Microsoft is running the same platform play that gave it 30 years of enterprise dominance. The question isn’t whether Copilot is good. It’s whether your team will use it.
Disclosure: I spent 26 years at Microsoft. That gives me a clear view of how Copilot fits when you’re already deep in Microsoft 365, and where it adds less if you’re not. LumenForge is vendor-neutral: we don’t sell Microsoft, we don’t earn from their stack, and we’ll say the same thing about Google, AWS, or anyone else. The question is always what works for your business, not whose logo is on the tool.
Bottom line: Microsoft’s Copilot strategy is sound, and the growth numbers are real. But a license is not adoption, and adoption is not transformation. If you’re a small business already living in Microsoft 365, the decision in front of you isn’t “should we buy AI?” It’s “are we going to get value out of what we’re already paying for, or let it gather dust?”
The strategy is sound: distribution is the moat
Copilot’s competitive advantage isn’t the model. It’s where the model lives.
Microsoft is doing what it has always done well: embed the new capability inside tools people already open every morning, make the platform harder to leave, and capture margin on the layer that sits on top. Word, Excel, Outlook, Teams. Copilot is already there. No new login, no new vendor, no procurement fight.
What we’re seeing in the numbers backs that up. Microsoft reported more than 30 million paid Microsoft 365 Copilot seats on its July 29, 2026 earnings call, up from roughly 20 million just three months earlier [1]. Customers holding 50,000+ seats grew more than sevenfold year over year [2].
And Microsoft keeps lowering the barrier. Copilot Chat is already included in Microsoft 365 and Office 365 business subscriptions at no extra charge, a deliberate move to show users the value before asking them to upgrade [3]. For a lot of small businesses, some form of Copilot is already in the building. You’re paying for the door whether or not anyone walks through it.
The grounding call: platform potential is not adoption
Here’s where I’ll be straight with you, as someone who spent 26 years inside Microsoft. Being positioned to win and winning are two different things, and Microsoft’s own history proves it.
- Cortana had the best distribution any assistant could ask for. It shipped inside Windows 10 starting in 2015. By 2019, Satya Nadella acknowledged it had fallen behind competitors, and the standalone Windows app was discontinued in 2023 [4]. Pre-installed on millions of machines; rarely used.
- Yammer cost Microsoft $1.2 billion in 2012. In 2023 the brand was folded into Viva Engage, with Microsoft itself citing customer feedback that overlapping apps had made it hard to drive adoption [5].
The key signal here is that distribution gets a tool in front of people. It does not make them change how they work. Even analysts who track Microsoft closely are skeptical. After the January seat disclosure, Directions on Microsoft analyst Greg DeMichillie concluded that Office isn’t much of an advantage when it comes to getting people to pay for AI [3].
Copilot is a far stronger product than Cortana ever was. But the failure pattern is the same one to watch for: a capable tool, sitting in the toolbar, that nobody builds into their day.
The metric that matters: usage, not licenses
License counts tell you what was sold. Seat adoption tells you whether anything changed.
Put the headline in context. Thirty million paid seats sounds massive, but against Microsoft’s commercial base it works out to roughly 6.5% penetration [1]. And buying is the easy part. A Gartner survey found that of organizations that finished a Copilot pilot, only 5% moved to a larger rollout [2].
Where this matters most is how “adoption” gets measured. According to Gartner analyst Dan Wilson, Microsoft’s default adoption setting counts one use in 28 days [2]. That’s not adoption. That’s someone clicking the button once a month. At $30 per user per month, that’s $360 a year for a curiosity.
Net: if you can’t answer “who used it this week, for what, and what did it save?” you don’t have a transformation. You have a line item.
The price tag is moving too
The per-seat license is no longer the whole bill. Microsoft is shifting Copilot from a flat subscription to a subscription plus metered usage, and most small businesses haven’t noticed yet.
What we’re seeing in 2026:
- Usage-based billing arrived June 16. Microsoft added charges on top of the $30 Microsoft 365 Copilot license, billed in “Copilot Credits” per task based on model use, context retrieval, tool calls, and runtime. Customers can pay as they go or pre-purchase for volume discounts [6].
- The new Copilot app splits the bill in two. The seat covers Chat, Office app integration, and model selection. Agent work (Cowork, Code, Autopilot, frontier models) runs on credits, added on top of the seat [7].
- The base suites got more expensive. Microsoft’s July 1, 2026 commercial price update raised prices on Business Basic, Business Standard, and Apps for Business, among others [8].
