Una AI Launches Orchestrator and AI Governance for FP&A Teams
Key Takeaways
- •Una AI Orchestrator coordinates modeling, data, and workflow agents for multi-step planning tasks.
- •Finance teams must approve Orchestrator’s proposed changes before they are applied to a plan.
- •Una AI Governance can control AI access and actions for both Una AI and external MCP-connected applications.
- •Administrators can use policies, tool permissions, and Data Zones to restrict where AI operates, what it does, and what data it can view.
- •An exportable activity log records each AI action, its user, outcome, and the reason for any denial.

Una AI, the AI workspace for finance, has announced the general availability of Una AI Orchestrator and Una AI Governance at the Finance & Accounting Technology Expo (FATE). Introduced together, the two products allow financial planning and analysis (FP&A) teams to hand multi-step planning work to AI agents, while finance teams retain control over what the AI can access and change.
"The more an agent can do, the more it matters who controls it," said Michael Morrison, CEO of Una AI. "An agent that builds your model and loads your actuals has to answer to the people who sign off on the numbers. They decide what it can change, and they can see everything it did. That is why we are shipping Orchestrator and Governance together."
One Request, Several Agents, One Plan
With Una AI Orchestrator, a user can ask Una, the in-app AI agent, to turn an Excel budget into a planning model, load actuals from the company's ERP, and set up the review workflow. Una passes the work from a modeling agent to a data agent and then to a workflow agent, returning a single set of proposed changes that shows each step. Those changes are applied to the plan only after finance approves them. That sequence — model creation, actuals load, and review setup — typically spans several manual steps for FP&A teams, which is why the step-by-step, approval-gated design sits at the core of how Orchestrator works.
Finance Decides What AI Can Access and Change
For finance leaders adopting AI, the concern is accountability. Una AI Governance gives administrators a single place to address that concern. It controls what AI can access and change within Una AI, whether requests come from the in-app agent or from connected external AI apps such as Claude and ChatGPT, which connect to Una AI through the Model Context Protocol (MCP), an open standard for linking AI apps to business systems. That scope matters in practice: once external AI apps can reach a planning platform through a protocol like MCP, the platform's own permission layer becomes the enforcement point for what those apps can see or change.
Three areas set permissions, and a fourth provides the audit trail:
- Policy sets where AI runs: inside Una AI, through external AI apps, both, or neither. Administrators can make AI read-only with a single switch.
- Tool permissions determine what AI can do. Administrators approve or block each type of AI action, such as running queries or changing scenarios.
- Data Zones determine what AI can see. Administrators can keep sensitive data, such as HR detail, hidden from AI even when the user is allowed to see it.
- Activity records what AI has done. Every AI action is logged with the user, the action, and the outcome, including why a request was denied, and the log can be exported for audit.
According to the announcement, AI never sees more than the user who launched it, and it cannot change its own permissions. The per-action permissions and exportable log map to the questions audits ask of any system change — who acted, on what, and why it was allowed or denied — here extended to AI actions.
What to watch next is how these controls get configured in practice: whether administrators start with AI in read-only mode, which action types they approve, and how widely they connect external AI apps over MCP — choices that determine how much execution authority finance teams delegate to agents.
Additional information is available on the Una AI website at una.ai.
Source: Global FinTech Series