Privacy Policy
Last updated: 2026-04-30. Companion document: Data Processing Agreement.
1. Who we are
AutomateIT is a trading name of AutomateIT Online Ltd, a private company limited by shares incorporated in England and Wales (company number 17096237), with its registered office at 71-75 Shelton Street, Covent Garden, London, WC2H 9JQ. In this notice, "we", "us", "our" and "AutomateIT" all refer to AutomateIT Online Ltd.
We are registered with the UK Information Commissioner's Office (ICO) under registration number ZC129488 (Tier 1 fee payer). The director responsible for data protection matters is Ivan Aguilar Mari.
AutomateIT is an AI phone agent built for UK tradespeople. Plumbers, electricians, gas engineers and similar small business owners sign up, complete a short setup, and get a UK phone number. From then on, an AI receptionist answers their calls, books jobs, sends payment links, and escalates emergencies.
You can contact us about anything in this notice in one of three ways:
- Email Ivan directly: ivan@automateitonline.co.uk
- Email our customer inbox: hello@automateitonline.co.uk
- Write to us at the registered office address above, marked for the attention of the Director.
For data protection matters specifically you can also email privacy@automateitonline.co.uk. This is a function alias forwarded to the accountable person at the time.
We aim to acknowledge data protection enquiries within 5 working days and give you a full reply within 30 days. (This is the timeline UK GDPR sets in Article 12.)
AutomateIT is currently operated by a single director. We do not have a separate compliance team. The response time targets in this notice reflect that, and the Director is personally accountable for every commitment in it.
Our company name, place of registration, registered number and registered office are also displayed in the footer of every page on https://automateitonline.co.uk in line with the Companies (Trading Disclosures) Regulations 2008 (SI 2008/495), made under sections 82 to 85 of the Companies Act 2006.
1.1 Data Protection Officer
AutomateIT has assessed whether a Data Protection Officer is required under UK GDPR Article 37. We do not meet the mandatory thresholds: we are not a public authority; the volume of end-caller data we currently process is not "large scale" in the ICO sense at the present number of customer accounts; and we do not process special category data on a large scale by design. The Director, Ivan Aguilar Mari, is the accountable person. We will reassess this position annually, on any material change to the volume or nature of processing, and immediately on either of the following thresholds being crossed: (a) 50 active customer accounts on the platform, OR (b) 5,000 inbound end-caller calls processed in any rolling 30-day window. The reassessment outcome is recorded in this notice's change log. We will appoint a DPO (internal or contracted) before reaching the relevant Article 37 threshold.
2. Scope of this notice
AutomateIT wears two legal hats and this notice covers both. (1) For tradespeople who sign up as our customers, and for visitors to the marketing site, AutomateIT is the data controller: we decide what data we collect and why. (2) For members of the public who phone a tradesperson's AutomateIT number, AutomateIT is a processor acting on the tradesperson's instructions; the tradesperson is the controller of those calls. Section 3 covers each group separately.
This notice does not cover external websites you reach by following a link from ours. Each of those will have its own privacy notice.
3. The personal information we collect, and why
We have grouped this section by the relationship you have with us. Each subsection lists the categories of data, the source, why we use it and where it is stored. Section 5 maps each purpose to a UK GDPR lawful basis.
3.1 Marketing-site visitors
If you simply browse the marketing site, we collect very little. The hosting and content delivery layers handle some technical data automatically:
| Data | Source | Why we collect it | Storage |
|---|---|---|---|
| IP address, browser user-agent, referrer, requested URL, response status, timestamp | Automatically from your browser | Edge security, abuse and bot mitigation, debugging | Cloudflare edge logs, Railway server logs |
Theme preference (light / dark) | Set by you via the theme toggle | Remember your visual preference between visits | Browser localStorage key theme |
| Preview-access cookie | Set when an authorised tester enters the preview password during the soft-launch phase | Restrict pre-launch site access | First-party cookie preview-access, httpOnly, sameSite lax, 30 days |
Swipe the table sideways ›
If you fill in our contact form to request a demo, we collect the following directly from you and store it in our contact_submissions table in our application database:
- Your name
- Your email address
- Your business name and the trade you operate in
- A phone number we can call you back on
- Optional: your business website URL or Google Place identifier
- Your typical weekly call volume (a rough banding you choose from a dropdown)
- The frustration that prompted you to contact us (a short choice from a list)
- Any free-text message you choose to add
We use this to call you back, run a personalised demo and follow up. When you submit the form we generate a unique signup token tied to your submission so we can pre-fill your account if you decide to sign up.
If you accept a demo call, we then create a row in our demo_bookings table covering:
- The booking type (immediate or scheduled), the scheduled time, and a reschedule token for the public reschedule link we email you
- Attempt count and the latest call status (whether we reached you, were sent to voicemail, etc.)
- A copy of the call transcript, summary, outcome and duration that our AI demo agent ("Lucy") returns after the conversation
- Whether and when we sent you a signup link by SMS or email
If you accept a demo call, your end of the conversation is recorded by Retell on the same basis as live calls. The recording is held at Retell for up to 6 months and the transcript on demo_bookings for up to 12 months.
PECR Regulation 19. PECR Regulation 19 requires specific prior consent for unsolicited automated calls. The AutomateIT contact form records explicit consent at submission, including the version of the consent text the customer agreed to and the timestamp. Automated demo calls are placed only where this consent is on file. Consent can be withdrawn at any time by emailing privacy@automateitonline.co.uk. AutomateIT demo calls are solicited (the prospect submitted the contact form to request the callback), so Regulation 19 may not strictly apply; the explicit consent is captured nonetheless as a defensive overlay and as the UK GDPR Article 6(1)(a) basis for the recording.
