Microsoft Just Put a Price Tag on Every AI Task. What That Changes for Your Budget.

Read that again. Consumption-based.

For the last few years, the cost of AI inside Microsoft 365 has behaved like a phone plan. You bought a per-seat Copilot license, and the price was the price whether your team used it once a week or once a minute. Flat. Predictable. Easy to put on a slide for the board.

That is no longer the only model on the table. With Work IQ now billed by usage, certain kinds of AI work get a meter attached. The bill moves with how hard your agents work. Worth understanding before it surprises you in an invoice.

What Microsoft shipped

Work IQ is the layer Microsoft calls "the intelligence layer behind how work gets done." It reads across email, calendar, meetings, chats, files, people, collaboration patterns, and line-of-business systems to build a working understanding of how your organization operates. Microsoft says the average Work IQ data footprint inside a Fortune 500 company runs over 600 terabytes, a figure from its own internal data in May 2026.

The new API, announced June 2, 2026 by Charles Lamanna, Microsoft's Executive Vice President of Copilot, Agents, and Platform, lets developers build custom agents that reason over that Microsoft 365 data while keeping existing permissions, compliance, and governance controls intact.

There are three pieces a builder can call:

• The Chat API, which Microsoft describes as "programmatic access to the full power and intelligence of Microsoft 365 Copilot," returning answers with citations.
• The Context API, which hands back agent-ready context and source data instead of a finished answer.
• The Tools API, which gives agents the verbs to do things like send an email or schedule a meeting.

You do not need a new subscription, a new SKU, or a new per-user license to use it. The currency is Copilot Credits, the same credits that cover Copilot Studio and other Microsoft AI services.

One important line for anyone worried about a runaway invoice: if your people use the prebuilt agents that ship inside Microsoft 365 Copilot, their Work IQ usage is covered by licensing and costs no credits. The meter starts running when you build custom agents in Copilot Studio, Foundry, or a third-party platform and ground them in your Microsoft 365 data through these APIs.

Per-seat and per-call are different animals

Here is the practical change for a P&L owner.

A per-seat license is a headcount question. You hire ten people, you buy ten licenses, you know the annual number. Budgeting is arithmetic.

A consumption model is a usage question. Microsoft's words again: "You pay for use of the Work IQ API through a usage-based model that uses Copilot Credits." The cost depends on how often your agents run and how heavy each task is.

Microsoft splits the bill into two parts. The Tools API carries a fixed charge of 0.1 Copilot Credits per call. The Chat and Context APIs are variable, priced by how complex the request is.

The published ranges tell the story:

• A light task, like identifying what needs to be done, runs $0.20 to $0.40 per call.
• A medium task, like analyzing a customer interview, runs $0.30 to $0.75 per call.
• A heavy task, like producing an executive summary from a stack of meetings and documents, runs $0.50 to $1.50 per call.

A dollar fifty sounds like nothing. Run that heavy summary across every meeting, for every manager, every day, and the arithmetic stops being trivial. This is the same math anyone who has watched a cloud bill already knows. Small per-unit costs are friendly until volume shows up.

The new job almost nobody has yet

Before a single credit gets spent, Microsoft requires an administrator to turn on consumptive billing, attach a payment method, and set access controls, limits, and alerts in the Microsoft 365 admin center.

That setup is a job in itself.

Microsoft built a Cost Management dashboard for it. Inside it, an administrator allocates Copilot Credits, applies policy-based access and limits, runs prepaid and pay-as-you-go side by side, and sets budgets, alerts, and hard caps to prevent overspending. Enablement can run on prepaid credits, pay-as-you-go, or existing capacity, and organizations can connect Azure subscriptions to handle billing at scale.

The closest existing role is usually license management, a once-a-year renewal task. Watching a live consumption dashboard, setting caps, and reading the spend the way a finance lead reads a credit card statement is a different muscle.

An agent left running without a ceiling is the metered-billing version of leaving the lights on, except the lights can think.

If you take one operational step from this story, it is this: decide now who owns that dashboard. Give them the authority to set hard caps.

Why designing the workflow first lowers the bill

The cost model rewards a discipline that is easy to skip.

Each call has a price that depends on which API you use and how much work you ask for. A Tools API call to send an email is a fixed, tiny charge. A heavy Chat API call to synthesize a quarter of meetings is the top of the range. An agent that fires the expensive call when a cheap one would do is an agent that quietly inflates your bill.

This is where capability inside your team becomes a line item, not a nice-to-have. A person who understands the work well enough to design the workflow before turning an agent loose will pick the right tool for each step. That choice is the difference between a controlled spend and a creeping one.

It connects directly to how we measure capability in The 7 Levels of AI Proficiency. The early levels are about using AI tools that someone else built and paid for. The higher levels are about building, governing, and orchestrating agents that act on your behalf, which is exactly the territory the Work IQ API opens. Metered billing raises the value of that climb. The person who can design an efficient workflow and read a consumption dashboard is now saving you money, not just moving faster.

What to tell the board

Three things are worth raising in your next planning conversation.

First, the cost of AI is starting to behave like the cost of cloud compute. It scales with use, it can spike, and it needs a budget owner. The flat per-seat era is not over, but it is no longer the whole picture.

Second, governance and spend now travel together. The same admin controls that keep permissions and compliance intact also hold the caps that keep the bill in line. Whoever owns one should probably own the other.

Third, the organizations that come out ahead will be the ones that treated this as an operating decision, not an IT setting. Who builds the agents. Who approves them. Who watches the meter. Those questions belong on a leadership agenda, not in a footnote.

A next step worth taking this week

You do not need to deploy anything to get ready. Ask one question inside your organization: if we started building custom agents on our Microsoft 365 data tomorrow, who owns the cost dashboard, and what hard cap would they set?

If the answer is a blank look, that is the work to do first. Name the owner. Set the ceiling. Then decide which workflow is worth building well before you turn an agent loose on it.

Related reading: Level 5: The Captain (Design Thinker).

Sources

1. Work IQ API general availability and Copilot Credits billing
2. Announcing the new Work IQ APIs (Microsoft 365 blog, Charles Lamanna)
3. Work IQ API overview (Microsoft Learn)
4. Usage-based billing and Cost Management for Copilot Credits (Microsoft Learn)

Harrison Painter

Executive AI Advisor. Founder, LaunchReady.ai and AI Law Tracker.

Harrison is an Indiana AI Advisor who helps business owners and executives get their time back by building AI systems that run the work for them. Nearly 20 years in business and author of You Have Already Been Replaced by AI. Creator of The 7 Levels of AI Proficiency.

Connect on LinkedIn