Skip to content

ESEN

Let's talk
Menu

Your data stays in-house.

Lindero runs on your server and works with your systems under your rules. By default, the AI does not see your figures.

Let's talk How the AI is governed
LOCALLever on cloud: only the structure of your data goes out through the slot, not a single row.
Envelope: the structure (table and column names, types, descriptions)Behind the window: your rows, which stay inLOCAL lever: the model runs on your serverBy default, the cloud AI only receives the structure (table and column names, types, descriptions) and what your people ask; not a single row. Your figures, documents and emails leave only if you switch that on. With the lever on LOCAL, the model runs on your server and nothing goes to a cloud AI.

How we look after your data

  • It runs on your server.

    Next to your ERP. What it syncs from your data never leaves it.

  • Reading and writing take separate paths.

    It reads with an account that cannot write. Anything that changes your ERP goes through actions that are rehearsed, approved and logged.

  • You decide what the AI sees.

    By default, it does not see your figures. You decide whether it does.

  • Everyone sees their own department.

    Someone looking at Purchasing sees Purchasing. As far as they are concerned, the Finance accounts do not exist.

  • Everything is logged.

    You know who asked what, and when. Passwords are never stored in any log.

Every AI, governed. Down to each piece of data.

Lindero does more than log what the AI does. Because it understands your business, you know which data each AI touches: not tables, but suppliers, salaries or sick leave.

You mark what is sensitive, Lindero suggests it to you, and each AI's record sums it up like this.

Purchasing Expert Made-up example
Can read
SupplierPurchase orderDelivery note
Does not see
Employee
Sees
Amounts
If an AI is about to read specially protected data, such as your people's health, it warns you or asks for permission. You choose.

  • Every AI, registered automatically.

    The chat, each digital expert and each automated task that uses AI has its own record: what it does, which model it uses and what it can read. All you add is who is accountable for it.

  • Fixed rules watch it, not another AI.

    If an AI reads outside its department, it is refused. If an automated task runs out of control, it is paused and you are told. An email from an AI to someone outside waits for a person's approval.

  • Hidden instructions, brought into the open.

    Lindero looks for instructions hidden for the AI in documents and emails. When it finds one, it flags the passage and holds back whatever the AI wants to do based on it.

  • Every incident, logged until it is closed.

    A rule that fires, or an answer someone marks as wrong, opens an incident on its own. It gets an owner and is closed with a note.

  • An auditor who sees everything and changes nothing.

    Your compliance officer sees the activity, each AI's record and its decisions. They cannot approve anything or change a rule.

  • The EU AI Act, obligation by obligation.

    You see what Lindero does for each article and what falls to you, with the evidence from the last 30 days, plus a dossier for your auditor. Helps you meet the EU AI Act; it is not a certification.

  • At least six months of records.

    What each AI does is kept for at least six months, as the law requires of those who use it. Lindero does not delete it on its own.

  • Your people learn to use it with judgement.

    A short guide: what the AI does, its limits and how to check a figure. And you know who has read it.

On trial first. Then into production.

Anything new built with Lindero, whether an app, an automated task or an agent, is tested on your real data before it writes to your systems.

ERPIn testing: the bridge is pulled back; every action drops into the proposals tray and nothing reaches the ERP.
Boxes: an AI's actionsSix plungers: the fixed rules W1–W6Paper tape: the log the auditor readsEvery action an AI takes goes through the same gate. Inside, fixed rules, not another model, check it; if one trips, the action is stopped and an incident is opened. While it is in testing it does not write to your ERP: its actions stay as proposals. It goes to production when a person who did not build it approves it. Everything stays on the log tape, which the auditor reads and cannot change. In the figure, a rule trips every five actions so you can see it.

  1. Draft

    It is being built. It writes to no system.

  2. On trial

    Does not write to the ERP

    It works with your real data, but it neither writes to your ERP nor sends emails. An agent proposes instead of acting, and Lindero compares what it would have done with what your people did.

  3. In production

    Once a person who did not build it approves it, with the evidence in front of them: days without incidents and, for an agent, how often it matches your people.

In a hurry? Only an administrator who did not build it can skip the trial, with a reason, and that is logged.

What leaves your server and what does not

Leaves, towards the AI you choose

  • How your data is organised, without your figures.
  • The questions your people type.
  • Your figures, documents and emails, only if you switch it on.

Stays on your server

  • The passwords to your systems.
  • What Lindero syncs from your data, and the log of what happens.

Would you rather nothing reached a cloud AI? Install the AI on your server.

Are you the IT manager?

We give you all the technical documentation, including what is still to be resolved.

Let's talk