Caller-identification opener. The first thing Lucy says on every demo call is the following, verbatim (with the recipient's first name interpolated where the form supplied one): "Hi, is this {customer_name}? This is Lucy from Automate I.T. about the demo you booked. Just so you know, calls are recorded for quality and you're welcome to hang up any time. Got a couple of minutes for me?" This identifies the caller (AutomateIT Online Ltd; the brand "AutomateIT" is spoken as "Automate I.T." so the text-to-speech reads each letter individually), states the purpose of the call (the demo the recipient booked), discloses that the call is recorded, and gives the recipient an immediate verbal opt-out. The opener is set as the Retell begin_message in scripts/setup-retell-demo-agent.mjs and pinned in Lucy's system prompt at src/lib/lucy-prompt.mjs; it cannot be shortened or skipped by the model. Lucy does not announce upfront that she is an AI (the demo is built around the prospect realising mid-conversation that they have just had a natural exchange with an AI agent), but if any caller sincerely asks whether they are speaking to a person or a machine, Lucy is required by her system prompt to confirm honestly that she is an AI. If a caller asks where to find this notice or how to contact us during the call, Lucy directs them to automateitonline.co.uk/privacy or to ivan@automateitonline.co.uk.
PECR Regulation 24's caller-identification information requirements (postal address or freephone number) apply to unsolicited calls. As above, AutomateIT demo calls are solicited and Regulation 24 does not strictly engage. The opener nonetheless identifies the caller (Lucy, calling for AutomateIT Online Ltd, with the brand "AutomateIT" spoken as "Automate I.T."), states the purpose of the call (the demo the prospect booked), discloses recording, and gives an immediate verbal opt-out, which meets the underlying transparency goal Regulation 24 protects.
When you use our chat widget on the marketing site, your messages and the bot's replies pass through our chat provider, ElevenLabs Conversational AI. We do not currently store transcripts of those chats in our own systems. ElevenLabs may hold them for a short period for service operation; their privacy policy applies to that processing.
3.2 Customer accounts
Once you sign up to AutomateIT we collect more, because we need it to run the product. The categories below are split into three sub-tables for readability.
3.2.1 Account and login
| Data | Source | Why we collect it | Storage |
|---|---|---|---|
| Email address and password (the password is hashed, we never see it) | You, at signup | Authentication and account recovery | Supabase Auth (auth.users) |
| Password-reset tokens | Generated by Supabase Auth | Allow you to reset a forgotten password | Supabase Auth |
| Sign-in cookies | Set on successful login | Keep you signed in to your account | Supabase Auth cookies (see section 10) |
Swipe the table sideways ›
3.2.2 Business profile and configuration
| Data | Source | Why we collect it | Storage |
|---|---|---|---|
| Owner name and trade type | You, in onboarding | Personalisation; choosing the right onboarding template | businesses row |
| Business name, business email, business phone, business address and postcode | You, or pre-filled from Google Places when you confirm a Google business | Showing the agent's caller ID, sending notifications, validating service areas | businesses row |
| Google Place identifier | Google Places API when you confirm your business | Preventing duplicate signups for the same physical location | businesses.google_place_id |
| Services you offer | You, in onboarding (some pre-filled from a public scrape of your website) | Telling the AI agent what you do | services table |
| Opening hours, including up to three slots per day | You, in onboarding | Letting the AI agent know when you are available | working_hours table |
| Service-area postcodes | You, in onboarding | Validating jobs against your coverage area | service_areas table |
| Deposit rules, alternative escalations, SMS preferences (six on/off toggles), dismissed banners, agent avatar status, error code and reason | You; some columns set automatically by the app | Configuring the AI agent's behaviour and the dashboard UI | businesses JSONB columns (deposit_rules, alternative_escalations, sms_preferences, dismissed_banners); agent_avatar_status, agent_avatar_error_code, agent_avatar_error_reason |
| Emergency scenarios (semantic descriptions, not literal keywords) | You, in onboarding | Helping the agent recognise emergencies during a call | businesses.emergency_keywords (text[]) |
| Greeting template, tone setting, AI agent name, voice selection | You, in onboarding | Configuring the voice and personality of the AI agent | businesses row |
| Plan name and minutes-used counter | You and our usage tracker | Billing, plan-limit enforcement | businesses row |
Swipe the table sideways ›
3.2.3 Integrations, billing, files and feedback
| Data | Source | Why we collect it | Storage |
|---|---|---|---|
| Stripe Connect account ID and "details submitted" flag | Stripe, after you connect | Sending payment links from your Stripe account, not ours; we do not store any card data | businesses.stripe_account_id and businesses.stripe_details_submitted |
| Google Calendar refresh token and selected calendar ID | Google, after you authorise the integration | Reading availability, writing booked appointments to your calendar | businesses.google_refresh_token and businesses.google_calendar_id |
| Twilio phone number, Twilio number SID, phone routing type, porting status | Issued by Twilio when you provision or port a number | Routing inbound calls to your AI agent | businesses row |
| Logo image file | Uploaded by you, or fetched as a Google favicon if you don't upload one | Displayed in the dashboard, used as a reference for AI avatar generation | Supabase Storage logos/{business_id}/logo.{ext} (public bucket) |
| Generated agent avatar image, the description you typed and the gender you selected | You, plus the AI image generator | Branded illustration of your AI receptionist | Supabase Storage avatars/{business_id}/ and businesses.agent_avatar_url |
| Avatar generation audit log (one row per generation, including the source description, prompt and outcome) | App, on every avatar generation attempt | Anti-abuse and rate-limiting | avatar_generations table |
| Customer-feedback submissions, replies, internal admin notes | You, if you click "Share feedback"; classification metadata added by Claude | Acting on your feedback, replying to you, building a roadmap | feedback_submissions table |
| Structured feedback tasks derived from your submission | App and Claude, on send-reply | Internal admin work tracking | feedback_tasks table |
| Weekly wins recap rows | App and Claude, on a weekly cron | Internal admin reporting on positive feedback | weekly_wins_recaps table |
Help-widget conversations and callback requests, including snapshot fields and the business_data JSONB | You, when you ask for help during onboarding | Live human follow-up when the AI hits a wall | callback_requests table |
| AI assistant chat history | You, when you use the dashboard AI assistant; assistant replies generated by Claude on your account's behalf | Render the assistant in your dashboard; carry conversational context across turns; preserve a transcript of decisions taken on your account | assistant_messages table |
| AI assistant memory facts | Extracted by Claude Haiku from your conversations with the assistant; embedded by Voyage AI for retrieval | Personalise the assistant's responses to your business by recalling stable preferences and configuration | assistant_memories table |
Swipe the table sideways ›
We do not collect special-category data (health, race, religion, political opinions, biometric or genetic data) for customer accounts. If you tell us such information voluntarily in feedback or a support email, we treat it as ordinary personal data and do not use it for anything other than responding to you.
