Summary
Microsoft is asking enterprises to double their Office costs for AI features they cannot measure. Yet, one year in, half of the technology leaders surveyed still don't know if it's worth it. The question they cannot answer is how to measure the ROI of productivity software that promises to save time.
When Microsoft launched Microsoft 365 Copilot for enterprise customers in November 2023, the product promised AI that drafts emails, summarises meetings, analyses spreadsheets, and generates documents within the applications that knowledge workers use daily. Microsoft priced it at $30 per user per month.
Roughly two years later, enterprises remain divided on whether the investment pays off. According to an October 2024 survey of the CNBC Technology Executive Council, when technology leaders were asked whether Copilot had been worth the $30 monthly cost, equal numbers said yes and no. But the largest group, fully 50% of respondents, answered that it was still too soon to know.
Most organisations remain uncertain if Copilot’s promised productivity gains justify the added cost.
That uncertainty extends to deployment decisions. The survey found that whilst 50% of organisations had decided to roll out Copilot to all employees, another 17% had decided not to fully adopt it, and a third remained in the testing phase.
Despite the survey being one year old, our own observations are similar to it even now, in October 2025. The most recent conversations are painting a brighter picture, with more optimism towards the Copilot’s future. Organisations cautiously report few successful cases and some voice plans to increase adoption. Yet, in general, not much has changed.
For a company with 5,000 employees, deploying Microsoft 365 Copilot company-wide means spending an additional $1.8 million annually – a substantial sum. But there is yet no answer whether the productivity gains justify the cost.
The Measurement Problem
More than 40% of companies struggle to define and measure the impact of generative AI initiatives, according to Deloitte's State of Generative AI in the Enterprise research. Under half have developed key performance indicators to effectively measure the return on AI. Many report that standard success measures don't work for evaluating AI productivity tools.
Finance teams evaluating multimillion-pound productivity software investments expect calculations, such as hours saved per employee and efficiency gains in specific workflows.
Standard ROI models fail to capture the value of AI tools integrated into daily workflows.
Traditional Microsoft 365 applications provide clear, measurable value: email enables communication, spreadsheets perform calculations. Microsoft 365 Copilot offers enhancements to these workflows. The software drafts email responses based on thread history, creates presentation outlines from prompts, generates Excel formulas from natural language requests, and summarises lengthy documents or meeting transcripts.
The features make existing work faster or easier. However, most organisations can function without them. Copilot presents a harder value proposition to quantify than core applications.
The Pricing Pressure
At $30 per user per month, Microsoft 365 Copilot uses a per-seat subscription model that compounds budgetary pressure. And unlike previous Microsoft 365 features that were bundled into existing subscriptions at modest price increases, Copilot requires a separate purchasing decision for each user.
Enterprise Agreement customers accustomed to tiered discounts based on spending volume found Microsoft's single-price approach particularly jarring. No volume discounts. No negotiation leverage.
The flat pricing structure has demonstrably affected adoption. In an attempt to address the issue, Microsoft introduced a consumption-based alternative, Microsoft 365 Copilot Chat, in January 2025, specifically to address enterprise resistance. As CNBC reported, "Some management teams have resisted paying Microsoft to give the 365 Copilot to thousands of employees because they weren't sure how helpful it would be at the $30 monthly price."
The new consumption-based model allows organisations to pay only for actual usage rather than committing to per-seat subscriptions. Microsoft's chief marketing officer for AI at work, Jared Spataro, said the new pricing addressed uncertainty: "As one customer said to me, this model lets the business value prove itself."
Indeed, this new model provides a reasonable alternative. Still, its impact has yet to be seen.
🖐 Reassess your Microsoft contract terms before scaling AI licences. Discover: Microsoft Enterprise Agreement Negotiation.
What Adoption Actually Looks Like
Microsoft stated that daily active users more than doubled quarter-over-quarter as of October 2024, whilst refusing to disclose actual numbers.
UBS reported having 50,000 Microsoft 365 Copilot licences in October 2024. Accenture committed to 200,000 users in November. The US House of Representatives announced in September 2025 that it would make up to 6,000 licences available for a one-year trial.
Organisations we advise report the same experience: pilot programmes that extend month after month without progressing to full deployment, as finance teams demand clearer ROI evidence before approving wider rollouts.
Most enterprises remain stuck in prolonged pilot phases as finance teams demand clearer proof of Copilot’s value.
Large organisations approach Copilot through pilots, trials, and phased rollouts rather than immediate company-wide adoption. The House of Representatives structured its deployment explicitly as a trial with evaluation metrics, even with government backing.
