Datacoves launches Atlas for governed AI definitions

Datacoves on Sept. 15 launched Atlas, an agreement layer for enterprise AI that turns team and agent conversations into a governed Definition of Record. The product is designed to help companies keep data meanings consistent across their existing stack, AI agents and security requirements.

Why it matters: - Atlas is meant to reduce one of enterprise AI's biggest failure points: different teams and agents using different definitions for the same metric. - The product aims to turn agreed-upon definitions into a reusable company asset instead of a one-off answer that has to be rebuilt each time. - Datacoves says Atlas can help organizations keep AI outputs tied to governance, ownership and existing data infrastructure.

What happened: - Datacoves announced Atlas on Sept. 15, 2026. - Atlas is an agreement layer that helps enterprise teams and their AI agents settle what data means before answering questions with it. - The launch coincides with dbt Summit 2026, held Sept. 15-18 in Las Vegas, where Datacoves is a Silver sponsor and the Atlas team is at booth 208.

The details: - Atlas scopes a question before answering it. - Atlas surfaces assumptions that can change the result. - Atlas routes open questions to the person or team that owns the answer. - The output is a plain-English Definition of Record that is versioned and attributed. - Atlas is built to connect with a company's warehouse, dbt transformation code, Git repository and orchestration through the open Model Context Protocol, or MCP. - Datacoves says this approach keeps definitions portable instead of locked to one vendor. - Because Atlas exposes its Definition of Record over MCP, existing AI tools can build against agreed business definitions instead of re-guessing them in each prompt. - Atlas can run entirely inside a customer's own network for organizations that cannot send sensitive data to an outside cloud. - Datacoves also offers Atlas as a managed service for teams without that constraint. - Datacoves co-founder Noel Gomez said Atlas is where a team agrees on what is true for the organization and keeps every answer tied to that agreement instead of drifting from it. - Datacoves says more information is available at datacoves.ai. - Media inquiries can be sent to media@datacoves.com.

Between the lines: - The launch targets a growing enterprise problem: AI can produce confident answers without resolving underlying disagreements about data definitions. - By building around MCP and existing data tools, Datacoves is positioning Atlas as an interoperability layer rather than a closed AI product. - The on-premises option suggests Datacoves is also aiming at regulated or security-sensitive buyers.

What's next: - Datacoves is using dbt Summit 2026 to demo Atlas in person. - The company is asking enterprise teams to evaluate Atlas against their current warehouse, dbt and AI workflows. - Datacoves says teams can learn more at the company's website or at booth 208 during the event.

The bottom line: - Atlas is Datacoves' bet that enterprise AI needs governed definitions before it needs faster answers.