We do not collect or store payment card data. Customer billing payments to AutomateIT itself, when self-serve billing is added, will run through Stripe Checkout; we will only store the Stripe Customer ID and the metadata Stripe returns to us.
All of the customer-account fields above are necessary for AutomateIT to operate. If you do not provide them you will not be able to complete onboarding or use the service. Optional fields are clearly marked in the form.
3.3 End-caller data (AutomateIT acts as a processor)
When someone phones your AutomateIT number, the information from that call belongs to your business, not to us. In data protection terms, you are the "controller" (you decide what the information is used for) and we are your "processor" (we handle it for you on your instructions). In practice that means:
- You decide why the data is collected (to take a booking, take a payment, escalate to you, etc.).
- You set the retention period within the limits we make available to you.
- We process the data only on your documented instructions, which are the configuration choices you make during onboarding and in Settings.
The categories we process on your behalf include:
- The caller's phone number (as supplied by Twilio in the call signalling)
- Their name, where they say it
- Their address or postcode, where they share it for routing or service-area validation
- The full audio recording of the call, the speech-to-text transcript, and the AI's post-call summary, outcome label and sentiment classification
- Job details captured during the call (service requested, scheduled date and time, deposit status)
- Outbound SMS we send on your instructions (booking confirmation, payment link, etc.)
- Cost and duration metadata for billing and analytics
The AI agent plays a short call-recording disclosure at the start of every call so callers know the call is being recorded. The disclosure currently says: "One quick thing, calls are recorded for quality." This is delivered in the agent's first spoken message before the agent takes any details from the caller. You as the controller are responsible for any further notice obligations to your callers, for example a written privacy notice on your own website that explains how you use the recording, transcript and any onward processing.
We sign or commit to UK GDPR Article 28 processor terms with you as part of our customer agreement. Those terms include sub-processor disclosure (section 6 of this notice), security obligations (section 11), assistance with data subject requests (section 9), and breach notification.
Joint-controller note (Article 26). For most aspects of the call (recording, transcription, routing, job creation, post-call analytics) AutomateIT acts as the customer's processor and the customer is the controller. There is one specific element where the design choice is AutomateIT's own and materially affects callers' rights: the auto-injected pre-call recording disclosure ("One quick thing, calls are recorded for quality.") is determined by AutomateIT, not the customer, and the customer cannot remove it. For that one element AutomateIT and the customer are joint controllers. The essence of the arrangement is that AutomateIT determines the disclosure wording (to satisfy UK GDPR Article 13 and PECR transparency), while the customer determines who is called, when, and what the call is about. Either party can be approached for a data subject rights request and the parties cooperate to respond.
3.3.1 Special category data in calls
Live phone conversations sometimes capture information that falls within Article 9 of the UK GDPR, most commonly health information shared by a caller during an emergency. We do not solicit such information and the AI agent is not configured to ask for it. Where it is volunteered we process it only to the extent necessary to (a) help the AI agent route the call appropriately and (b) provide the customer with the recording and transcript they need to do their job. To the limited extent AutomateIT itself processes special category data as a controller (for example, when an admin reviews a call transcript to support a customer or to investigate a service incident), we rely on Article 9(2)(c) of the UK GDPR for vital interests in emergencies involving risk to life, Article 9(2)(f) for the establishment, exercise or defence of legal claims, and Article 9(2)(g) substantial public interest read with paragraph 10 of Schedule 1 to the Data Protection Act 2018 (preventing or detecting unlawful acts) and, for any criminal-offence content under Article 10 of the UK GDPR, paragraph 33 of Schedule 1 (legal claims). An Appropriate Policy Document is maintained as required when relying on the Schedule 1 substantive paragraphs above (DPA 2018 Schedule 1 paragraph 39 sets the required content). Customers (acting as controllers) are responsible for ensuring they have a corresponding Article 9(2) condition for any onward use of the recording or transcript and for telling their callers what they will do with the information.
3.3.2 Member-of-the-public requests
If you are a member of the public who has called a tradesperson using AutomateIT and you have a question about your data, you can either contact the tradesperson directly or email us at ivan@automateitonline.co.uk. Where we hold your data only as the tradesperson's processor, we will route your request to them within 5 working days and copy you on the routing. We will help the tradesperson respond and, where they fail to respond within a reasonable period, we will respond ourselves to the extent permitted under our processor obligations.
3.3.3 No direct relationship with end-callers
End-callers have no contractual or commercial relationship with AutomateIT. They dialled the tradesperson's published number and reached our infrastructure as a consequence. This is why our role is processor-only, why the recording disclosure is mandatory and auto-injected by our backend, and why we publish this notice in language that is meaningful to a member of the public who has just had their voice recorded.
3.3.4 Recording disclosure as a built-in control
Unlike many call-recording setups where the disclosure depends on the customer remembering to read it, AutomateIT injects the disclosure into every customer's begin_message server-side at call start, by way of applyRecordingDisclosure() in src/lib/retell-recording-disclosure.ts. The customer cannot accidentally remove it from their saved greeting; the helper re-applies it on every call. This is a deliberate compliance-by-design control under UK GDPR Article 25.
3.3.5 Vulnerable callers
We recognise that emergency callers may include vulnerable adults, callers in distress, callers with hearing or cognitive impairments, and (occasionally) minors using a parent's phone. The recording disclosure played at the start of every call is short and clear by design. The customer (tradesperson) remains the controller and is responsible for any further accessibility-of-notice obligations under the Equality Act 2010. We do not perform automated profiling of caller vulnerability and we do not sell or share end-caller data with anyone outside the sub-processors listed in section 6.
3.3.6 Stripe Checkout payment-link flow
When the agent sends a payment link during a call, AutomateIT also creates a Stripe Checkout Session on the customer's connected Stripe account containing the deposit amount, a reference to the job, and (depending on session config) the caller's phone or email so they can be sent a receipt. AutomateIT does not store any card data.
