The terms we work on

Last updated September 29, 2026

These terms are between Agentic Integrations ("we", "us"), a company based in the United Kingdom, and you. If you are browsing the site or booking a call, only the sections on the website apply. If your firm becomes a client, the rest apply to your firm too, alongside the proposal or order form you sign with us. Where the signed proposal says something different, the signed proposal wins.

The service is sold only to businesses. By accepting these terms for a firm, you confirm you have the authority to do so.

Using the website

You may use this website to learn about our service, contact us and, if your firm is a client, sign in. Please do not try to break it, probe it for weaknesses without our written permission, scrape it at volume or use the contact form to send anything other than a genuine enquiry.

The information on the site is a description of our service, not an offer. What we will do for your firm, and the price, is set out in the written proposal we send you. A call with us is free and places you under no obligation.

What we provide

We set up AI Experts for your firm: AI agents that each own one stage of one of your processes, draft that work inside the Officeware your firm already uses, and hand it to a person at your firm to approve. Two things sit underneath every AI Expert, and your firm pays for both:

A model trained on your firm's own past work, so it drafts the way your firm writes. How we do that is under “Training your model” below.
A library of your firm's past work, indexed so an AI Expert can find and cite the right documents for each piece of work. How we build it is under “Indexing your library” below.

We connect the AI Experts to the systems listed in your proposal, configure them and run them for you. Your proposal names the processes and stages each AI Expert covers.

The service does not do these things, and you should not rely on it to:

Send, sign, file or submit anything without a person at your firm approving it, unless your firm has told us in writing that a stage is safe to automate
Give legal, financial, tax or other professional advice, or make the judgements your firm's professionals are responsible for
Guarantee any outcome from the work it drafts

Training your model

We do not build AI models from scratch, and we do not send your firm's work to a model owned by someone else to learn from. Instead:

We choose an open model: one whose publisher has released it for anyone to download, run and adapt under an open licence that allows commercial use. Your proposal names the model we start from and its licence.
We anonymise the past work your firm has chosen to give us: names, contact details and anything else that identifies a person, such as your firm's clients and their contacts, are removed or replaced.
We download a copy of the open model and train that copy on the anonymised work, so it learns your firm's structure, terminology and style without learning who your firm works for.
We test the trained model on examples from your own work before any AI Expert uses it, and tell you what we found.
The trained model runs on graphics processors (GPUs) we rent from RunPod, for your firm alone.

The model trained on your firm's work is used only for your firm. We never use it for another client, never use your work to train a model for anyone else, and never give your work back to the model's publisher.

A trained model can reproduce what it was trained on, so anyone at your firm who can use an AI Expert may see wording drawn from any document in its anonymised training set. Leave out anything that not everyone who will use the AI Experts should see, and tell us if part of your work must be kept to a smaller group; we can train a separate model for it.

We retrain your model when your firm asks, for example to learn from new work, and when your proposal schedules it; the GPU time for each run is billed as described under “Fees and payment”.

Indexing your library

Your library is the searchable record of your firm's past work that the AI Experts draft from. To build it we:

Collect the past work your firm chooses from the places it already keeps it (the file stores, email, CRM and other systems listed in your proposal), or from files your firm sends us.
Read the text out of every document, including PDFs and scanned pages.
Tag each document with what it is, who it was for, when it was written and the other details your proposal lists, and split it into passages.
Index every passage for search, so an AI Expert can find the passages that matter for the piece of work in front of it and cite the document each one came from.

Once the library is built, new documents added to the connected folders and systems are indexed as they land, so it stays current without anyone uploading anything. Search indexing is included in the fee and is not billed as GPU time.

Your firm's library is kept separate from every other firm's and is never searched on anyone else's behalf.

Fees and payment

Training and indexing: $40,000 one-time, half on signing and half when your trained model and library go live. It covers choosing the open model, preparing your work for training, training and testing your model, indexing your library, and connecting and configuring your AI Experts.
GPU time: every hour a GPU runs for your firm on RunPod, including training and retraining your model and keeping it running for your AI Experts, is billed at RunPod's published rate for that GPU at the time it ran, multiplied by 1.5. It is measured the way RunPod measures it and billed monthly in arrears.

Each GPU time invoice lists every run: the GPU type, the hours, the rate it ran at and what it was for. While a model is kept running so an AI Expert can answer at any time, that time counts even when no work is being drafted. Your proposal says how your model will be run, and we will agree any change with you before making it.

