Microsoft’s ambitious pivot to become an AI-first company faces an unexpected obstacle: its own flagship product. Copilot, the artificial intelligence chatbot that Chief Executive Satya Nadella positioned as the cornerstone of Microsoft’s transformation, is struggling to gain traction with users despite massive marketing investments and the company’s embedded position in corporate America.

The challenges facing Copilot extend far beyond typical product launch hurdles. According to The Wall Street Journal, current and former employees describe a product plagued by confusing brand positioning and persistent interoperability problems that have frustrated both enterprise customers and individual users. The stakes couldn’t be higher for a company that has invested billions in AI infrastructure and positioned itself as OpenAI’s primary commercial partner.

Previously unreported data reveals the depth of Copilot’s user retention problem. From July through late January, the percentage of Copilot subscribers who use the product as their primary option plummeted from 18.8% to 11.5%, according to a survey of more than 150,000 respondents in the U.S. by market research firm Recon Analytics. During the same period, Google’s Gemini saw its share of paid users who choose it as their first option increase from 12.8% to 15.7%, suggesting that Microsoft isn’t just failing to grow—it’s actively losing ground to competitors.

The Numbers Tell a Troubling Story

Microsoft’s latest earnings report revealed that the company had sold 15 million corporate Copilot “seats” within its 365 productivity business, which maintains a base of more than 450 million paid seats overall. While Microsoft touted more than 150 million monthly active Copilot users across its first-party platforms, these figures pale in comparison to competitors. Google’s Gemini boasts over 650 million monthly users, while ChatGPT has approximately 900 million weekly active users, according to the Journal.

The utilization rates among paying customers present an even more concerning picture. Some companies are using only about 10% of the Copilot subscription seats they are paying for, according to a recent note by analysts at Citi Research. The analysts identified “disorganized data silos” as a significant issue preventing broader adoption. When workers have access to subscriptions for Copilot, ChatGPT, and Gemini simultaneously, they choose ChatGPT and Gemini at higher rates than Copilot, the Recon data shows.

Jared Spataro, Microsoft’s chief marketing officer of AI at work, pushed back against these findings in an interview with the Journal. He claimed that daily active usage of 365 Copilot has grown 10-fold year-over-year and is outpacing growth of other 365 enterprise offerings. “The pace of growth that we’re seeing is unlike anything we’ve seen before,” Spataro said, though he declined to provide specific figures to substantiate these claims.

A Product Identity Crisis

At the heart of Copilot’s struggles lies a fundamental confusion about what the product actually is. Microsoft has deployed several versions of Copilot woven into different apps and services, including its 365 productivity tools such as PowerPoint, the GitHub developer platform, and a consumer-facing version available through its Edge browser and via a standalone app. These different iterations are divided into three main categories: enterprise tools sold to companies and professionals, Copilots for developers and IT personnel, and the general consumer chatbot led by Microsoft AI CEO Mustafa Suleyman.

Customer surveys conducted by Microsoft have shown that users remain confused by the multiple versions of Copilot, according to people familiar with the matter cited by the Wall Street Journal. Some users have long complained about Copilot being forced onto them, with the AI assistant popping up on everything from documents to the browser without clear explanation of which version they’re using or what capabilities each possesses.

The lack of a cohesive experience across different Copilots has become a persistent pain point that even Nadella has acknowledged. About a year ago, the CEO sent a frustrated email to Rajesh Jha, executive vice president of experiences and devices, detailing an incident in which Nadella had asked the enterprise version of Copilot on the Edge browser to help with a public webpage he was viewing, but it couldn’t fulfill his prompt, according to people familiar with the email. While that specific issue was resolved, similar interoperability difficulties continue to plague users.

Organizational Silos Hamper Progress

Microsoft hired Mustafa Suleyman in 2024 to oversee its consumer AI products and build AI models that could compete with those of OpenAI and others. However, some employees told the Journal that an organizational silo between Suleyman’s consumer-focused team and teams working on enterprise versions has made it challenging to achieve a unified vision for the product. This internal fragmentation manifests in the user experience, where different versions of Copilot often feel like entirely separate products rather than variations of a cohesive platform.

Spataro defended this separation in his interview, arguing that the divide between the 365 Copilot and the consumer version is by design, allowing users to keep their work and personal information separate. Yet this explanation doesn’t address why users find the experience disjointed or why interoperability remains problematic even within enterprise contexts.

The organizational challenges extend beyond team structure. The effort to train proprietary models has been hampered by a shortage of computing capacity, with the company rationing server time to ensure availability for OpenAI and other customers of its Azure cloud service. On numerous benchmark tests, Microsoft’s flagship proprietary AI model ranks well below competitors, raising questions about whether the company can develop competitive technology independent of its OpenAI partnership.

The OpenAI Paradox

Microsoft’s close partnership with OpenAI, once viewed as a strategic masterstroke, has become a double-edged sword. The company has relied heavily on OpenAI and its competitor Anthropic to power its various Copilots, stating it will use the best models available. However, this dependence has left Microsoft vulnerable as its relationship with OpenAI evolves and potentially diminishes.