3.3.7 Where the call audio lives
Call audio is held at Retell. AutomateIT proxies the audio on demand from Retell to the authenticated customer's browser via /api/calls/[id]/recording. The proxy streams the bytes through; AutomateIT does not write the audio to disk or to any persistent store.
3.3.8 Job rows
Each booked job is stored on the jobs row with customer_name, customer_phone, customer_email, service, description, postcode, scheduled_at, deposit_amount, deposit_status, stripe_session_id and google_event_id. We retain those rows for 6 years from job completion to align with HMRC accounting record requirements (see retention policy section 4.7).
3.3.9 International transfers of end-caller data
Where end-caller data is transferred outside the UK to a sub-processor (currently Retell AI in the United States, and onward to Retell's own infrastructure provider), we transfer it under the UK-US Data Bridge (where the sub-processor is DPF/UK-Extension certified) or under the UK Addendum to the EU SCCs (where it is not), in each case as the customer's authorised sub-processor under their processor agreement with us. The customer is the controller of those onward transfers and our processor agreement records the customer's general authorisation under UK GDPR Article 28(2) for us to engage the sub-processors listed in section 6.
3.4 Where we got your data from (Article 14)
Most of the data we hold about you comes directly from you. The exceptions are:
- Your business name, address, business phone, opening hours and Google Place identifier may be pre-filled from the public Google Places record you confirm during onboarding. The source is the public Google Places API.
- A candidate list of services and service areas may be extracted from the public homepage of the business website you give us. The source is your own public website. We perform this extraction once at onboarding using Anthropic Claude as described in section 4. The fetched HTML is not retained after the extraction completes.
- A 128px favicon may be fetched from your Google Places record if you have not uploaded a logo.
- During onboarding, your typed company-name search is sent to the public Companies House API for a register lookup.
For end-callers (people who dial your AutomateIT number), the caller's phone number is supplied by Twilio in the call signalling. We do not look up the caller in any external database. All other end-caller data is supplied by the caller themselves during the conversation.
4. How we use AI and automated processing
Several parts of the product are powered by third-party AI services. The table below summarises each one. None of these systems make decisions that produce a legal effect on you or anything similarly significant within the meaning of UK GDPR Article 22.
| AI service | What it does | What it sees | Where output goes |
|---|---|---|---|
| Anthropic Claude (onboarding help widget) | Suggests edits and answers questions while you fill in onboarding | Full wizard state, including services, opening hours, service areas, escalation phone numbers and emergency scenarios; plus messages to the widget | Returned to your browser as suggested edits you can accept or reject |
| Anthropic Claude (website-scrape extraction) | Extracts a candidate list of services and service areas from your public homepage, once at onboarding | The public HTML of the homepage you provide | Pre-fills onboarding Step 2 and Step 3; the HTML is discarded after extraction |
| Anthropic Claude (feedback workflow) | Classifies feedback (bug, feature, preference, positive), polishes admin reply drafts, and sends a 24-hour positive auto-reply | Your feedback text and admin notes | feedback_submissions.category, the email reply we send you, and feedback_tasks for bugs and feature requests |
| Anthropic Claude (weekly wins recap) | Aggregates positive feedback into a short prose summary used by our admin | Positive feedback rows from the past week | weekly_wins_recaps table, internal admin only |
| Anthropic Claude Haiku (avatar description sanitisation) | Rewrites the customer's typed avatar description before it is sent to Gemini, neutralising crude language and rejecting content-policy violations | Your typed description and the gender you selected | An expanded, professional description passed only to Gemini; the raw description you typed is what we store |
| Google Gemini ("Nano Banana 2") image model | Generates the AI receptionist avatar | Your Haiku-sanitised description, your uploaded logo as a reference image, and the gender you selected | Stored in Supabase Storage and businesses.agent_avatar_url |
| Retell AI | Speech-to-text, natural-language understanding, function calling, text-to-speech, post-call sentiment and emergency classification | Live audio of every call to your AutomateIT number | Audio held at Retell; transcript, summary, outcome and metadata flow back to your calls row |
| Anthropic Claude (dashboard AI assistant) | Powers the in-product chat that reads your account data and can take configuration actions on your behalf; Claude Haiku also extracts memory facts from your conversations | Your messages, the conversation history, and a tenant-scoped read view of your business profile, services, working hours, calls, jobs and customers | assistant_messages table; resulting actions write to the relevant business tables; extracted facts go to assistant_memories |
| Voyage AI (memory embeddings) | Generates 1024-dimensional embeddings of memory facts so the assistant can retrieve relevant context on each turn | The fact text only (no surrounding conversation; no business profile data) | assistant_memories.embedding column (vector(1024)) |
Swipe the table sideways ›
When you generate an agent avatar we send Gemini (a) your typed description (after sanitising by Claude Haiku to remove crude language and content-policy violations), (b) your uploaded business logo as a reference image, and (c) the gender you selected. Gemini returns a generated image that we resize and store. We never send Gemini your customer or call data.
The auto-classifier on feedback has a manual override built in. If the classifier picks the wrong category, our admin can change it from a dropdown. The classifier output is metadata only and does not change what we send to you.
4.1 Article 22 and significant decisions
None of the AI features in AutomateIT currently produces a "significant decision" within the meaning of Articles 22A to 22D of the UK GDPR (as inserted by Section 80 of the Data (Use and Access) Act 2025). They suggest, polish, summarise, transcribe or classify, and a human or the customer's own pre-configured rules drive the customer-facing outcome. The Retell agent makes automated routing decisions during a live call (whether to treat a call as an emergency, whether to escalate to a human, whether to send a payment link). We treat these as decisions with potentially similar significance and apply the following safeguards: a human (the tradesperson or their nominated escalation contact) is always reachable as a fallback; the agent does not finalise a payment without explicit caller confirmation; a caller who is unhappy with how the agent handled their call can ask for human review by contacting the tradesperson or AutomateIT directly. If we ever introduce a feature that would amount to a significant decision based solely on automated processing, we will tell the affected data subject, give them a route to make representations and to contest the decision, and enable meaningful human intervention.