Prices are in US dollars and exclude any sales, use, value added or similar taxes, which are added where they apply. Invoices are due within 30 days unless your proposal says otherwise. If an invoice is more than 30 days overdue we may, after telling you in writing and giving you 14 days to pay, pause the service until it is paid; your data and your trained model are kept safe while it is paused.

There are no per-seat fees.

Your firm's responsibilities

Review every draft before it is used. An AI Expert hands its work to a person at your firm, and your firm is responsible for what it sends, signs, files and commits to.
Only give us work your firm has the right to use for training and indexing, and have the permissions needed for any personal information in it.
Decide what goes into the training set, knowing that anyone who uses the AI Experts may see wording drawn from it.
Give us the access and documents set out in your proposal, and keep your people's sign-ins to your own systems secure.
Use the service only for your firm's own work, lawfully, and not to build a competing product, resell it, or try to get around its security or another firm's separation.

AI drafts

Drafts are produced by AI from your firm's own model and library. Every answer cites the document it came from so it can be checked, and flags anything it could not support. AI can still get things wrong: it can misread a request, draw on an out-of-date document or repeat a mistake from your past work. Drafts are a starting point for your team, not advice, and we do not promise any particular result.

Who owns what

Your firm owns its data: the work it gives us, its library, the drafts produced for it and every export. You give us permission to store and process that data only to train your firm's model, index its library, provide the service to your firm and support it. We never use it to train a model for anyone else or to draft for anyone else.

The model trained on your firm's work is held by us for your firm and used only for your firm. The open model it started from stays under its publisher's licence, and your firm's use of the trained model must follow that licence; your proposal names it.

We own the service itself: the software, the training and indexing methods, the AI Experts' configuration and the improvements we make, including improvements that come from feedback you give us. Your firm has the right to use it while it is a client. Nothing in these terms transfers our software or code to your firm.

Other companies' services

The service works inside and alongside services from other companies: the Officeware your firm connects, RunPod for GPUs, the publishers of open models and the providers listed in our privacy notice. Your firm's use of its own accounts is governed by its own agreements with those companies. We are not responsible for their availability or for changes they make, including changes to RunPod's rates, but when one affects the service we will tell you and work to restore it.

Confidentiality and data protection

Each of us will keep the other's confidential information private, use it only for our work together and protect it at least as carefully as our own. This does not cover information that is already public, was already known, or must be disclosed by law.

How we handle personal information is set out in our privacy notice. Where we process personal information on your firm's behalf, our data processing agreement applies and forms part of these terms.

Availability and support

We host, monitor and update the service and aim to keep it available at all times, but it may occasionally be unavailable for maintenance, because GPUs are not available from RunPod, or for other reasons outside our control. We plan maintenance for outside your business hours where we can. Support is by email and your firm's direct line to the team that runs your account.

We improve the service over time. We will not remove something your firm has paid for without agreeing it with you.

Cancelling and leaving

After go-live there is no minimum term. Either of us can end the agreement with one month's written notice. Either of us can also end it straight away if the other seriously breaches these terms and does not put it right within 30 days of being told, or becomes insolvent.

When the agreement ends, the AI Experts stop, your model stops running and no further GPU time is billed. Fees and GPU time already due remain payable, and the one-time fee is not refunded. On request we give your firm a full export of its data, then delete its data, library and trained model as described in our privacy notice. You should disconnect the services your firm connected; we will send you the steps.

Warranties and liability

We will provide the service with reasonable skill and care. Apart from what these terms and your proposal say, the service is provided as it is, and we make no other promises, including that it or its drafts will be free of errors.

Neither of us is liable to the other for loss of profit, revenue or contracts, or any indirect or consequential loss. Each party's total liability under these terms in any 12 months is limited to the fees your firm paid us in the 12 months before the claim.

Nothing in these terms limits liability for death or personal injury caused by negligence, for fraud, for your firm's obligation to pay fees, or for anything else that cannot be limited by law.

General

If we change these terms we update the date at the top. Changes that affect clients are sent by email at least 30 days before they apply to your firm, and do not apply to a period already paid for unless you agree.

Neither of us is responsible for delays caused by events outside our reasonable control. Your firm may not transfer the agreement without our consent; we may transfer it to a buyer of our business, who would be bound by it. If any part of these terms is found unenforceable, the rest still applies.

These terms are governed by the law of England and Wales, and the courts of England and Wales have exclusive jurisdiction.

Questions about these terms go to hello@agenticintegrations.app.