Hilde AI

A colleague that reads everything, and shows its work

Hilde lives in the workspace, sees the same data you do with the same permissions, and produces drafts a human approves.

What it actually does

Not a chat box beside the system, but an agent inside it — with the permissions and the audit trail that come with that.

Drafts in customer service
Replies written with the customer's orders, shipments and previous cases in mind — ready for review.
Suggestions in bookkeeping
Account, VAT code and accrual suggested, with the uncertain parts raised rather than hidden.
Discrepancies and follow-up
Finds the difference in the freight invoice, the missing receipt, the order that is stuck.
Questions about your own numbers
"Which customers dropped most in August?" answered against the database, with the query visible.
The same permissions as you
The agent cannot see more than the user it works for. The company boundary applies to it too.
Models in the EU
Runs on Vertex AI in a European region. Sending anything outside the EU requires a switch to be turned on.

Hilde writes, suggests and prepares. It does not email customers, does not post to the ledger without approval, and pays for nothing.

That line is not temporary caution — it is the design. An agent acting on its own inside a ledger is an agent nobody can audit.

Every run leaves a trail: what it read, what it suggested, what it changed and who approved it.

That is the difference between a tool you can rely on in production and a demo that impresses once.

We set the agent up in a test environment with your permissions.