AI applications and agents

When connections are not enough, we build something of your own: an assistant that knows your data and answers in plain language, an agent that does the repetitive work, or a portal and internal tool where AI is part of the daily process. We develop from design to publication and then keep it running. Agents get the minimum permissions per tool, written limits and human approval before any action with an external effect. We measure quality on a set of real cases before a solution goes into use. Our own products Imprion, FabriqAI and ConnectiCom are built the same way.

What we build

An assistant over your data

Click for more info

An assistant over your data

Answers employees' questions from ERP, CRM, documents and e-mail, with the source cited; in the browser, Teams or your own tool.

An AI agent for repetitive work

Click for more info

An AI agent for repetitive work

Reads an enquiry, drafts a quote, reconciles data between systems or prepares a report; actions with an external effect are approved by a person.

Portals and internal tools

Click for more info

Portals and internal tools

A web portal or internal tool with AI as part of the process: from design to publication and maintenance.

Taking over your solution

Click for more info

Taking over your solution

If you started a solution yourselves, we review it, secure it, publish it to production and take over maintenance.

Safety rules for agents

Click for more info

Safety rules for agents

Minimum permissions per tool, written limits, human approval before actions with an external effect and protection against prompt injection.

Measured quality

Click for more info

Measured quality

Every solution has a set of 50 to 200 real cases and a minimum accuracy it must reach before use; we keep measuring after publication.

How it works

  1. 01

    Design

    Together we define what the solution should do, where its limits are and how we will measure success.

  2. 02

    Prototype in a test environment

    Within a few weeks we build a working prototype on your data, separate from production, and try it with your users.

  3. 03

    Measurement and security review

    We check the solution on the set of real cases and with a security review; it goes to production once it reaches the agreed accuracy.

  4. 04

    Publication and maintenance

    We publish, roll out to the users and then take care of operation, updates and new requirements.

Frequently asked questions

What is the difference between an assistant and an agent?

An assistant answers questions and drafts; a person decides. An agent carries out a sequence of steps on its own within written limits, and a person approves actions with an external effect before they happen.

How do you make sure the AI does no harm?

The agent gets the minimum permissions per tool and written limits; actions with an external effect, such as sending a quote, are approved by a person first. The solution is protected against prompt injection and every step is logged.

How do you measure quality?

Before development we assemble a set of 50 to 200 real cases from your work and set a minimum accuracy. The solution goes into use only once it reaches it; after publication we keep measuring.

We started a solution ourselves. Can you take it over?

Yes. We review the code and permissions, add safety limits and logging, publish it to a separate production environment and take over maintenance.

How long does it take and what does it cost?

A prototype is usually ready within a few weeks; the scope and price follow the assessment and we prepare the offer after the introductory call. Payment is per completed step, with no advance.

Tell us what the solution should do

Describe the process and who runs it; after the introductory call we propose the design, the scope and the offer.