Microsoft shares tumbled after its earnings report last week sparked investor concern that growth in its most important unit, the Azure cloud-computing business, is slowing, and that its AI business remains overly reliant on OpenAI while Copilot remains unproven. Shares fell nearly 3% on Tuesday amid a broader slide in software stocks prompted by fresh concerns that AI tools will make enterprise subscriptions less necessary—a particularly troubling development for a company built on subscription revenue.

The timing couldn’t be worse. In recent weeks, a new AI product from Anthropic called Claude Cowork has drawn praise for its ability to work in and across 365 applications in ways that Copilot users have found difficult. The release of new Cowork features was a factor behind Tuesday’s slide in software stocks, according to the Journal. Microsoft product leaders have held discussions about Cowork, and the company has internally worked on a similar product, a person familiar with the matter said.

Infrastructure Constraints and Strategic Trade-offs

Scott Guthrie, executive vice president of Microsoft’s Cloud and AI Group, acknowledged in an interview with the Journal that Suleyman’s team was only created in 2024. He explained that the long timeline involved in building cloud infrastructure accounted for the team’s lack of computing capacity until recently. This explanation, while technically accurate, underscores how Microsoft’s infrastructure decisions have prioritized serving OpenAI and Azure customers over developing its own competitive AI products.

In its recent earnings report, Microsoft indicated it was allocating more computing power to improve its Copilot products as it had gained confidence in its ability to monetize them. However, shareholders aren’t yet convinced that this trade-off makes sense, according to a note from investment firm UBS. The market’s skepticism reflects broader concerns about whether Microsoft can successfully balance its roles as both an AI infrastructure provider and an AI product company.

“We have moved past the initial phase of discovery” of AI, Nadella wrote in a December blog post, adding that the industry was entering a phase where “we are beginning to distinguish between ‘spectacle’ and ‘substance.'” The irony is that Microsoft’s own flagship AI product may be struggling to demonstrate the substance Nadella champions.

Internal Adoption as a Testing Ground

One area where Microsoft is having success getting more people to use Copilot is within its own workforce. Nadella has urged company leaders and staff to lean into AI tools as part of what the company calls a push to be a “frontier firm.” Microsoft has instructed managers to include questions about AI use in performance discussions, and employees are asked to quantify how they are using AI tools like Copilot in their workflows, according to people familiar with the matter.

Pam Maynard, chief AI transformation officer for Microsoft’s Customer and Partner Solutions organization, told the Journal that the adoption rate within Microsoft’s sales organization has surged from about 20% to more than 70% in the past year as more employees have become comfortable using AI. This internal push represents both a genuine effort to transform work practices and a necessary testing ground for products that haven’t yet proven themselves in the broader market.

The company, which laid off more than 15,000 workers last year, created a boot camp for software engineers, designers, and product managers to immerse themselves in the latest AI tools, with the goal of using AI to work more efficiently together. The camp is about “getting people really comfortable with changing their mindset from ‘I used to think of myself as a great coder, now I’m a product builder,'” Katy George, corporate vice president of workforce transformation, said in an interview with the Journal.

Marketing Blitz Meets Market Resistance

To counter Copilot’s user retention problems, Microsoft has launched an unprecedented marketing offensive. In 2025, the company spent roughly $60 million on television ads for Copilot, according to ad tracker iSpot. In comparison, it spent less than $1 million on ads for LinkedIn, its next-most-heavily advertised brand on TV. The company planned to run an ad for Copilot during the Super Bowl, its second since 2024, with a 30-second spot during the game costing more than $8 million.

This massive marketing investment reflects both Microsoft’s commitment to the product and the magnitude of the challenge it faces. Unlike previous Microsoft products that could rely on enterprise sales channels and existing customer relationships, Copilot must compete in a crowded market where users have genuine alternatives and can easily switch between competing services. The heavy advertising spend suggests Microsoft recognizes that brand awareness alone won’t solve its retention problems, but hopes it can at least drive trial and initial adoption.

Those surveyed who switched from Copilot cited finding better quality elsewhere, with some specifically mentioning poor user experience and restrictive usage limits. These qualitative findings align with the quantitative data showing declining usage rates, painting a picture of a product that struggles to retain users once they’ve tried alternatives.

The Long Game and Competitive Advantages

Despite current struggles, analysts remain cautiously optimistic about Microsoft’s long-term prospects in the AI space. Even if Microsoft is lagging behind in the chatbot race, it continues to generate billions from AI-driven cloud-computing demand, keeping it among the world’s most valuable companies. Analysts see the company as well positioned to close the gap because its productivity software is used by hundreds of millions of corporate users—a captive audience to whom it can easily promote new AI products.

Chad A. Morganlander, senior portfolio manager at Microsoft investor Washington Crossing Advisors, offered a pragmatic perspective on the company’s challenges. “While Copilot is struggling now, our bet is they have this embedded client base, and that they will get it wrong until they get it right,” he told the Wall Street Journal. “They have plenty of money for the marathon.”

