Responsible AI

We build AI systems that read company documents, answer questions and take actions in business software. When they get something wrong, real people deal with the result. These are the commitments we work to on every project.

We build on existing models

We use foundation models from established providers and connect them to your data and workflows. We don't create or train our own models. For each system, we tell you which model and provider it uses and where your data goes when the model processes it.

We start with the problem

Every project starts with the process or product you want to improve. Some problems are better solved with a form, a report or a simple rule than with AI, and when that's the case we'll say so.

People approve sensitive actions

The agents we build ask a person before they do anything sensitive, such as paying money, changing records or contacting customers. We agree with you up front which actions need approval. Every action an agent takes is logged, so you can see what it did and why.

Answers you can check

Our knowledge assistants link each answer to the document it came from, so people can check it instead of taking it on trust. When an assistant can't find an answer in your documents, it says so rather than guessing.

Security testing before launch

We plan security tests when we scope a project, and run them during the build and again before launch. As well as standard application testing, we test AI features against attacks such as prompt injection, data leakage and misuse of the tools an agent can reach. We fix what we find before the system goes live.

Governance you can show an auditor

We set up AI governance that fits ISO/IEC 42001, the international standard for AI management systems. That includes a record of each AI system, its risks and the decisions made about it, so you have evidence ready when a customer or auditor asks.

You own what we build

At the end of a project we hand over the code, configuration and documentation. You can run, change or move the system without us.

Questions

If you have a question or concern about an AI system we built, email hello@ikarmic.com.