Standing at the crossroads - PwC on the difficulties of getting productivity gains from Copilot

There continues to be speculation that Microsoft Copilot may be an early AI market casualty as enterprise adoption rates struggle to keep pace with Microsoft’s ambition for the tool. 

On Microsoft’s recent earnings call CEO Satya Nadella said Microsoft 365 Copilot now has 15 million paid seats. This is out of a customer base of 450 million subscribers for Office 365, which works out at around three percent uptake of Copilot.

Wanting to find out more about Copilot enterprise adoption I took the opportunity to speak with Stephanie Mosticchio, Global Microsoft Alliance Leader for PwC, just after she had returned from attending Microsoft’s AI Tour New York event.

It can be difficult to gather productivity gains from Copilot

I began by asking Mosticchio what her main takeaways were from the New York event. Her reply was that:

Customers are confused about their focus and where they are getting the value from AI – the productivity piece of what AI is doing.

PwC itself has 270,000 users on Copilot and Mosticchio reports that between October and January they had 49 million interactions using it. She suggests that in terms of metrics, each interaction saves that individual two to six minutes of time. She acknowledges that other areas are more measurable within a business, that, for example, call centers using an agentic tool can save a lot of time. With O365+ some companies are moving super-fast with adoption, but Copilot needs to be applied to areas with a more metric-based process to see the productivity gains.

Mosticchio suggests part of the problem is that Copilot is embedded in many direct human interaction processes. As she says:

Are you comfortable with an AI agent reaching out on your behalf? Would I use an AI agent to email execs at Microsoft? Do I want it to send this autonomously? No, I do not. For productivity gains you need to look at AI agents in supply chain apps or predictive maintenance apps. You need to take use cases and trudge through what that looks like. Then take the top three to five things to do and run with them. But you do need a groundswell from the bottom of the organization up as well as from the top down.”

While PwC has embraced the tools and Copilot, Mosticchio feels that, “the rise will come from new employees who are maybe younger because in a senior role there is not enough time to play around with everything although Foundry and Copilot Studio do make it easier to experiment.”

Customers are deploying and then not turning on

In 2025, Microsoft declared the emergence of the Frontier Firm that looks markedly different from organizations of the past. As defined by Microsoft in its ‘Work Trend Index Annual Report 2025: The Year the Frontier Firm is Born’:

These firms are structured around on-demand intelligence and powered by “hybrid” teams of humans plus agents, these companies scale rapidly, operate with agility, and generate value faster. Frontier Firms are already taking shape, and within the next 2–5 years we expect that every organization will be on their journey to becoming one.”

PwC is part of the Microsoft Frontier Firm program, which means they get access to all Microsoft early releases. Mosticchio explains:

We take things super-early and things are a little bit buggier. We are re-imagining our business, as technology is core and central to what we do for clients, enabling agents is important. It is super important for us to see what is happening early, so that we can direct customers to what is right for them.

With Copilot she feels that:

Finding those who want to play around with it is important as it helps people become less nervous. For example, for me it has helped me prepare staff reviews, and create partner plans. You can just load, say, six documents in and ask it to suggest areas to discuss and why. This gives you significant amounts of time back, so you don’t have to do this work in your off-time. My CTO inside the Alliance product engineering team made these suggestions. This type of informal sharing is really important. We now have a repository to share examples of how we have used it and we are building similar capabilities for customers.

With Copilot some customers are deploying it and then not turning it on. We need more education around it generally and Microsoft is looking to companies like PwC to hear examples of how it is being used. What it enables you to do is more things throughout your day. Use cases that are working need to be explained better.

I asked why she thinks Copilot is not growing as fast as expected. She believes it is because there is a lot of competition, the market is crowded, tools are expensive and a lot of internal training is required. Having said that she also said that she has seen 200% growth in pipeline for Copilot over the past six months.

Her parting words were that:

Organizations need to start somewhere and they should just go and do it. Take 10 use cases and if six fail, run with the four that work.

My take

I think Mosticchio is correct that the challenge with AI adoption at the moment is finding the ROI, and that it is easier to find this in business areas where embedding agents in the process yields obvious cost benefit analysis, areas such as call-centers, or field services. The challenge for Copilot is that it is largely being deployed as part of Office 365, software that is so ubiquitous that it is generally upgraded for everyone with minimal or no training. Small wonder that organizations that switch Copilot on then struggle to see where returns are coming from. Employees that are using it are probably getting small increments of time savings that mean they do not have to do so much in their own (unpaid) time. And putting a metric around that, is a real ROI conundrum for the enterprise.