• It has been about one year since Microsoft began offering its Microsoft 365 Copilot AI as an enterprise solution to companies, and adoption is growing rapidly.
• But many companies remain in the testing phase, and many say they are still not certain the AI is worth the cost, according to the latest bi-annual survey of the CNBC Technology Executive Council.
• OpenAI's own enterprise product which runs on the Microsoft Azure cloud and Meta's Llama are other gen AI tools that companies indicate they are using.

Roughly one year after its introduction and forecasts that it would add billions of dollars in annual sales, use of Microsoft's flagship generative AI tool for the office market, Microsoft 365 Copilot, is growing rapidly within corporations, but many remain on the fence when it comes to full deployment.

That's according to the latest bi-annual survey of technology executives on the CNBC Technology Executive Council, which is a sampling of views from tech leaders including CTOs and CISOs at a wide range of companies, from retailers to industrials, health care, energy and financial services firms.

Virtually no firms are on the fence when it comes to AI adoption, broadly speaking. Roughly two-thirds (63%) of the tech executives surveyed say new investments in AI are accelerating, while another third say they are evaluating AI investments but describe their efforts as cautious.

Among the clearest findings from the CNBC survey is how far and fast Microsoft's Copilot, introduced roughly one year ago, has emerged as the early winner within the corporate enterprise, with 79% of survey respondents saying their company is using it.

But the numbers on full, permanent deployment aren't as clear a win for the tech giant, at least not yet. While 50% of survey respondents said their organizations have made the decision to deploy the tech to all employees, another 17% say they did not decide to fully adopt Copilot, and another a third of companies say they remain in the testing phase.

When asked whether it's been worth the cost ($30 monthly per user in the case of Copilot), tech leaders say they aren't sure. An equal one-quarter of respondents answered either yes or no to that question, while 50% of respondents said it is still too soon to know. Those numbers should trend higher over time, according to the CNBC survey findings, with only one-third of companies saying they have no plans to use Copilot in the future.

According to recent Deloitte research, two-thirds of organizations are increasing gen AI investments based on "strong early value to date." However, when it comes to scaling that value, a different picture emerges. Many enterprise AI projects remain in what Deloitte refers to as the pilot or proof-of-concept stage, with over two-thirds of organizations saying they have moved 30% or fewer of their gen AI experiments into full production.

There are multiple reasons why companies will still be short on confidence when it comes to enterprise AI adoption, according to Deloitte. More than 40% of companies are struggling to define and measure the impact of gen AI initiatives, and under half have developed key performance indicators to measure AI return effectively, while many say standard success measures don't work.

Many companies remain unconvinced that they have the infrastructure in place to benefit from AI adoption, according to Deloitte's research, and that includes concerns that their organizations lack sufficient data needed for custom models to go into production. Risks, regulations (such as those in Europe), and governance issues are also weighing heavily on companies when it comes to moving with more speed to scale gen AI.

How the gen AI competition stacks up against Microsoft

Microsoft, which has been holding adoption data relatively close to the vest, has publicly said the number of people who use Copilot daily at work nearly doubled quarter-over-quarter, and nearly 60% of the Fortune 500 now uses Copilot. The number of customers with more than 10,000 seats more than doubled quarter-over-quarter, including Capital Group, Disney, Dow, Kyndryl, and Novartis. At a price of $30 per month per user, back-of-the-envelope math equates to $3.6 million annually for 10,000 employees. 

Jim Rowan, who is head of AI and principal at Deloitte Consulting, says companies across the economy have learned from their cloud adoption experiences that the total cost of ownership can get expensive quickly, and that experience is causing companies to be cautious about "overcommitting" to vendors. 

Microsoft is not alone in enterprise wins, though it has a significant lead, according to the CNBC survey. The Azure cloud offering from its biggest AI partner and AI investment, OpenAI, is in use at nearly two-thirds of companies. Meta's Llama large language model is the next most popular gen AI for the corporate enterprise, in use with 38% of firms, followed by Anthropic (13%), and Mistral — the French AI startup backed by many top Silicon Valley VC firms, as well as Microsoft and former Google chairman Eric Schmidt — which is in use at 8% of companies.

Microsoft continues to spread its own bets within gen AI, announcing on Tuesday that its GitHub subsidiary will let developers use models from OpenAI competitors like Anthropic and Google for its chatbot tool designed for coding.

A recent Morgan Stanley study on Copilot found 94% of CIOs expected to adopt Microsoft generative AI products over the next year, up from 63% in Q4 2023. Microsoft 365 Copilot was cited by 68% of CIOs; followed by Azure OpenAI Services, cited by 41% of CIOs.

It is not a surprise to see Microsoft with a lead in the enterprise AI race as it can become harder for firms to make the investment case, at least in the early days, the farther away it moves from Microsoft 365 Copilot. That tool is an example of one that fits into an existing tech platform already rolled out across the enterprise and embedded in the way we work. For many AIs, Rowan said, "there is no really easy button to press."

With an enterprise-wide solution like Copilot, Rowan says it's still critical to put as much thought into specific training and engagement of employees as into the technology investment itself, or it will be hard to ultimately justify the cost. "There needs to be a continual message around AI from leadership, on how they are using it, and how employees are using it," he said.

Forty-six percent of CNBC survey respondents said that within their organizations at least half of employees are now using AI. Another 42% of respondents say one-quarter of their workforce is using AI. According to the survey, most organizations (84%) do believe that some suite of gen AI tools already fits, or will fit, into their business plans.

At this point in the AI adoption curve, it is not logically inconsistent for executives to say they are not sure what the KPIs or ROI will be, but that their firms need to be "on this AI journey," according to Rowan.

There will be some quick wins, such as chatbots which are being widely deployed by companies for both customers and employees, but Rowan said it's not the quick wins which should dictate long-term thinking, but rather, a core approach to R&D decision making that has not changed. There is always a question of where a firm should start, he says, but there isn't the question of whether the R&D effort should get started. "Start with a set of use cases and get in AI value creation mode," Rowan said. "Fail fast, make changes, and address." 

The alternative, in his view, is worse. "Companies risk missing the moment where tech moves super-fast and you're now behind," he said.

The bi-annual survey was conducted among a sampling of members on the CNBC Technology Executive Council between Oct. 8-Oct. 21.