4.2 No training on your data
None of our AI sub-processors are permitted to use your customer or call data to train or improve their general models. Anthropic's Commercial Terms exclude commercial API inputs and outputs from training. Google's paid Gemini API does not use prompts to improve products. Retell does not train on customer data. We monitor each provider's training-default position and will switch off any flag or stop using the provider for the affected workflow if a change would put customer or end-caller data into training without an opt-in.
5. Lawful bases for processing
In plain English: most of what we do with your data is to provide the service you signed up for. Where that's not the legal hook, the table below shows what is.
5.1 Special category and criminal-offence data
To the limited extent AutomateIT processes special category data as a controller (see section 3.3.1), we rely on Article 9(2)(c) of the UK GDPR for vital interests in emergencies, Article 9(2)(f) for legal claims, and Article 9(2)(g) substantial public interest read with paragraph 10 of Schedule 1 to the Data Protection Act 2018 (preventing or detecting unlawful acts). For criminal offence data under Article 10 of the UK GDPR (for example where a call recording happens to capture an allegation of an unlawful act), we rely on paragraph 33 of Schedule 1 (legal claims) and paragraph 36 read with paragraph 10 of Schedule 1 (preventing or detecting unlawful acts), and only to the extent strictly necessary. An Appropriate Policy Document is maintained as required when relying on the Schedule 1 substantive paragraphs above (DPA 2018 Schedule 1 paragraph 39 sets the required content).
5.2 Article 6 lawful bases
| What we do with your data | Why we are allowed to (legal basis) | Notes |
|---|---|---|
| Creating and operating your AutomateIT account | Article 6(1)(b) contract | Necessary to provide the service you signed up for. |
| Onboarding, including the Google Places lookup, the website scrape, and Claude-assisted extraction | Article 6(1)(b) contract | All of these are part of provisioning the service. |
| Storing your business configuration and using it to drive call handling | Article 6(1)(b) contract | |
| Sending you transactional emails (signup link, demo confirmation, reschedule confirmation, "we missed you" follow-up, feedback reply) | Article 6(1)(b) contract for booked customers; Article 6(1)(f) legitimate interests for prospects who submitted the contact form | These are not marketing emails. The legitimate-interests outcome for prospects is that completing the demo workflow they asked for outweighs the limited intrusion of a single follow-up sequence. |
| Logging, monitoring, rate-limiting and abuse prevention on our APIs and crons | Article 6(1)(f) legitimate interests | Outcome: keeping the service available and safe for paying customers outweighs the limited intrusion of short-lived server logs. |
| AI image generation for your agent avatar (Gemini) | Article 6(1)(b) contract | The avatar is a feature you opt into. |
| Internal AI classification of customer feedback to triage admin work (Claude) | Article 6(1)(f) legitimate interests | Necessary for operating an efficient feedback queue. Outcome: the metadata-only classifier does not change what we send you and lets us respond faster. You can object at any time. |
| AI-assisted reply to your feedback, including the 24-hour auto-reply for positive feedback (Claude) | Article 6(1)(b) contract | Part of the customer support layer of the service. Email us to switch to fully manual replies. |
| Recording, transcribing and analysing your demo call with Lucy | Article 6(1)(a) consent, recorded at the contact form, with Article 6(1)(f) legitimate interests as the residual basis if consent is later withdrawn but the recording is needed to establish, exercise or defend a legal claim | Consent is captured at submission with the consent-text version, timestamp, IP and user-agent. The call is announced as recorded at the start. The demo call is solicited (you requested the callback when you submitted the contact form), so PECR Regulation 19 may not strictly engage; the explicit consent is captured as a defensive overlay. We keep the recording for up to 6 months and the transcript for up to 12 months. You can withdraw consent or ask us to delete both at any time by emailing privacy@automateitonline.co.uk. |
| Sending you product update or marketing emails | Article 6(1)(a) consent | We will only send marketing emails to people who have positively opted in. Until that opt-in mechanism is built, we do not send marketing emails. |
| Retention of call traffic and cost metadata (duration, cost, time) | Article 6(1)(b) contract, with PECR Regulation 7(2) for billing-purpose retention | We retain only the minimum traffic data needed for billing reconciliation, for 24 months from the call date, after which the row is anonymised per the data retention policy. |
| Retaining invoicing records for HMRC | Article 6(1)(c) legal obligation | Six years from the end of the relevant accounting period. |
| Responding to ICO requests, court orders, law-enforcement requests | Article 6(1)(c) legal obligation | We will tell you about a request unless we are legally prohibited. |
| Processing end-caller phone-call data | Acting as processor on the customer's instructions; the customer's own lawful basis applies | See section 3.3. |
Swipe the table sideways ›
All SMS sent by AutomateIT, both to prospects and to customers, are transactional under PECR. We do not send SMS marketing. If we ever introduce SMS marketing in future, we will obtain prior consent under PECR Regulation 22 and update this notice and our forms before doing so.
PECR Regulation 22 boundary for prospect follow-up emails. Regulation 22 governs unsolicited commercial email. AutomateIT sends two types of email to prospects who have submitted the contact form but not yet signed up:
(a) Transactional follow-up of a demo the prospect explicitly requested. Examples: the signup link sent after Lucy successfully demos; the rescheduling email sent if Lucy could not reach the prospect; the "we missed you, here is your signup link" email sent after the third failed call attempt. These are continuations of the demo the prospect actively requested. They contain no promotional content beyond the link to complete the signup the prospect already initiated. Lawful basis: Article 6(1)(b) UK GDPR (steps at the prospect's request prior to entering into a contract). PECR Regulation 22 does not engage because these are not unsolicited marketing communications.
(b) Marketing emails to prospects (e.g. promotional newsletters, product update announcements, re-engagement campaigns). AutomateIT does not currently send these. If we ever did, Regulation 22 would require prior consent (the soft opt-in only covers existing customers, not prospects). The contact form would need an additional, unbundled consent checkbox before any such marketing could be sent.
AutomateIT operates on the strict (a) basis only. Any future move to (b) requires a privacy notice update, a contact-form consent change, and a written scope decision recorded in section 13.