This view reflects a broader truth about Microsoft’s position: the company has both the financial resources and the customer relationships to persist through product missteps that would sink smaller competitors. With more than 450 million paid seats in its 365 productivity business, Microsoft has an enormous potential user base that competitors cannot easily reach. The question is whether the company can fix Copilot’s fundamental problems before user frustration hardens into lasting preferences for alternative products.

Workforce Transformation as Product Development

George, whose workforce transformation team was created last year, said Microsoft has found that frequent AI use correlates to success in divisions such as sales. This internal data provides some evidence that AI tools can drive productivity gains, though it remains unclear whether these benefits stem specifically from Copilot or from AI tools more generally. The company’s push to make employees quantify their AI usage in performance reviews creates strong incentives for adoption that don’t exist in external markets.

While employee use might lead to a more AI-literate workforce and help refine the product through feedback and usage data, Microsoft still needs substantially more consumer and business users to choose Copilot over alternatives. The gap between internal adoption rates—driven by corporate mandate and performance incentives—and external adoption rates suggests that Copilot’s value proposition remains unclear to users who have genuine choice in the matter.

The internal focus on AI transformation also raises questions about resource allocation. As Microsoft dedicates significant energy to training its own workforce on AI tools, competitors are focused on making their products so intuitive and valuable that such extensive training becomes unnecessary. The need for boot camps and performance review mandates to drive adoption may indicate that Copilot hasn’t achieved the user-friendly design that characterizes successful consumer technology products.

Strategic Crossroads and Future Directions

Microsoft finds itself at a critical juncture. The company has made enormous investments in AI infrastructure, marketing, and product development, yet its flagship AI product is losing users to competitors. The organizational silos between consumer and enterprise teams, the computing capacity constraints, and the persistent interoperability problems all point to deeper strategic questions about how Microsoft approaches AI product development.

The emergence of Claude Cowork as a potential alternative that better integrates with Microsoft’s own 365 applications represents a particularly stinging challenge. If Anthropic can build a product that works more seamlessly with Microsoft’s ecosystem than Microsoft’s own offering, it suggests fundamental problems with how Copilot was conceived and executed. The fact that Microsoft has internally worked on a similar product indicates awareness of the gap, but also raises questions about why such capabilities weren’t prioritized from the beginning.

Nadella’s vision of transforming Microsoft into an AI-first company, much as he transformed it into a cloud-first company around a decade ago, depends heavily on Copilot’s success. The cloud transformation succeeded because Microsoft built genuinely superior products that addressed real customer needs. Whether Copilot can achieve similar product-market fit remains an open question, with current data suggesting significant work remains.

Market Dynamics and Competitive Pressures

The broader market context adds urgency to Microsoft’s Copilot challenges. The recent slide in software stocks, prompted by concerns that AI tools will make enterprise subscriptions less necessary, threatens Microsoft’s core business model. If AI assistants can accomplish tasks that previously required multiple specialized software subscriptions, the value of Microsoft’s extensive software portfolio could erode. This makes it essential that Microsoft controls the AI tools that might cannibalize its existing products, rather than ceding that position to competitors.

Google’s Gemini and OpenAI’s ChatGPT have established strong positions in both consumer and enterprise markets, with user bases that dwarf Copilot’s. These competitors continue to innovate rapidly, introducing new features and capabilities that set user expectations for what AI assistants should deliver. Microsoft’s challenge isn’t just to match these competitors but to differentiate Copilot in ways that leverage its unique advantages, particularly its integration with productivity tools that billions of people use daily.

The Recon Analytics data showing declining Copilot usage among paid subscribers is particularly concerning because it suggests problems with the core product experience rather than just awareness or trial. Users are choosing to pay for Copilot access—either directly or through enterprise agreements—but then opting to use competing products instead. This pattern indicates that marketing investments and distribution advantages cannot compensate for fundamental product shortcomings.

The Path Forward

Microsoft’s response to these challenges will likely determine whether Copilot becomes the transformative product Nadella envisions or a cautionary tale about the difficulties of product execution even for well-resourced technology giants. The company’s decision to allocate more computing power to Copilot development represents a necessary step, but computing resources alone won’t solve the interoperability, user experience, and brand confusion problems that current and former employees have identified.

The organizational questions may prove most critical. Breaking down silos between Suleyman’s consumer-focused team and the enterprise product teams could enable the unified vision that has proven elusive. However, such organizational changes often prove difficult to execute, particularly in large companies with established power structures and competing priorities. The fact that these silos persist despite Nadella’s known frustration with interoperability problems suggests deeper institutional resistance to change.

Microsoft’s embedded position in corporate computing provides a substantial advantage, but also creates risks. If enterprise customers grow frustrated with Copilot being “forced onto them,” as some users have complained, it could damage Microsoft’s broader relationship with these customers. The company must balance its desire to promote Copilot with the need to maintain trust and goodwill with the hundreds of millions of users who depend on its productivity tools for their daily work. How Microsoft navigates this tension will significantly influence whether its AI-first transformation succeeds or stumbles.