissive framework built on safeguards: information about significant decisions, a way to challenge them and access to human intervention.
Sources
- EUR-Lex: Regulation (EU) 2016/679 (GDPR), Articles 5, 6, 9, 12, 13, 14, 21, 22, 28 and 30 on lawful bases, information at collection, automated decisions and processor contracts (read 6 November 2026)
- European Data Protection Board: Data controller or data processor, the controller determines purposes and means and the processor acts only on its instructions under a contract (read 6 November 2026)
- European Data Protection Board: Opinion 28/2024 on AI models, the three-step legitimate interest test and a conversational agent as an example of a legitimate interest (read 6 November 2026)
- EU Artificial Intelligence Act, Article 50: people must be informed that they are interacting with an AI system, at the latest at the first interaction, from 2 August 2026 (read 6 November 2026)
- GOV.UK: Data (Use and Access) Act 2025, data protection and privacy changes, including a more permissive framework for automated decisions with safeguards (read 6 November 2026)
- legislation.gov.uk: The Data (Use and Access) Act 2025 (Commencement No. 6) Regulations 2026, bringing section 80 into force on 5 February 2026 (read 6 November 2026)