Security and management

Using AI safely inside your own company.

Where your data sits, who can reach it, what an agent may do by itself and what happens when something goes wrong. All in one place, also for whoever was forwarded this link.

Reply to a report
within 4 hours
Hosting
AgentWorks or n8n

Where your data goes.

From your own systems to you checking it. Per step you see what happens and which agreement goes with it.

  1. 1

    Your own systems

    Data stays where it is

    Your mailbox, bookkeeping or CRM. The agent reads only what it needs for its task.

    Access per agent recorded in the AI agent design

  2. 2

    AgentWorks or n8n

    The agent does the work

    With UnifyAI by default, in your own environment by exception. Everything it does is logged.

    • AgentWorks
    • n8n

    Data processing agreement and SLA, signed before the start

  3. 3

    The language model

    AI reads and writes along

    The model may run outside the EU. It reads only what the agent sends it and does not train on your data.

    Transfer described in the data processing agreement

  4. 4

    You

    You check what matters

    An email to a customer or a booking only goes through once a person has looked at it.

    What the agent may do by itself, we agree per agent

People, data and ownership.

Agreements that hold in every project, with every agent.

  • A person always looks along

    The agent prepares, you check. What leaves the building, such as an email to a customer or an entry in your books, only goes through after a person has checked it.

  • Your data stays yours

    The data an agent uses stays your data. We use it only for the work set out in the agreements.

  • What we build is yours

    The agent, the connections and the documentation are fully transferable. Stopping with us does not mean starting over.

  • Only the access that is needed

    An agent gets access to the systems and data it needs for its task, and no more. That is recorded per agent in the AI agent design.

GDPR and the EU AI Act.

No certificate on the wall, but agreements in how we work. From the first day of a project.

  • Data processing agreement

    If we process personal data for you, we set that out in a data processing agreement: which data, what for, how long and who else is involved.

  • Keeping and deleting

    Which data an agent keeps and for how long, we record in the agreements with you. What is no longer needed goes out.

  • Visibly made by AI

    Anyone talking to an AI can see it. What the UnifyAI Agent makes carries the label “Made by the UnifyAI Agent”. The EU AI Act asks for that transparency.

  • Governance in the team

    The AI governance & legal role watches the GDPR, the EU AI Act and your own rules on what is allowed, per project.

What the UnifyAI Agent does with your data.

The agent on this site is an AI. This is how it handles what you tell it.

  • Which model we use

    The UnifyAI Agent works with Claude, the language model by Anthropic. Processing may therefore happen outside the EU. We arrange that with a data processing agreement, and the privacy statement describes the transfer.

  • Anonymous until you save something

    At the start you see that the conversation is stored. You stay anonymous until you give your name and email to save something. At that moment we ask explicitly. The PDF of your AI scan comes only after that. What you fill in on the scorecard we keep with your conversation, so we can discuss it in an intro call.

  • No access to our database

    The language model cannot read or write in our database itself. Every write goes through a route on our own server that checks what is allowed.

  • No training of external models

    We use conversations anonymised to improve the UnifyAI Agent itself. They are not used to train other parties’ models.

  • Bounded

    Detection of misuse and prompt injection, a limit per session and a cost ceiling. The agent only fetches public web pages, never internal addresses.

De volledige afspraken staan in de privacyverklaring. Lees de privacyverklaring

Management after delivery.

An agent that is live needs looking after. Hosting keeps it running, maintenance keeps it up to date.

Hosting

Verplicht

€ 49

per maand, per agent, excl. btw

  • AgentWorks
  • n8n
  • Draait dag en nacht, gekoppeld aan je eigen systemen
  • Monitoring: we zien het als hij stilvalt
  • Kostenbewaking van het verbruik
  • Support met reactie binnen 4 uur, op kantooruren
  • Afspraken vastgelegd in de SLA

Onderhoud

Optioneel

vanaf € 100

per maand, per agent, excl. btw

  • Bijwerken als je pakket of een koppeling verandert
  • Bijsturen waar de agent vaker fouten maakt
  • Nieuwe regels of uitzonderingen toevoegen
  • Periodiek nalopen of de agent nog doet wat hij moet doen

Zonder onderhoud draait de agent gewoon door. Verandert er iets in je pakket of koppeling, dan passen wij hem niet vanzelf aan: dat doen we dan los, op offerte.

Prijzen en wat erbij hoort. Hosting en onderhoud

Questions about security and management.

Where does our data sit?

By default the agent runs with UnifyAI, on AgentWorks or n8n. Only by exception in your own environment. Per agent we record which systems it uses and which data it keeps. The agreements are in the privacy statement.

Is our data processed outside the EU?

It may be, because we pick the language model that is best for the task. We arrange that with a data processing agreement and describe the transfer in the privacy statement. So we do not claim everything stays inside the EU.

Can an agent send or book something by itself?

What matters, the agent prepares and a person checks. Which actions those are, we agree per agent in the AI agent design. What that checking looks like is on the AI agent project.

How do you deal with the EU AI Act?

This is how we work: a person in the loop, visible when something is made by AI, and agreements on what an agent may do. The AI governance & legal role watches that per project. We do not give legal advice for your situation.

Who manages the agent after delivery?

We do, with hosting and maintenance. The agent is fully transferable, and by exception it can run in your own environment.

What if something goes wrong?

You report it and get a reply within 4 hours, during office hours. With maintenance we then review the agent and adjust it.

Which systems do you connect with?

Exact Online, AFAS, HubSpot, SnelStart, e-Boekhouden and Salesforce, among others. Per package you can read on integrations what is possible and how we secure the connection.

DaarnaWerkwijzeDe route, de specialisten en je klantomgeving.
Erwin Berkouwer

Erwin Berkouwer

AI consultant and architect, your single point of contact

Book an intro call.

30 minutes to an hour, online or by phone. Within 2 working days a proposal is ready in your personal environment.

Loading calendar