Microsoft has announced a deep overhaul of the Copilot app: from a single interface to three distinct sections — Home, Code and Autopilot. The last one is an AI agent that runs continuously in the cloud, monitors Teams channels and completes tasks on its own, without anyone having to start it each time.
The change that affects company finances most isn't the agent itself: it's the pricing model. For Autopilot and Code, Microsoft is dropping the flat-rate subscription and moving to usage-based billing. You pay for what you use, not the same monthly fee as everyone else.
For anyone managing an IT budget, this means one thing: you need to understand right away how many actions the agent will perform each month, before rolling it out at scale. An agent active across several Teams channels, with no defined spending cap, can generate costs that are hard to predict.
Three sections, one always-on agent
Microsoft has redesigned Copilot by splitting it into three areas: Home (the classic conversational assistant), Code (tools for developers) and Autopilot, the most significant addition for anyone running business operations.
Autopilot is an AI agent — software that acts autonomously on predefined instructions — built on OpenClaw, the model Microsoft is investing in for complex tasks. The difference from an ordinary chatbot is substantial: it doesn't wait for someone to ask it a question. It runs continuously in the cloud, monitors Microsoft Teams channels and completes tasks on its own, as reported by The Decoder.
In practice, you can instruct it to read the requests coming into a Teams channel, classify them and then reply or delegate, with no human intervention on each individual message.
The pricing shift that really hits your budget
The technical announcement is interesting, but the real management shift is something else: Microsoft is dropping the flat rate for Autopilot and Code and introducing usage-based billing — you pay for every action the agent performs, not a monthly fee that's the same for everyone.
This model has two sides:
- Upside: if you use the agent a little, you pay a little. You don't pay for an unused license.
- Risk: an agent active across several channels, with lots of incoming requests, can generate an unpredictable monthly bill if you don't set usage limits from the start.
The logic is the same one we've seen with other AI agents on the market. If you're following the evolution of AI agents for business automation, you know the shift from subscription to usage-based pricing is a broad trend, not an isolated Microsoft decision.
Where it can help an SME, and where it can't
Autopilot makes sense where Teams has repetitive workflows: internal ticket handling, sorting sales requests, project status updates. If your company uses Teams as its main operational channel, the agent can take real manual work off your plate.
It doesn't make sense, however, if:
- Teams is only used for the occasional video call;
- your processes aren't documented yet — an agent doesn't know what to do if there are no rules;
- no one is available to monitor variable costs month by month.
The same logic applies to voice agents: the technology works when the underlying process is already clear. We saw this when analyzing how Ringg AI handles customer service with voice agents — 65% of calls resolved autonomously is a real result, but it requires a precise setup of the flows.
What to do before you turn it on
If you're evaluating Autopilot for your organization, there are three things to do before switching it on at scale:
- Map your active Teams channels: how many messages do they receive per day? How many actions would the agent generate in a typical month?
- Set a spending cap: without a defined maximum budget, usage-based billing can catch you off guard. Ask Microsoft or your reseller which spending controls are available.
- Start with a single channel: activate Autopilot on a limited workflow, measure costs and results for four weeks, then decide whether to scale.
The same gradual approach applies to any AI agent. If you've read how the ChatGPT Work Data Agent for SMEs works, you know the safest adoption curve always starts from a small, measurable scope.
The typical mistake with usage-based billing
The main risk isn't technical: it's governance. Companies that turn on AI agents under usage-based models without someone in charge of monitoring end up, after the first month, with an invoice that's hard to justify to the CFO.
Usage-based billing rewards those with clear processes and predictable volumes. It penalizes those who switch the technology on experimentally, without controls. It's not a Microsoft problem: it's the very structure of the model, which OpenAI has already introduced with its sponsored agents integrated with HubSpot and Shopify and which also shows up in more recent integrations like the automated WhatsApp Business setup with AI agents.
If your team doesn't yet have a process for tracking AI tool costs month by month, now is the time to build one — regardless of Autopilot.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
