Microsoft Just Priced AI at $23.50 a Seat. Here Is How to Know If It Is Worth It.

TL;DR

Microsoft just turned its Copilot add-on into a permanent $23.50-a-month line item for every Business Standard seat. That number is not really the story. The story is that you now have a clean benchmark for judging any AI tool's price, and most small businesses are about to make the same mistake with this one they made with the last one: buying seats before naming the task.

What did Microsoft actually change?

As of July 1, 2026, Microsoft 365 Business Standard with Copilot and Business Premium with Copilot became permanent subscription tiers, priced at $23.50 and $32 per user per month. Both are available to organizations with one to 300 seats on annual billing, according to Microsoft's own Partner Center announcement (Microsoft Learn).

Before this change, Copilot was a short-term promotional add-on layered onto an existing Business Standard or Premium subscription. Now it is baked into the plan itself, at a fixed price with no expiration date. Windows Forum's write-up on the change frames it the same way: Microsoft is done selling AI as an optional extra and is treating it as a standard part of the core productivity seat (Windows Forum).

Existing Business Standard and Premium customers who add Copilot separately still see a $21 list price, discounted to $18 through September 30, 2026. The $23.50 bundled figure is the one that matters for anyone choosing a plan from scratch, since it is the price of the whole seat, productivity suite included.

That is a real change in how Microsoft sells AI. It is a smaller change in whether your team actually gets value from it.

Why does a $23.50 number matter more than the product itself?

Because for the first time, a large vendor has put a specific, durable number on "AI, included." That gives every small business owner a benchmark to hold every other AI subscription up against, whether or not you ever touch Copilot.

This lands at the same moment the broader AI pricing market is shifting toward outcome-based charging: paying per case resolved, per document reviewed, per hour saved, instead of a flat seat fee regardless of use (SaaS Intelligence). A fixed per-seat number like Microsoft's is the opposite bet. It only works out if the seat gets used.

We have run enough AI Efficiency Audits to know exactly what happens next at a lot of companies. Someone in finance approves Copilot for the whole team because the price looks reasonable next to a full software license. Three months later, half the team never opened it past the first week, and the bill is still $23.50 times every seat, every month, whether it gets used or not.

That is not a Microsoft problem specifically. It shows up with nearly every per-seat AI tool we have reviewed for clients, and it connects directly to something we found when we looked at why small businesses stall on AI adoption: the barrier is rarely the price tag. It is not knowing where to start, which means the seat sits unused at whatever price you paid for it.

The outcome-based pricing shift is a useful contrast here, not a competing headline. A vendor charging per case resolved or per hour saved cannot bill you for a seat nobody touches, because there is no seat, only usage. Microsoft's flat $23.50 puts the entire burden of proving value back on your business, since the invoice arrives whether the tool gets opened or not.

That is the part worth sitting with before you sign up for seats across the company. A flat price is not a bad deal. It is a deal that only pays off if someone actually owns making it pay off.

How do you know if an AI seat is actually worth its price?

Run one calculation before you buy a single seat, let alone a whole team's worth: name the task, time the task, price the time.

Pick the one recurring task you would actually hand to the tool, not a vague "help with productivity" hope. Drafting client emails, summarizing meeting notes, building a first-pass spreadsheet model, cleaning up a report template. Something specific enough that you could describe it to a new hire in one sentence.

Then time how long that task takes a person right now, honestly, not the best-case estimate. Multiply that time by what an hour of that person's time actually costs the business, wages plus overhead, not just their hourly wage.

If $23.50 a month is small next to the hours you would actually save on that one task, the seat pays for itself and the rest is upside. If you cannot point to a specific task at all, you are buying a subscription on the hope that a use case shows up later, and that is exactly the pattern that leaves an unused seat on next year's software bill.

What is the mistake almost every SMB makes with a per-seat AI tool?

Buying company-wide before piloting narrow. It feels efficient to roll out one decision for the whole team instead of managing a patchwork, but it is the single most common way we see AI budgets get wasted.

The fix is not complicated, and it does not require a technical team to run. Give a small group, three to five people who already do the target task daily, two to four weeks with the tool on that one task specifically. Have them log roughly how much time it actually saved, not how impressive the demo felt.

If the pilot group's time savings hold up, expand seat by seat to the people doing that same task, not to the whole company at once. If they do not hold up, you have spent a few seats' worth of budget finding that out instead of a whole team's worth.

We see the same failure mode with almost every category of AI tool, not just Copilot: a scheduling assistant, a customer-service chatbot, a meeting-notes app. The tool itself is rarely the problem. The rollout order is. Whole-team access on day one skips the one step, a small proof, that tells you whether the task you picked was actually a good fit for AI in the first place.

This is the same review-and-expand discipline we described for letting an AI agent act in your business: start narrow, prove it, then widen the rollout, rather than granting the whole team access on day one and hoping adoption follows.

We build this same staged rollout into every Perpetua deployment for exactly this reason. An always-on AI employee earns its next task by proving the last one, and the client sees the hours it actually saved before the scope grows, rather than paying for a full team's worth of capability on day one and finding out later which parts of it went unused.

So what does this mean for your business?

It means you now have a clean number to test any AI subscription against, Microsoft's or anyone else's: does the task it replaces cost you more in time than the seat costs in dollars. That is the whole decision, and it takes one afternoon to work out per tool, not a procurement process.

It also means the $23.50 figure itself is not a signal to buy or skip Copilot specifically. It is a floor for the conversation. An unused $23.50 seat and an unused $60 seat are the same mistake at different prices, and a used $23.50 seat that saves five hours a month is money made back many times over.

Pick the one task, pilot it narrow, measure the hours, then decide. That sequence works whether the tool in question is Copilot, a competitor, or something you have not tried yet. You can read more on how we structure these evaluations on the AI Insights Hub.