Skip to main content

Legal

AI Transparency Notice

Version 2026-10-04-v1.1 · effective 2026-10-04

Where Outturn uses automation and AI, what it never does, and how people stay in control.

1.Our principles

  • AI never sets an authoritative figure. Amounts, quantities, dates and deadlines come from your data and the published check rules, each with its source.
  • Every action that changes data or leaves the platform (for example sending an e-mail or exporting a claim pack) is a proposal until a named person approves it.
  • You always know when you are dealing with AI.

2.Checks are rules, not AI

The checks that find discrepancies (for example storage days, weights, wagon hours or invoice lines) are deterministic rules. The same data always gives the same result, and every finding shows the records and the rule that produced it.

3.The built-in assistant

The assistant answers questions about your terminal's work and connected data. Its answers cite their sources, and figures in an answer are taken only from tool results over your data.

Which mode answers depends on your plan. On the Starter and Growth plans and during the free trial, once Outturn has switched it on, the assistant can use Claude, an AI model provided by Anthropic (listed in our Sub-processors), for open questions: the model receives your question and the data excerpts the assistant's tools retrieve for it, and its answers are labelled AI-generated. In every other case — another plan, the model not switched on, a monthly AI limit reached, or the model not answering — the assistant uses the knowledge-base mode (deterministic tools and the Outturn help) and the interface says so.

The assistant only handles questions about work processes and connected systems. Personal or off-topic requests are refused; they are recorded only as "out of scope", without their text.

4.AI agents used by Outturn support

Outturn's own team uses an AI agent (Claude, by Anthropic) to run the platform. It reads operational information about Workspaces — plan and trial dates, connector and sync status, error messages, billing status — never your ledger data, documents, findings, e-mail contents or passwords, and no contact e-mail addresses. It can only propose actions; a named Outturn person approves each one, and every proposal and decision is logged.

If you turn on Support access, those proposals may also cover your Workspace configuration (see the Terms of Service). Every change made that way appears in your staff & support access log as "Configured by Outturn support".

On the Starter and Growth plans and during the free trial, the digital workers may ask the AI model for column suggestions for a file they do not recognise. The model then sees only the column headers, never the values, and a person confirms every mapping.

5.EU AI Act

When the assistant uses an AI model, we meet the transparency duties of Article 50 of Regulation (EU) 2024/1689 (the AI Act): users are informed that they are interacting with an AI system and AI-generated content is marked as such.

Outturn is not designed for any use listed as high-risk in Annex III of the AI Act. It does not evaluate, monitor or rank individual workers and makes no decisions with legal or similarly significant effects on people.

6.Your data and AI

We do not use Customer Data to train AI models, our own or anyone else's. If an external model provider is enabled, it is bound by a data processing agreement that forbids such use.

7.Questions

Questions about AI in Outturn: [email protected].