Industry analysts characterised Microsoft's Copilot push as "slow/underwhelming", according to CNBC reporting in January 2025. Another analysis noted that "some customers have been hesitant to pay more for Office AI after bugs in experimental features", pointing to quality concerns alongside pricing questions.
The Productivity Studies
Microsoft has published its own research on Copilot's productivity impact. A study of 297 early access users found that 70% reported being more productive and 68% claimed improved work quality. Users were 29% faster across a series of tasks involving searching, writing, and summarising.
The UK government's cross-department Microsoft 365 Copilot experiment found that trial participants saved an average of 26 minutes daily. Over 70% agreed that Copilot reduced time spent searching for information and performing mundane tasks.
However, the UK study also noted limitations. The quality of Copilot's output required editing in most cases, reducing the claimed efficiency gains. Benefits varied significantly by job function and grade level.
The studies rely on self-reported productivity improvements from early adopters in controlled conditions. How these gains translate to real organisations with real constraints remains unknown.
The ROI Question Without An Answer
Finance teams cannot convert productivity metrics into financial justification. If Copilot saves employees 26 minutes daily on average, as the UK study suggested, what does that translate to in business value?
The calculation is borderline impossible. Time saved doesn't automatically equal cost reduction unless it results in fewer staff hours required. Time saved doesn't automatically translate to a revenue increase unless staff utilise the freed time for higher-value work that generates income.
Enterprise clients consistently tell us the same story: they cannot justify scaling Copilot deployment when they cannot demonstrate value in terms that their CFOs accept.
Microsoft's own Copilot Dashboard calculates "Copilot-assisted value" by multiplying assisted hours by an average hourly rate (defaulting to $72 based on US Bureau of Labor Statistics data). But this assumes that all time saved has equal economic value, which does not reflect how organisations actually operate.
Beyond the economic assumptions, the measurement approach is flawed. Copilot integrates so deeply into Office applications that distinguishing deliberate usage from passive interactions is nearly impossible. The software can register activity even when users haven't deliberately used it.
Organisations base purchasing decisions on these metrics. Microsoft's Copilot Dashboard and Viva Insights reports supply the primary data that many enterprises use to evaluate ROI. The measurements come from Microsoft itself. The vendor controls both the product and the metrics used to justify purchasing more of it.
Enterprises rely on Microsoft’s own data to assess Copilot ROI, with no independent validation available.
No independent audits of Microsoft 365 Copilot productivity claims exist. Enterprises must either trust data from Microsoft itself, conduct their own measurement programmes to verify Microsoft's claims, or conduct their own measurement programmes to verify the claims.
Even setting aside measurement accuracy, projected ROI figures are unreliable. Forrester research commissioned and paid for by Microsoft conveniently projected ROI ranging from 132% to 353% for small and medium businesses over three years. However, these projections model hypothetical benefits rather than measuring actual returns from deployed implementations.
The gap between theoretical ROI models and practical budget justification explains why 50% of technology leaders in the CNBC survey answered "too soon to know" about whether Copilot justified its cost. Without clear metrics tied to their specific organisational contexts, technology leaders cannot confidently make the business case to finance teams.
🖐 Evaluate your Microsoft 365 costs with independent insight. Learn more: Microsoft 365 Planning and Optimisation.
The Uncertainty Problem
Microsoft 365 Copilot operates in an AI market where pricing models, capabilities, and competitive offerings are constantly evolving. Measurement challenges and unclear long-term value create rational hesitancy among enterprise buyers.
The Deloitte research found that whilst 74% of enterprises report that their most advanced generative AI initiatives meet or exceed ROI expectations, organisations acknowledge needing at least a year to overcome adoption challenges, including governance, training, and integration. The measurement problem is widespread.
Microsoft's introduction of consumption-based pricing acknowledges the resistance. The pay-as-you-go model offers a chance to avoid committing to thousands of $30 monthly subscriptions without proven value. But is it a fix? That’s yet to be realised.
Do AI-enhanced productivity tools justify premium pricing when productivity gains cannot be measured? Two years after launch, half of the technology leaders still cannot answer. Not "maybe yes," not "maybe no." They genuinely do not know.
Two years after launch, enterprises still cannot determine whether Copilot delivers measurable business value.
That's not what success looks like for a product that Microsoft positioned as transformational. Yet Microsoft collects $30 per user per month regardless, and enterprises that are funding year-long experiments deserve independent validation of productivity claims before committing to company-wide deployment.