AI agents

    An AI agent for described work, overseen by your team

    We agree with you the agent’s task, its access to data, what it may do on its own and where a person decides.

    Book a call →

    30 minutes, free.

    Prefer to write? Send a message

    Illustrative example · fictional companyAn agent checks an invoice against the order and the contract terms

    Example: an agent checks a supplier invoice against the order and the contract terms. Product and quantity match, but the invoice price is PLN 2.40 instead of PLN 2.15, so the agent passes the invoice to the accountant with a note of the PLN 125 difference. When everything matches, it marks the invoice as checked. When the order is missing, it also passes the invoice on for checking. The agent does not pay or approve invoices.

    The agent compares documents and describes the differences. It does not pay or approve invoices; the accountant decides.

    What we do

    • We describe the task: what the agent does and what it does not do.
    • We agree access to data and permissions: what the agent only reads and what it may do on its own.
    • We set the points where a person decides, and how the agent hands over cases it is not sure about.
    • We test the agent on real test data from your company and estimate the cost of using it.

    What you get

    • A description of the agent's task: data, permissions, where a person decides and the estimated usage cost.
    • An agent working in the agreed process, on the systems you use.
    • Test results: where the agent works well and where it hands a case to people.

    How we start

    1. Initial call

      30 minutes, free

      You tell us what you want to improve or build. We say plainly whether we can help and where to start.

    2. Business Discovery

      PLN 5,000 net

      The default first stage of the work, about 1–2 weeks from receiving the agreed access. NDA before access, payment on completion. You get a report, recommendations and a plan for the next steps; three months of basic Steering Center to review the report are included.

    3. Implementation

      Scope and price agreed separately

      If we go further, we agree the scope and price. Discovery carries no obligation to continue; exceptions to this order are agreed case by case.

    With agents, Discovery helps when it is not yet clear which work is worth handing to one.

    Questions about this service

    What if the agent makes a mistake?
    It can make mistakes, so it has a narrow task, limited permissions and points where a person decides. Before rollout we test it with your team on real test data.
    How much does it cost to run the agent?
    The usage cost depends on the number of cases and the model chosen. We estimate it before rollout; the implementation itself is priced individually.
    What data does the agent have access to?
    Only the data its task needs, within the scope agreed with you before rollout.
    How is an agent different from automation?
    Automation runs fixed steps by rules. An agent helps where content has to be read and compared in a way a rule cannot describe. They often work together.

    Let's talk about the work you want to hand to an agent

    Describe the task and the data your team works with. We'll check whether an agent makes sense here.

    Tomasz Pijanowskifounder of ToruAI

    Book a call →

    30 minutes, free.

    Prefer to write? Send a message