The lawful-basis row above ("Sending you transactional emails ... Article 6(1)(f) legitimate interests for prospects") should be read together with this boundary paragraph: the legitimate-interests language reflects an operator-side balancing test for the residual logging and abuse-prevention layer around the email send, while the customer-facing communication itself rests on the Article 6(1)(b) pre-contractual basis described in (a) above.
6. Sharing and sub-processors
We never sell personal information. We do share it with carefully chosen sub-processors who run parts of the platform on our behalf. Each one is contractually bound to use the data only for the purpose we set, to keep it secure, and to delete it when we ask.
The list below is current at the effective date of this notice. We will update it when it changes.
| Provider | Role | Location | Transfer safeguard |
|---|---|---|---|
| Supabase (Supabase Inc., US-incorporated) | Application database, authentication, file storage | Project region: EU (eu-west-1, Dublin, Ireland) | Counter-signed Data Processing Addendum incorporating EU SCCs and the UK Addendum for any access by Supabase staff in the United States. |
| Retell AI (Retell, Inc., US) | Live call handling, speech-to-text, AI dialogue, recording, post-call analysis | United States | UK Addendum to the EU SCCs under Retell's published DPA. We have signed Retell's click-through DPA before relying on this. |
| Twilio (Twilio Inc., contracting through Twilio Ireland Limited for the EEA) | UK phone number provisioning, inbound call routing, SMS delivery | EEA and United States | EU-US Data Privacy Framework with UK Extension (Twilio Inc. is DPF-certified). EU SCCs and UK Addendum apply as fallback. |
| Stripe (Stripe Payments Europe Limited as UK-customer contracting entity; Stripe LLC as US data importer) | Stripe Connect Express accounts; no card data on our servers | Ireland and United States | UK adequacy regulations (EEA) for the Ireland leg. EU-US Data Privacy Framework with UK Extension for the US leg (Stripe LLC is DPF-certified). EU SCCs and UK Addendum apply as fallback. |
| Google (Google Ireland Limited for Cloud, Workspace and Gemini contracting; Google LLC, US) | Google Calendar API, Google Places API, Geocoding API, Gemini image generation, Workspace email aliases | EEA and United States | UK adequacy regulations (EEA) for the Ireland leg. EU-US Data Privacy Framework with UK Extension for the US leg (Google LLC is DPF-certified). EU SCCs and UK Addendum apply as fallback under the Cloud DPA. |
| Anthropic (Anthropic, PBC, US) | Claude API for help widget, website-scrape extraction, feedback classify, polish and auto-reply, weekly wins recap, avatar description sanitisation | United States | UK Addendum to the EU SCCs under Anthropic's Commercial Terms (Anthropic is not DPF-certified per public verification on 2026-04-30; SCCs and UK Addendum are the primary safeguard, and a Transfer Risk Assessment is held in docs/legal/tra-register.md). Default API retention 30 days for trust-and-safety review (longer for content flagged for safety review); commercial inputs and outputs are not used for model training. |
| Voyage AI (Voyage AI Inc., US) | Embedding generation for dashboard AI assistant memory retrieval (voyage-3, 1024-dim) | United States | UK Addendum to the EU SCCs (Voyage AI is not DPF-certified per public verification on 2026-05-04). Per Voyage AI's published terms, API inputs and outputs are not used for model training. A Transfer Risk Assessment for Voyage AI is scheduled to be added to docs/legal/tra-register.md before the Phase 9 dashboard assistant ships to live customers. |
| Resend (Resend, Inc., US) | Transactional email delivery (signup links, demo confirmations, feedback responses) | United States | EU-US Data Privacy Framework with UK Extension (Resend, Inc. is DPF-certified). EU SCCs and UK Addendum apply as fallback. Resend deletes customer data within 90 days of account termination. |
| ElevenLabs (ElevenLabs Inc., US; Eleven Labs Poland sp. z o.o. as EU controller for some voice data) | Conversational AI chat widget on the marketing site only | United States | EU-US Data Privacy Framework with UK Extension (ElevenLabs Inc. is DPF-certified). EU SCCs and UK Addendum apply as fallback. |
| Cloudflare (Cloudflare, Inc., US, as data importer; Cloudflare, Ltd. for UK presence) | DNS, CDN and edge proxy for our marketing site | Global edge network | EU-US Data Privacy Framework with UK Extension (Cloudflare, Inc. is DPF-certified). EU SCCs and UK Addendum apply as fallback. |
| Railway (Railway Corp., US) | Application hosting and cron services | United States (with Netherlands and Singapore options offered for paid services) | EU-US Data Privacy Framework with UK Extension (Railway Corp. is DPF-certified). EU SCCs and UK Addendum apply as fallback. |
| Companies House (UK government) | Public register lookup for company-name searches during onboarding | United Kingdom | UK government, no transfer outside the UK. |
Swipe the table sideways ›
For clarity, the categories of recipients of your personal data are: (a) the sub-processors listed above; (b) our professional advisers (accountants, solicitors, insurers) under professional confidentiality; (c) regulators, courts and law-enforcement bodies where we are legally required to disclose; (d) a buyer of our business in the event of a sale, on terms at least as protective as these. Other AutomateIT customers do not see your data; the application enforces row-level security on every customer-facing table so each customer can only see their own rows.
We will notify customers at least 30 days before adding or changing a sub-processor that processes their personal data, and customers may object on reasonable data-protection grounds during that window.
Several sub-processors above use their own infrastructure providers (typically Amazon Web Services, Google Cloud Platform or Microsoft Azure). Each sub-processor publishes its own sub-processor list and is contractually required to flow our transfer protections down to those onward recipients. Links to the relevant sub-processor lists are available on request.
7. International data transfers
In short: data we send abroad goes to the EEA (covered by UK adequacy) or to the US (covered by the UK-US Data Bridge where our provider is certified, otherwise by a contractual safeguard).
The UK has made adequacy regulations for the European Economic Area (the Data Protection (Adequacy) (European Union) Regulations 2021 and the Data Protection (Adequacy) (European Economic Area) Regulations 2021). Transfers to providers in Ireland, Germany or any other EEA state therefore do not need additional safeguards. We will update this policy if the EEA adequacy regulations are not extended beyond their current sunset date.
