Following recent additions to Copilot Studio that ensure human judgment in decision-making, Microsoft is also delivering new functions supporting additional autonomy for AI, albeit while preserving a level of human oversight when and where that’s needed.
Copilot Studio’s new AI Approvals automate rules-based decision-making steps that take place as part of multistage approval processes. AI evaluates such requests and makes consistent approval or rejection decisions automatically.
Although AI approvals use a company’s business rules, they’re different from traditional rules-based automation in that they can analyze unstructured data, understand complex documents, and apply nuanced logic. Human reviewers can still finalize critical decisions.
“Humans always stay in control, as they can veto approval decisions,” said Derah Onuorah, product manager for Microsoft Copilot Studio, in a LinkedIn post on the latest updates. He added that AI approvals will increase speed, deliver cost savings through outcomes such as early payment discounts, and increase human productivity by offloading many routine functions.
How Approvals Work
AI Approvals integrate into existing multistage approvals in Copilot Studio. Users or developers can add automated decision points, also called AI stages, to streamline approvals that are considered routine. Creating and launching approval flows involve three steps:
- Defining decision criteria. At this stage, a user writes instructions for AI. One example: “reject expense reports over a threshold of $1,000 without pre-approval from a person classified as a manager in the system.”
- Provide inputs for guidance. Supply documents, images, text, or knowledge of the organization so that AI has information needed to make informed decisions that comply with company rules.
- Review decisions. As AI evaluates individual requests, it applies the rules it’s been given and makes an approve or reject decision, while providing an explanation for how and why it decided.
Microsoft has also published a prompting guide to help customers derive the most value from AI Approvals.
AI Agent & Copilot Summit is an AI-first event to define opportunities, impact, and outcomes with Microsoft Copilot and agents. Building on its 2025 success, the 2026 event takes place March 17-19 in San Diego. Get more details.
Where Approvals Work
In his LinkedIn post and a related blog, Microsoft’s Onuorah identified several common processes where AI approvals can provide a productivity boost, including:
- Expense reports. AI can approve or reject reports automatically when they comply with policy thresholds, receipts are valid, and spending is in line with company requirements.
- Purchase orders. AI approvals can be issued when requests are within a departmental budget, meet vendor requirements, and fall within approved authorization levels.
- Vendors and suppliers. AI can approve or reject new vendor applications based on supplied qualification criteria, compliance considerations, and risk thresholds that have been established.
- Invoice processing. AI can validate and approve invoices when documentation is complete, financials are accurate, and payment terms fit within policies.
- Document reviews. AI can approve contracts, proposals, or reports that meet formatting standards – which includes required elements – and are complete.
- Time off requests. Based on available balances, as well as employee and functional coverage needs, and scheduling rules, AI can automatically approve, reject, or flag requests for further human review.
- Travel authorizations. If requests comply with policy — including budgetary requirements — and are booked in advance, AI can approve them.
AI Approvals are available in preview in agent flows in Copilot Studio
Concluding Thoughts
Any manager who’s struggled with the time involved in reviewing documents, policies, or other elements in the processes detailed above can relate to the time and energy they can consume, so capitalizing on AI to alleviate as much of that burden as possible means more productivity and less angst.
By empowering AI to make more routine and repeatable decisions — while also taking steps to ensure humans stay at the helm of the most important business decisions — Microsoft continues to work toward a balance whereby customers can get the most impact from AI while keeping people fully in the mix for those decisions that only humans can make. Achieving that balance remains a work in progress, but this is another step forward.
Ask Cloud Wars AI Agent about this analysis
