Partners say paid Copilot take-up remains selective as customers weigh AI options

Microsoft CEO Satya Nadella used the company’s FY26 Q3 earnings call at the end of April to share that the company now has over 20 million Microsoft 365 Copilot paid seats, with the number of customers with over 50,000 seats quadrupling year-over-year.

That 20 million figure, however, represents just 4.4% of the “more than 450 million” paid Microsoft 365 commercial seats cited during the company’s FY26 Q2 earnings call.

Partners that techpartner.news contacted recently shared that while Copilot interest is high, customer adoption remains highly selective as the realities of scale deployment sink in and alternate AI tools and LLMs gain attention.

“Most customers are taking a cautious, use‑case‑driven approach rather than broad rollout, said Warren Hill, senior sales executive at Dynamics 365 solutions provider Focus Dynamics Group.

“There’s a strong desire to understand governance, data readiness, and real productivity impact before making larger commitments.”

Katrina Marques, head of Microsoft and cloud services for Atturra Managed Services, agreed with that assessment, stating that most Atturra customers are still in a pilot or evaluation phase, with only a small number moving into broader production rollout.

“We’re seeing some organisations deploy Copilot to targeted user groups, but the majority are still working through readiness - particularly around data governance, security, and understanding where it will deliver meaningful value,” she explained.

Marques said that while there’s definitely strong and growing interest from customers, driven by the broader push around AI, there is also a level of uncertainty remaining.

“Many customers are still figuring out what the right use cases look like within their business, and how to approach adoption in a structured way,” she said.

“In a small number of cases, customers have committed more broadly to Copilot licensing, but even then, the focus is still on building the right foundations - defining governance, identifying use cases, and planning adoption rather than simply rolling it out at scale.”

"Rapid expansion” from pilot to rollout once foundation in place

Daniel Whittle, GM of AI managed services at end-to-end digital transformation partner Wild Tech, said that while most organisations are still in a readiness and foundation phase in terms of Copilot use rather than full-scale rollout, that's not necessarily a bad thing.

“What we're seeing consistently is that initial pilot groups are straightforward to stand up. The harder work is what comes next, as broader deployment tends to be held back by data and governance readiness,” he said.

“Key blockers often include the state of SharePoint and file structures, the implementation of Purview, information protection and DLP controls, and having clear data access and security boundaries in place.”

He said that as organisations are recognising that Copilot is only as effective and safe as the data it can access, that's subsequently shifted the conversation from 'how many Copilot licences do we buy?' to 'are we operationally ready to deploy this at scale?'.

“It's a more mature question, and the right one to be asking,” he noted.

“Once that foundation is in place, we're seeing rapid expansion from pilot to broader rollout, particularly in environments that have already invested in Microsoft's security and compliance stack. The groundwork pays off quickly.”

Specific use cases creating progress

Fuse Technology’s head of sales Peter Limnios said that for his customers, Copilot takeup is more varied, with hesitance at one end, full adoption at the other and most sitting somewhere in between.

“Interest is strong, but there's a genuine question customers are working through: where does Copilot fit relative to everything else they're already using or evaluating?,” he said.

“There's no shortage of AI tools in the market right now. The customers making the most progress are the ones who've defined specific use cases rather than trying to figure out AI broadly."

Tecala’s senior product manager Andre Schoeman, however, had a different view, stating that a “meaningful portion” of Tecala's customer base has moved out of pilot and into productive use.

“We've run security assessments alongside those deployments rather than after them, and customers are connecting the dots themselves between AI productivity and the data exposure it creates,” he said.

Schoeman stated that data loss prevention (DLP) becoming “inseparable” from the AI conversation has resulted in customers pursuing DLP foundations actively rather than reluctantly.

“This is also prompting harder conversations about SharePoint hygiene and information  governance, because Copilot outcomes are directly affected by the quality and structure of the underlying data,” he explained.

“We're also seeing real interest in non-Microsoft LLMs, which we host inside our Agentverse platform so customers can use multiple models without compromising on security posture.

"The organisations getting genuine value are treating AI as a change and productivity program, not just a licensing decision.”