Privacy notice

How we handle your information

Last updated September 29, 2026

Agentic Integrations is a company based in the United Kingdom that sets up and runs AI Experts for firms: AI agents trained on each firm's own past work that draft inside the Officeware the firm already uses. This notice explains what we collect, what we do with it and the choices you have.

It covers two different kinds of information. Information about you as a visitor, prospect or user, which we decide how to use. And the work a client firm gives us to train its model and build its library, which belongs to that firm and which we handle only on its instructions.

What we collect

The contact form: your name, email address, company and message, plus the link tags described below if you arrived from a tagged link.
Booking a call: your name, email, the time you chose and any answers you give on the booking form, through our scheduling provider.
Client accounts: the name, email address and role of each person a client firm adds, sign-in records, and a log of the settings they change.
Using the service: what each AI Expert worked on, which drafts were approved, edited or rejected, and the GPU time used, so each firm can see its own usage and check its invoices.
Connections: when a firm connects one of its systems, such as Microsoft 365 or Google Workspace, the access tokens that system issues. We never see or store the passwords.
Emails you send us and our notes from calls with you.

We do not run advertising trackers, analytics that follow you across sites, or sell or share personal information for advertising.

A firm's work

A client firm gives us its past work, such as reports, letters, applications, statements and the records behind them, to train its model and build its library. That work often contains personal information about the firm's staff, clients and contacts. The firm decides what goes in, and we process it only to provide the service to that firm, under our agreement with them and a data processing agreement.

Each firm's work, library and model are kept separate from every other firm's. No firm can read another's, and no firm's work is used for another.
Our staff look at a firm's work only to prepare it for training and indexing, to test the firm's model, or to provide support the firm has asked for, and that access is logged.
Nothing an AI Expert drafts is sent, signed or filed without a person at the firm approving it, unless the firm has decided a stage is safe to automate.

If you are a firm's employee, client or contact and want to know what it holds about you, ask the firm first; we will help it answer.

Training a firm's model

We train a copy of an open model (one its publisher has released for anyone to download and adapt) on the past work a firm chooses, so its AI Experts draft the way it writes. Downloading the open model sends none of the firm's information to its publisher.

Before any of a firm's work is used for training, we anonymise it: names, contact details and anything else that identifies a person, such as the firm's clients and their contacts, are removed or replaced. The model learns how the firm writes, not who it writes about.
Training runs on GPUs we rent from RunPod. The firm's work is copied to them for the run and removed when it finishes; the trained model is stored there so it can run for the firm's AI Experts.
A firm's trained model is used only for that firm. It is never used for another client, and a firm's work is never used to train a model for anyone else or given to a model's publisher.
A trained model can repeat wording from what it was trained on, so people at the firm who use its AI Experts may see passages drawn from the anonymised training set. The firm decides what goes in with that in mind.

Indexing a firm's library

We collect the past work a firm chooses from the systems it connects, read the text out of each document, tag it, split it into passages and index those passages for search. When an AI Expert works on something, only the passages relevant to it are retrieved from the firm's own library and given to the firm's model, and each draft cites the documents it drew from.

New documents in the connected folders and systems are indexed as they arrive. Each firm's library is kept in its own records in our database, separated from every other firm's by row-level security.

Why we use it

To reply to your enquiry, hold the call you booked and send the proposal you asked for. Legal basis: our legitimate interest in responding to people who contact us, and taking steps at your request before a contract.
To train a firm's model, index its library, run its AI Experts, support it and bill for it. Legal basis: performing our contract with the firm, and processing its work on its instructions.
To keep the service secure, prevent misuse and fix faults. Legal basis: our legitimate interest in running a safe service.
To know which of our posts and adverts bring enquiries, from the link tags described below. Legal basis: our legitimate interest in marketing responsibly.
To keep records the law requires, such as invoices. Legal basis: legal obligation.

We will only send you marketing email if you have asked for it or are a client, and every such email lets you stop them.

Cookies and local storage

We use only what the site needs to work. There are no advertising or cross-site tracking cookies, so there is no cookie banner.

Sign-in cookies, set when a client signs in, which keep you signed in. They are removed when you sign out.
Your light or dark theme choice, stored in your browser.
Link tags: if you arrive from a link that says where it came from (for example a post or an advert), your browser stores that link's tags and the referring site for up to thirty days. It holds no personal information, stays on your device, and is sent to us only if you submit the contact form.

We use a small number of providers to run the site and the service. Each handles information only to provide its service to us, under a written agreement.

RunPod: the GPUs that train each firm's model and run it for the firm's AI Experts.
Supabase: the database, sign-in and file storage, including each firm's library.
Vercel: hosting for this website.
Railway: hosting for the service that connects to a firm's systems and runs its AI Experts.
Voyage AI: turning library text into search indexes, so the right passages can be found.
Reducto: reading the text out of PDFs and scanned documents.
Resend: sending email, such as notifications of new enquiries and sign-in links.
Calendly: booking calls.

If a firm also chooses to draft with a commercial AI model provider (currently Anthropic or OpenAI), that runs under the firm's own API key and its own agreement with the provider. Those providers' commercial terms do not allow them to train on the content sent to them.

The systems a firm connects are the firm's own accounts, governed by the firm's agreements with them.

We may disclose information if the law requires it, or to a buyer of our business, who would be bound by this notice. We do not sell personal information.

A current list of providers, and our data processing agreement, are available to clients on request from hello@agenticintegrations.app.

Where it is processed

We are based in the United Kingdom and work with firms in other countries, so information is processed in the UK, in the firm's own country and by providers who may process it elsewhere, including the GPUs a firm's model runs on. Where personal information from the UK or the European Economic Area is transferred to a country without an adequacy decision, we rely on the UK International Data Transfer Addendum or the European Commission's standard contractual clauses, or the UK Extension to the EU-US Data Privacy Framework where the provider is certified.

How long we keep it

Enquiries and call notes: for as long as we are in conversation with you and up to two years afterwards, unless you become a client.
A client firm's work, library, trained model and account: for as long as the firm is a client. When it leaves we give it a full export on request, then delete its work, library, trained model and accounts within 30 days, unless the firm asks us to hold them for a legal reason. Backups roll off on their normal cycle after that.
Copies of a firm's work made for a training run: deleted when the run finishes.
Invoices and contracts: for as long as tax and company law requires, usually six years.

How we protect it

Everything is encrypted in transit and at rest. API keys and connection tokens are held in an encrypted vault that only our service can read. Access to production systems, including the GPUs that hold a firm's model, is limited to the people who need it and is logged. If an incident affects a client's data, we tell the client in writing within 72 hours of confirming it.

Your rights

You can ask us to show you, correct, delete or send you a copy of the information we hold about you, to stop using it for a purpose, or to object to how we use it, at any time, by emailing hello@agenticintegrations.app. We answer within a month and do not charge for it. For work a client firm holds about you, we will pass your request to that firm and help it answer.

If you are in the UK you can complain to the Information Commissioner's Office (ico.org.uk). If you are in the European Economic Area you can complain to your local data protection authority.

If you live in a US state with a consumer privacy law, such as California, you have similar rights to know, correct and delete your information, and not to be treated differently for using them. We do not sell or share personal information for cross-context advertising, and we do not use sensitive personal information to infer anything about you. You can make a request, or have an authorized agent make one, at the same address; we will verify it before acting.

The site and the service are for businesses. They are not directed at children and we do not knowingly collect information about anyone under 16.

Changes and contact

If we change this notice we update the date at the top, and tell clients by email before a change that affects their data takes effect. Questions go to hello@agenticintegrations.app.