The UK Government has made adequacy regulations (in force from 12 October 2023) covering transfers to US organisations that are self-certified under the EU-US Data Privacy Framework and have opted into the UK Extension. Where a US sub-processor is so certified, we rely on the UK-US Data Bridge.
For transfers to US providers where we rely on the UK IDTA or the UK Addendum to the EU SCCs (rather than on the UK-US Data Bridge), we will complete and document a Transfer Risk Assessment in line with ICO guidance, considering the laws of the destination country, the supplementary measures in place, and the practical likelihood of access by foreign public authorities. The current TRA register is held by the Director and available to the ICO on request.
We do not rely on the Article 49 derogations as a routine transfer mechanism. They are reserved for occasional, non-systematic transfers where neither adequacy nor an Article 46 safeguard is available.
The UK Addendum we use is the International Data Transfer Addendum to the EU Commission Standard Contractual Clauses, Version B1.0, in force from 21 March 2022.
If you would like a copy of the relevant transfer mechanism for any sub-processor, ask using the contact details in section 1.
8. How long we keep information
Detailed retention periods, trigger events and disposal methods are in the AutomateIT Data Retention Policy at docs/legal/data-retention-policy.md.
| Data | Retention |
|---|---|
| Marketing-funnel data (contact form or demo booking from a prospect who did not sign up) | 12 months from last activity |
| Lucy demo call recording | Up to 6 months at Retell |
| Lucy demo transcript and booking | Up to 12 months from last activity |
| Customer account, business profile and onboarding data | Lifetime of the account, then 30 days grace, then deleted |
| End-caller call recording | Up to 12 months from call date (held at Retell) |
| End-caller transcripts and call metadata | 24 months from call date, then anonymised |
| Job records | 6 years from job completion (HMRC accounting record requirement) |
| Customer feedback | 24 months from submission |
| Invoicing records | 6 years from end of relevant financial year |
| Server logs | Short-lived; exact figures listed in the data retention policy as we confirm them with each provider |
| AI assistant chat history | Lifetime of the account; anonymised after 24 months from message creation |
| AI assistant memory facts | Lifetime of the account; deleted within 24 hours of a customer deletion request submitted via the in-product memory viewer |
Swipe the table sideways ›
9. Your rights under UK GDPR
If you're a tradesperson who has signed up: these rights apply to your account data. If you're a member of the public who phoned a tradesperson on an AutomateIT number and want a copy of the call: ask the tradesperson, they're the controller. We will help them respond.
| Right | What it means | How to use it |
|---|---|---|
| Right to be informed | Know how we use your data | This notice is the main way we provide that. |
| Right of access | A copy of your personal data and information about how we use it (a "subject access request") | Email ivan@automateitonline.co.uk. We will provide your data as a JSON file by default and as PDF or CSV on request. |
| Right to rectification | Correct inaccurate or incomplete personal data | Email or fix it yourself in Settings. |
| Right to erasure | Have your personal data deleted in certain circumstances (the "right to be forgotten") | Email ivan@automateitonline.co.uk. We may need to keep certain records (for example invoicing) to meet legal obligations. |
| Right to restriction | Pause our use of your data while a dispute is investigated | Email ivan@automateitonline.co.uk. |
| Right to data portability | Get your data in a structured, commonly used and machine-readable format | Default is a JSON export; CSV available on request. |
| Right to object | Object to processing carried out under legitimate interests on grounds relating to your particular situation. We will stop the processing unless we can show compelling legitimate grounds that override your interests, rights and freedoms, or the processing is needed to establish, exercise or defend a legal claim. You also have an unconditional right to object to direct marketing. If you object to direct marketing we will stop without question. | Email ivan@automateitonline.co.uk. |
| Right not to be subject to automated decisions with significant effects | As explained in section 4.1, the AI features in AutomateIT are assistive rather than fully automated decision-makers. The Articles 22A to 22D safeguards (DUAA 2025) apply if we ever introduce one. | Email ivan@automateitonline.co.uk to ask for human review of any automated outcome. |
| Right to withdraw consent | Where we are relying on consent (for example marketing emails to a non-customer). Withdrawing consent does not affect anything we did before you withdrew it. | Click the unsubscribe link in any marketing email (when we begin sending them), or email ivan@automateitonline.co.uk with the subject line "Withdraw consent". Withdrawal takes effect within 5 working days. |
| Right to lodge a complaint | Complain to the ICO | See section 14. |
Swipe the table sideways ›
To exercise any of these rights, email ivan@automateitonline.co.uk or write to the registered office. We may need to verify your identity before we act, particularly for access and erasure requests. We will respond within one calendar month, and we will tell you within that period if we need to extend by a further two months because the request is complex.
There is no fee for exercising your rights, except in the rare cases UK GDPR allows (for example if a request is "manifestly unfounded or excessive").
Some of these rights are subject to exemptions in Schedule 2 to the Data Protection Act 2018. We do not routinely rely on any Schedule 2 exemption. We may, on a case-by-case basis, rely on paragraph 2 of Part 1 (prevention or detection of crime) when responding to a lawful police or HMRC request, and on paragraph 5 (legal proceedings) to the extent strictly necessary to defend a legal claim. We will record our reasoning whenever we do so. We do not rely on the immigration exemption in section 23 of the DPA 2018 in any context.
AI assistant memory access and erasure. Customers can view their AI assistant memory facts and request deletion of any individual fact at any time from /settings/agent/memory in their dashboard. Deletion requests soft-delete the row immediately (so the assistant stops using the fact on the next turn) and a daily retention cron hard-deletes within 24 hours. Subject access requests covering the wider AI assistant chat history can be made by emailing ivan@automateitonline.co.uk.
