Executive summary:
AI value depends on the progression from adoption to sustained activity that ultimately yields measurable returns. Grant Thornton combined training, role-specific use cases and ongoing support to achieve 84% adoption and 12.5 million actions that are yielding process-level returns. As employees integrate AI into their work, they can identify opportunities to redesign or create processes, establish controls and measure results against credible baselines, turning everyday experimentation into repeatable, scalable business value and an AI value story.
Beyond deployment
Generative AI requires an implementation that is unlike other technologies.
Many organizations have been anxious to capture the power of AI. They deployed access and watched for the benefits to follow. But leaders are finding that access does not create value on its own. Utilization stalls, enthusiasm fades and even promising pilots struggle to scale.
AI is a dynamic tool that drives value in different ways for every user, depending on their unique role, processes and other factors. It can be difficult to tell users how to achieve the most value for their work, but organizations do need to guide success. Leaders must decide how to:
- Drive adoption: Adoption is the prerequisite.
- Foster activity: Ongoing activity refines use, to show where tools can empower processes and meet needs.
- Track returns: Value emerges when people build tools into processes, meet needs and measure the difference.
Grant Thornton confronted this challenge. The firm approached Microsoft 365 Copilot as a workforce and operating-model change.
Over the past year, Grant Thornton sustained 84% adoption, recorded 12.5 million Copilot actions and saw users create more than 10,000 agents. Those figures tell a progression story: broad adoption, significant activity and measurable business returns.
Adoption is the start
Adoption of generative AI varies across a range of factors, but it is growing. In the last three years, generative AI has reached 23.8% adoption among the U.S. population (53% globally), according to The 2026 AI Index Report by Stanford University and Human-Centered Artificial Intelligence.
The intuitive Copilot interface and user curiosity drove early adoption, but continued growth required ongoing training, open question sessions, a champion network, on-demand resources and a framework for users to engage their own exploration.
The 84% adoption rate at Grant Thornton required more than just deployment. The intuitive Copilot interface and user curiosity drove early adoption, but continued growth required ongoing training, open question sessions, a champion network, on-demand resources and a framework for users to engage their own exploration.
“The biggest thing that we encourage is to think of Copilot as a person. If you ask a person to do something, think about the direction, knowledge and information you should provide,” said Grant Thornton Enterprise Architecture Director Tim Zavertnik.
This perspective can change the aim of training. Feature instruction remains useful, but lasting adoption requires a new way of thinking about how to frame a request, provide context, evaluate an output and refine the interaction.
That mindset remains useful even as specific interfaces and capabilities continue to evolve.
Once users had the right mindset, Grant Thornton needed to help them find their unique use cases. These use cases drive the daily Copilot activity that ultimately yields business value.
How we can help you
SERVICE
SERVICE
Activity is a leading indicator
An AI business case starts with a simple condition: People must use the technology enough to learn where it fits. “We had to get to a granular level, to go beyond doing a standard demo,” Zavertnik said.
People needed repeated opportunities to apply AI to familiar work, compare approaches with colleagues and see examples in their own professional language.
People needed repeated opportunities to apply AI to familiar work, compare approaches with colleagues and see examples in their own professional language. A tax professional, an auditor and a marketer may use the same underlying capability, but each needs to recognize the task, source material and expected output as relevant to their role.
Once users recognized these use cases and built AI capabilities into their daily work, Grant Thornton saw their Copilot activity metrics continue to rise.
Grant Thornton users completed 12.5 million actions in Microsoft 365 Copilot over the past year, creating more than 10,000 agents across Microsoft 365 Copilot Chat, SharePoint and Copilot Studio. A vendor-standard estimate of six minutes per action would put that near 1.2 million hours, though Grant Thornton deliberately does not book that figure as value because an assumption applied at scale is not a measured return.
These figures measure use. They show that employees are using Copilot at scale and building new ways to work, but they do not prove returns on their own. Adoption and activity are leading indicators because they identify where behavior has changed and where deeper measurement should begin.
Returns are defined by processes
Hard-dollar value becomes visible when a defined process is rebuilt or created around AI and measured against a credible baseline.
Hard-dollar value becomes visible when a defined process is rebuilt or created around AI and measured against a credible baseline. The best candidates can be processes like financial close, since they have identifiable inputs, repeatable steps, accountable owners and measurable outputs.
Redesigning processes
In a process redesign, start with current process metrics for elapsed time, labor hours, handoffs, rework, quality measures and cost. Then, define the AI-enabled process, establish controls and compare performance over a meaningful period. Relevant measures may include cycle time, cost per transaction, throughput, first-pass quality, rework and the amount of human review required.
