← Latest AI News

October 12, 2026 · Daily brief

Microsoft agents now bill by the task, not the seat

Sovereignty angle
An agent that bills by the task instead of the seat can save you money or surprise you with a bill three times what you budgeted. The difference is whether you set the spending cap before the agent starts running.

Microsoft rebuilt Copilot with new agent modes that can draft, code, and run tasks while you're offline. Some now bill by usage instead of a flat monthly fee.

Microsoft announced a rebuilt Copilot on October 1 with three new capabilities: Home merges chat and work surfaces, Code lets you build trackers and dashboards in plain language, and Autopilot is an always-on agent that keeps working when you're offline. Home and Code enter preview in October; Autopilot expands to private preview by month-end.

What you can do with it now

If your team already uses Microsoft 365, Copilot can now do more without you watching. Autopilot drafts follow-ups, proposes schedule changes, and surfaces forgotten tasks with the context you'd need to act. Code lets non-technical people build simple apps — a lead tracker, an inventory dashboard — by describing what they want instead of hiring a developer.

The catch is billing. Some of these new agent features moved from the flat Copilot subscription to usage-based billing. You pay per task the agent completes, not per person per month. Microsoft calls this UBB. New agent work in SharePoint, OneDrive, and Copilot Studio will bill this way starting in October.

Microsoft says it's giving admins cost controls: you can set spending caps, choose which model families different groups can access, and track usage in Agent 365 dashboards. Early pricing isn't published yet for most agent tasks.

What to keep in your own hands

Set a spending limit before you turn agents loose. Usage-based billing means the bill can grow without you noticing if agents run more tasks than you expect. Go into the Microsoft 365 admin center, find Copilot cost management, and set your cap before the October rollout.

Decide which tasks an agent can finish versus which need a person to approve first. An agent that drafts a follow-up and waits for you to send it is different from one that sends it while you sleep. Check the Autopilot settings for approval rules, and default to requiring confirmation until you trust the output.