Company

Building the confidence behind enterprise AI.

We believe the future of enterprise AI will be defined not only by what agents can do, but by how reliably they do it.

Our evolution

The same question, with higher stakes.

Arman Kayhan and Mert Menekşe founded Co-one in 2021. The company began with data labeling and quality work that helped enterprise teams build AI systems on more dependable data.

That work moved into human feedback and evaluation. In customer-facing AI, it meant understanding which requests a system misunderstood, how responses changed across versions, and where people needed to step in.

Now agents can update records, call services, and move work across systems. We build tools to simulate those workflows, evaluate behavior against business requirements, and apply controls at the production integration points teams manage.

Explore our published enterprise work

The work behind the platform

Where research meets implementation.

Our work spans AI evaluation research, agent simulation, enterprise software, and deployment. The hard questions cross those disciplines. Which workflow should we reproduce? Which evidence shows that the agent followed the rule? Where can an integration enforce policy?

See the engineering in the platform
AI LAB · Log ExplorerProduct capture
AI LAB Evaluation Log Explorer with version and score filters and demo rows showing token usage and reliability scores.

Company milestones

From data quality to agent reliability.

As AI systems took on more responsibility, our work shifted from improving the inputs to examining what the systems did with them.

  1. Co-one OÜ registered in Estonia

    The founders started with data labeling and quality for AI teams.

    View the Estonian registration
  2. Distinguishing customer intents

    Co-one published its work with Türkiye İş Bankası on labeling customer queries and distinguishing similar intents for Maxi.

    Read the Maxi case study
  3. Forbes 30 Under 30 Europe

    Arman Kayhan and Mert Menekşe were named to the 2024 Technology list for their work building Co-one.

    Read the announcement about Forbes 30 Under 30 Europe
  4. Pre-Series A announcement

    Co-one announced a €1 million round led by Ak Portföy as it expanded its work on enterprise AI.

    Read the announcement about Pre-Series A announcement
  5. Estonian applied research funding

    Co-one announced funding through the Estonian Business and Innovation Agency’s programme to advance evaluation and uncertainty-aware AI research.

    Read the announcement about Estonian applied research funding

Engineering philosophy

We believe reliability must be demonstrated.

Evidence over assumptions
Inspect the agent’s actions and the state it leaves behind. A persuasive response can still accompany an incorrect result.
Realistic environments
Use stateful workflows and simulated tools to test against the requirements that matter in real operations.
Enterprise control
Design for the deployment boundary. Decide what data, credentials, and traffic each integration is allowed to handle.
Built for action-taking AI
Follow a tool call through to its consequences for approvals, records, and the people who rely on them.

Start a conversation

Let’s build AI enterprises can depend on.

Meet the team behind Co-one and explore how we can help bring reliable AI into your organization.

Talk to Co-one

Choose whether to allow optional analytics and marketing tags.