Breeze
The agent runtime. Breeze plans multi-step work, assigns it to specialised agents and keeps state across hours or days, including retries and approvals.
- Graph and loop orchestration
- Human approval steps
- Durable memory per case
Pier AI builds autonomous agents for operations teams. They plan the work, carry it out across your systems and hand off cleanly to the next agent or to a person. Built in Stockholm, running in your cloud.
Most agent projects stall between the demo and production. Pier gives you the runtime, the gateway, the monitoring and the deployment model in one stack, so the agent you tested on Friday is the one running on Monday.
The agent runtime. Breeze plans multi-step work, assigns it to specialised agents and keeps state across hours or days, including retries and approvals.
The model gateway. Gust routes every call to the cheapest model that can do the job, caches what it can and keeps prompts and data inside your tenant.
Observability for agents. Beacon traces every decision and tool call, and flags unusual runs before your customer notices.
Private deployment. Harbor installs the whole stack in your own Azure subscription, with your identity provider, your network rules and your keys.
Every agent follows the same four steps. You decide where a person signs off, and Beacon records each step so you can see exactly why the agent did what it did.
Picks up a new case from email, a ticket queue, a form or an API event, and gathers the context it needs from your systems.
Breaks the case into steps, checks them against your policies and asks for approval when the stakes are high.
Carries out the steps with the tools you allow: your ERP, CRM, SharePoint, Dataverse or any REST API.
Passes the result to the next agent or a person with a clear summary and closes the case. Nothing is left half done.
A supplier invoice arrives at 07:42. Here is what four Pier agents do with it before anyone in finance has had coffee.
We have shipped agents into finance, logistics and public sector teams across the Nordics. Three things have held true every time.
An agent improves fastest in front of real cases. We put a narrow version into production in weeks, then widen what it is allowed to do.
Gust sends simple calls to small models and big problems to big ones. Our customers cut inference spend and energy use by more than half.
Every decision an agent makes is logged in plain language. If it can't explain a step, it doesn't take it.
Our headcount is flat and our agent count is not. Every new role at Pier AI starts as an agent, built and tested on our own work before it ever reaches a customer.
Tell us about one process you would like an agent to take off your hands. We reply within one working day.