For small businesses, here is where the published prices sit today [9]:
| Plan | Price (per user/month) |
| Microsoft 365 Copilot Business (add-on) | $21 list; $18 current promo, annual. $25.20 monthly billing |
| Business Standard with Copilot | $23.50, annual |
| Business Premium with Copilot | $32, annual |
| Microsoft 365 Copilot (enterprise add-on) | $30, plus credits for agent work |
The key signal here comes from Directions on Microsoft analyst David Berry: the Copilot license is becoming “the cover charge, not the whole bill” [10]. Where this matters for a 50-person firm: a few enthusiastic users running agents can create a bill nobody budgeted, while everyone else’s seat sits idle. Usage tracking used to simply be about value. Now it is also about cost control.
What small businesses should do instead
Don’t deploy broadly and hope it sticks. Go narrow, prove it, then expand.
- Pick two or three high-frequency workflows. Things your team does every day or every week: meeting recaps, customer email replies, proposal first drafts, month-end report summaries. Frequency is what turns a feature into a habit.
- Measure before and after. Time the workflow today. Time it again in 30 days. Set your own adoption bar at daily use, not Microsoft’s 28-day default.
- Make use non-optional for those workflows. It is critical that the entire organization, from top to bottom, uses it. Ensure that leadership goes first and uses it visibly. If the owner still writes meeting notes by hand, nobody else will change.
- Decide where the saved time goes. This is the step everyone skips. Microsoft-affiliated researchers in an NBER field experiment found engaged users cut about two hours a week of email time, but saw no detected shift in how much other work got done [2]. Saved minutes do not reinvest themselves. That is a management job.
- Expand seats only where usage earns it. Not every role needs a paid license on day one.This is where a pilot is crucial, as is building a collection of “champions” that will embrace, test, and help integrate and land it.
The math only works when people actually use it. The UK government’s cross-government trial reported about 26 minutes saved per user per day, self-reported [2]. On a $30 seat, that’s under $3.50 per hour saved. On the $21 Business tier, the hourly cost drops further. For someone who opens it once a month, the math collapses, and metered agent work can push the real cost above the sticker price.
One bold move worth considering: run a 30-day “Copilot-first” sprint on a single team. Every recap, every first draft, every summary starts in Copilot, no exceptions. At the end, you’ll have real numbers, real champions, and a clear answer on whether to scale.
Net
Microsoft has built the right platform. Whether it becomes your operating layer or your most expensive shelfware isn’t Microsoft’s call. It’s yours.
The difference comes down to three things: change management, workflow specificity, and leadership making use non-optional. None of that ships in the box. AI is a force multiplier, not a replacement, and it only multiplies what your people do with it.
That’s the work we do at LumenForge: Education so your team knows how to use it, governance so they use it safely, and integration so it lands in real workflows. Vendor-neutral, people-first, deliberate.
If you’re on Microsoft 365 and not sure whether Copilot is earning its keep, book a free AI Reality Check Call. We’ll identify your first two or three workflows and how to measure them.
References
- 2Data. “Microsoft 365 Copilot Passed 30 Million Seats: What the FY26 Numbers Actually Tell Enterprise Buyers.” 2026.
- Mr. SharePoint. “Microsoft 365 Copilot Adoption Statistics for 2026.” August 10, 2026. Includes summaries of Microsoft FY26 Q4 earnings, Gartner research, the UK government trial, and NBER Working Paper 33795.
- Directions on Microsoft. “Microsoft Claims 15 Million Paid M365 Copilot Seats.” January 29, 2026.
- NewsBytes. “Microsoft Discontinues Cortana Digital Assistant on Windows 11.”
- TechCrunch. “Microsoft Ditches Yammer Brand and Goes All-In on Viva Engage.” February 14, 2023.
- Directions on Microsoft. “Microsoft’s Copilot Cowork Gets Model Choice Plus Usage-Based Billing.” June 16, 2026.
- Directions on Microsoft. “Microsoft Shows Off Its New Copilot ‘Super App’.” September 25, 2026.
- Microsoft. “2026 Microsoft 365 Packaging and Pricing Updates FAQ.” 2026.
- Microsoft. “Microsoft 365 Copilot Business.” Pricing accessed October 5, 2026.
- Directions on Microsoft. “Exploring the Hidden Costs of Microsoft 365 Copilot.” September 11, 2026.