A practical measurement sequence
| Stage | Question | Example evidence |
|---|---|---|
| Baseline | How does the process perform today? | Cycle time, labor hours, rework, quality and cost |
| Redesign | Where can Copilot or an agent change the workflow? | Removed handoffs, faster synthesis, draft generation or automated routing |
| Control | How will quality, security and accountability be maintained? | Human review, permissions, testing, exception handling and audit trail |
| Measure | What changed against the baseline? | Time, cost, throughput, quality and capacity redeployed |
| Scale | Is the result repeatable across teams? | Stable performance, documented ownership and sustained adoption |
The process level can create a natural boundary for governance. Teams can specify approved data, required reviews, exception paths and ownership before they scale an AI-enabled workflow. This connects value and control rather than treating them as separate workstreams.
Creating processes
When people within your organization have an AI mindset, they can even use the technology to solve needs that have no current solution. “A practitioner can identify a problem, outline how they would want it solved and use AI to help bridge that gap,” Zavertnik said. “Once they’ve proved it works, and made it repeatable, then they make it scalable.”
A new opportunity
As an example of innovation with Copilot, Zavertnik cited a Grant Thornton team that was asked to define the requirements for a company’s new CRM solution. The company needed help with capturing its manual workflows in wireframes and requirements documentation that it could provide to the CRM development team.
Rather than automating the traditional process, the Grant Thornton team saw a way to be more dynamic and effective. The team used M365 Copilot to quickly create an interactive HTML prototype that went beyond traditional wireframes, providing a platform for stakeholder feedback and data structure. The prototype let stakeholders experience how the information would move through the system, significantly improving their engagement, informing conversations about workflow design and data management, and ultimately accelerating requirement validation. Then, the CRM development team adopted the prototype as the foundation for the actual solution build.
Based on similar projects, the team estimated the traditional wireframe process at 2.5 weeks, including 1.5 weeks for the initial wireframes, plus a week to gather feedback and incorporate changes. With Copilot, the team reduced that work to one week. After the initial mockup, changes could be incorporated in a few hours or even minutes by using natural language input rather than manually updating and repositioning objects in a traditional wireframe.
This approach reduced a 2.5-week cycle to one week, improved requirements quality, gathered stakeholder inputs before they became costly rework, and accelerated the entire project from concept to implementation.
With measurable returns captured, three points converge:
- Pervasive adoption creates a population capable of spotting opportunities.
- Sustained activity builds the muscle to rebuild or create new AI-driven processes.
- Returns emerge when teams can identify baselines and measurable outcomes.
Your AI value story
Leaders can write their AI value story by tracing three questions: Are people adopting the tools? Are they using them to change their work? Are they producing measurable returns?
Leaders can write their AI value story by tracing three questions: Are people adopting the tools? Are they using them to change their work? Are they producing measurable returns?
Each question builds on the one before it, and the Grant Thornton experience illustrates why the sequence matters. The pursuit of AI-driven value comes from building or improving processes, then measuring the results. “Start small,” Zavertnik suggested. “Start simple. Even if you just have Copilot do simple things, the value compounds when you do them enough.”
High action counts also point leaders to recurring tasks that are worth standardizing. Agent creation can reveal the processes that employees are already trying to redesign. Those activity signals can help leaders prioritize use cases for controlled measurement. Those measurements can become your AI value story.
AI value is not created at deployment. It is created through leadership commitment to drive AI adoption, activity and measurement. Grant Thornton built this sequence inside its own firm before taking it to clients, which is why the firm can show baselines, controls and results, not just the activity counts.
Contacts:
Content disclaimer
This Grant Thornton Advisors LLC content provides information and comments on current issues and developments. It is not a comprehensive analysis of the subject matter covered. It is not, and should not be construed as, accounting, legal, tax, or professional advice provided by Grant Thornton Advisors LLC. All relevant facts and circumstances, including the pertinent authoritative literature, need to be considered to arrive at conclusions that comply with matters addressed in this content.
Grant Thornton Advisors LLC and its subsidiary entities are not licensed CPA firms.
For additional information on topics covered in this content, contact a Grant Thornton Advisors LLC professional.
Ready to talk? We’re ready to listen.
Request a meeting and a member of our team will be in touch to see what we can do to meet your needs.
Want to submit an RFP? Please submit your request through our RFP submission page.
Trending topics
No Results Found. Please search again using different keywords and/or filters.
Share with your network
Share