10. Cookies and similar technologies
We try to keep cookies to the minimum necessary to run the site. The current set is:
| Cookie or storage entry | Set by | Purpose | PECR category | Lifetime |
|---|---|---|---|---|
sb-<project-ref>-auth-token (possibly chunked across .0, .1) | @supabase/ssr on our app | Keep you signed in to your account | Strictly necessary | Session-bound, refreshed periodically by Supabase |
preview-access | AutomateIT marketing site | Holds the preview password during the soft launch (will be removed at public launch) | Strictly necessary | 30 days, httpOnly, sameSite lax |
| OAuth state cookies | Set during Google Calendar and Stripe Connect connection flows | Prevent CSRF on OAuth round-trips | Strictly necessary | Short-lived, cleared on completion |
__cf_bm | Cloudflare | Bot-management identifier on requests served through the Cloudflare proxy | Strictly necessary | Around 30 minutes |
theme (browser localStorage, NOT a cookie) | AutomateIT site | Remembers your light or dark mode choice | Strictly necessary for a service explicitly requested by the user (PECR's cookie rules also cover localStorage; we treat the theme toggle as such a service) | Until you clear browser storage |
Swipe the table sideways ›
We do not currently set advertising, analytics or third-party tracking cookies. If we ever do, we will do so only with your prior consent through a clear cookie banner.
11. Security
Phone calls and customer data need real protection. Here is what we do:
- Live call audio is carried over encrypted SIP between Twilio and Retell; the recording is stored at Retell, encrypted at rest, and only fetched on demand by our backend (never the customer's browser) when a tradesperson plays it back through
/api/calls/[id]/recording. - Transcripts, AI summaries, business profile data and feedback rows live in Supabase (Dublin), encrypted at rest by Supabase. Logos and generated avatars sit in Supabase Storage with the same encryption.
- All transport between browser, our app on Railway, and every sub-processor is TLS 1.2 or higher.
- Row-level security is enforced on every customer-facing table. Each customer can only see their own rows. Service role keys are server-only and never reach the browser.
- Admin access from developer endpoints is restricted to the Director, MFA-protected, and logged.
- Dependency vulnerability scanning runs on every deploy. Access logs are reviewed monthly.
No system can be guaranteed completely secure. If you discover a security issue with AutomateIT please email ivan@automateitonline.co.uk rather than disclosing it publicly, and we will work with you to fix it.
11.1 Privacy by design and by default
We bake the following controls into the product so they are always on:
- The recording disclosure is server-injected into every customer's
begin_message(section 3.3.4) so it cannot be turned off by accident. - Row-level security on every customer-facing table prevents cross-tenant data access.
- End-caller transcripts are scrubbed at 24 months without admin action (see retention policy).
- We do not collect a "caller name" field by default; the agent only captures it if the caller volunteers it during the conversation.
11.2 Breach process
Internal breach process: any AutomateIT person who detects or suspects a personal data breach must report it to the Director within one working hour. The Director assesses the risk against the Article 33 and Article 34 thresholds, notifies the ICO via the online form within 72 hours where required, notifies affected data subjects where the risk to their rights and freedoms is likely to be high, and notifies any affected customer-controllers within 24 hours so they can meet their own 72-hour ICO notification deadline. Every breach (including those that did not require notification) is recorded in our internal breach log per Article 33(5).
All time periods in this section are calendar hours, not working hours. The 72-hour ICO notification clock and the 24-hour customer-controller commitment run over weekends, bank holidays and Christmas Day. The internal 1-hour-from-becoming-aware reporting target is also calendar hours.
12. Children
AutomateIT is built for adults running their own trade business. Our customer-facing surfaces (signup, onboarding, dashboard) are not directed at anyone under 18. Under UK law, the age at which a child can consent to an information society service is 13 (Article 8 of the UK GDPR as retained in UK domestic law). The phone agent itself may incidentally record a call from a child (e.g. a teenager dialling the plumber on a parent's phone). We treat any such recording as ordinary call data under section 3.3, and the customer-controller is responsible for handling it appropriately under their own privacy notice. If you believe we hold personal data of a child where there is no legitimate reason to, contact us and we will delete it.
In particular: an end-caller dialling a tradesperson on AutomateIT's number may incidentally be a child (e.g. using a parent's phone in an emergency). The customer (tradesperson) is the controller of those calls; AutomateIT, as processor, ensures (a) the recording disclosure is played at the start of every call regardless of caller age, (b) the AI agent is not configured to ask for age, identity verification, or any data category that would treat a child differently, (c) the customer-controller is responsible for any further age-aware decisions on what to do with the recording (e.g. seeking parental consent if they decide to retain it for their own purposes).
13. Changes to this notice
We will update this notice when our processing changes, when we add or remove a sub-processor, or when the law requires us to. The current version and effective date are at the top and bottom of the document. We will tell you about material changes by email if you have an active customer account, and by updating the version note on this page in any case. The previous version remains available on request.
For material changes that affect how your data is used, we will update this notice within 30 days of the change taking effect, and notify active customers by email.
14. Complaints
If you are unhappy with how we have handled your personal data, please contact us first using the details in section 1 so we have a chance to put it right.
You also have the right to complain to the UK Information Commissioner's Office. The ICO's contact details are:
Information Commissioner's Office Wycliffe House Water Lane Wilmslow Cheshire SK9 5AF
Helpline: 0303 123 1113 Online complaint form: https://ico.org.uk/make-a-complaint/
The ICO is the UK's independent regulator for data protection. It can investigate complaints, require us to provide information through an information notice, audit our processing through an assessment notice, require us to take or stop a particular action through an enforcement notice, and impose monetary penalties of up to £17.5 million or 4% of our worldwide annual turnover (whichever is higher) for serious breaches of UK GDPR or the DPA 2018.
You can go straight to the ICO without contacting us first, but we'd appreciate the chance to put things right ourselves.
15. Effective date and version
Version: 1.0 (AI-drafted) Effective date: 2026-05-04 Document owner: Ivan Aguilar Mari, Director, AutomateIT Online Ltd.
16. Data Protection Impact Assessment
Because the AutomateIT platform involves call audio recording, AI sentiment classification, and innovative AI technology applied to UK consumer voice data, we have committed to carrying out a Data Protection Impact Assessment in accordance with UK GDPR Article 35. The DPIA covers (a) the systematic description of the processing, (b) the necessity and proportionality assessment, (c) the risks to the rights and freedoms of data subjects (in particular UK end-callers), and (d) the technical and organisational measures we have adopted to mitigate those risks. The DPIA is reviewed when processing materially changes and at minimum annually. A redacted summary is available to customers and to the ICO on written request.

