How user cohorts reveal what's really driving AI adoption across your organization
How user cohorts reveal what's really driving AI adoption across your organization
TLDR
→ McKinsey's 2025 State of AI report found that function-level ROI from AI varies significantly across the enterprise. Software engineering and IT report 10 to 20% cost reductions. Marketing and product development show revenue uplift above 10%. The same investment produces different returns depending on who is using it. → Cohort analysis, segmenting users by department, role, geography, and seniority, reveals which groups are generating value, which are not, and why. This is the visibility that makes targeted improvement possible. → Adoption trajectories diverge across cohorts over time. Departments whose usage deepens are generating genuine productivity gains. Departments whose session depth is declining are giving an early signal that the workflow fit is not right. Seeing this early is what makes intervention possible. → Geographic cohorts surface adoption patterns invisible at the company level. A tool that performs inconsistently for specific languages or regions will show this in geographic data before it shows up anywhere else. → Cohort-level measurement changes the quality of AI ROI reporting. It answers not just whether the deployment is growing, but which functions are generating return, how deeply, and where the next investment should focus.
Enterprise AI adoption is rarely uniform. The legal team that uses the AI agent every day to draft and review contracts is having a completely different experience from the operations team that logs in once a week for simple lookups. The London office that has embedded the AI tool into its morning workflow looks nothing like the São Paulo office that is still figuring out how to make it useful.
Understanding which groups are getting value, how much, and why, is what turns an AI deployment into a measurable business programme.
According to McKinsey's 2025 State of AI report, function-level ROI from AI varies significantly across the enterprise. Software engineering and IT report 10 to 20% cost reductions. Marketing and product development show revenue uplift above 10%. The same AI investment produces different returns depending on who is using it and for what. That variation is not random. It reflects workflow fit, adoption depth, and how well the AI agent has been configured for each group's specific needs.
Organizations that can see this variation can act on it. Those that cannot are making investment decisions based on a picture that averages over the most important signal.
What cohort analysis reveals
A cohort is a defined group of users, segmented by department, role, seniority, geography, or any other organizational dimension that is meaningful for your context. Tracking adoption and outcomes by cohort gives you a view of AI performance that a single company-wide number cannot.
At the department level, cohort analysis shows which functions have genuinely embedded the AI agent into daily workflows and which are engaging with it peripherally. In most enterprise deployments, a small number of departments account for a disproportionate share of productive use. Knowing which ones, and understanding what they have in common, tells you both where to scale and what conditions make scaling possible.
At the role level, cohort analysis reveals how different job functions interact with the AI agent. Senior leaders tend to use AI for synthesis and strategic questions. Individual contributors use it for task-specific work. New employees use it more extensively as they learn workflows. Each group has different needs, different expectations, and different definitions of success. Treating all of them as a single user base produces support and training interventions that fit none of them precisely. Treating them as distinct cohorts makes targeted improvement possible.
At the geography level, cohort analysis surfaces adoption patterns that would otherwise remain invisible. A multinational enterprise where the AI agent performs well for English-language queries but inconsistently for other languages will see this reflected in geographic adoption data. Teams in those regions are not failing to adopt the tool. The tool is failing to serve them. That distinction changes what the right response is, and you can only see it at the geographic cohort level.
How adoption evolves differently across cohorts
One of the most valuable things cohort analysis reveals is how adoption trajectories diverge over time. Different groups follow different paths after a deployment launches.
Some departments enter an adoption cycle that deepens over time. Users ask progressively more complex questions. They apply the AI agent to higher-value tasks. Return rates stay high and conversation depth grows. These groups are generating the function-level productivity gains that show up in business outcomes.
Other departments follow a different pattern. Initial curiosity gives way to declining engagement. Sessions get shorter. Return rates drop. The AI agent becomes something people use when it is convenient but avoid for anything that matters. This pattern often reflects a mismatch between what the tool does well and what that department actually needs.
Seeing these trajectories by cohort, and catching the divergence early, is what makes proactive intervention possible. A department showing declining session depth in week four is giving a clear and actionable signal. That signal is invisible in company-wide usage figures until the problem has already compounded.
The organizational dynamics cohorts reveal
Cohort analysis also surfaces organizational dynamics that usage data alone cannot explain. McKinsey's research consistently finds that AI value is concentrated in organizations that redesign workflows rather than layering AI onto existing processes. High performers are 2.8 times more likely to have fundamentally redesigned workflows around AI.
At the cohort level, this shows up as a pattern: departments that have changed how they work generate measurably more value from the same AI tool than departments that have not. The difference is not the technology. It is the workflow. Cohort data makes this visible by showing not just who is using the AI agent, but how their usage is evolving and what it is producing.
Geographic and cultural variation adds another layer. Adoption in different office locations, across different languages, and in different regulatory environments reflects factors that a single adoption metric will never surface. An AI deployment that is performing well on average can be underserving specific regions in ways that have real business and governance implications. Cohort analysis by geography is what makes those implications visible before they become problems.
What this means for reporting AI ROI
For enterprise leaders who need to report AI value to boards and senior stakeholders, cohort-level measurement changes the quality of the conversation.
Company-wide figures answer deployment questions. A board presentation that shows departmental adoption trajectories, hours saved by function, and AI proficiency growing in specific cohorts over time answers the business questions that matter: where is this investment generating return, where should we focus next, and how confident are we that the productivity gains are real and durable.
Deloitte's 2026 State of AI in the Enterprise report, based on a survey of 3,235 senior leaders, found that only 34% of organizations are truly redesigning business processes around AI, despite two-thirds reporting efficiency gains. The gap between nominal use and genuine transformation is exactly where cohort analysis does its most important work. It shows not just that departments are using the AI agent, but whether and how deeply they are changing how they work because of it.
That level of visibility is what turns an AI deployment from a cost centre into a programme with a defensible, evidence-based business case.
Nebuly
Nebuly is the ROI platform for enterprise AI. It connects to the AI agents your business runs on, the assistants your customers interact with, and the tools your employees use every day, including Claude, ChatGPT, and Copilot, and translates that activity into business value. How much time is being saved across teams. What revenue your AI is influencing. What adoption and AI proficiency look like in practice, across departments and geographies. All aggregated at the organizational level, never tied to individuals.
If you need clarity on what your AI investment is actually delivering, book a demo.
FAQs
What is user cohort analysis in the context of enterprise AI?
User cohort analysis segments AI users into defined groups, typically by department, role, geography, or seniority, and tracks adoption patterns, usage depth, and business outcomes for each group separately. It gives enterprise leaders visibility into how different parts of the organization are using and benefiting from AI, rather than a single company-wide average that flattens meaningful variation.
Why does AI adoption vary so much by department and role?
Different departments have different workflows, different task types, and different definitions of what the AI needs to do to be useful. McKinsey's research consistently finds that the departments generating the most value from AI are those that have redesigned workflows to take advantage of what AI does well, rather than layering it onto existing processes. Role-based variation reflects different use cases: senior leaders use AI for synthesis, individual contributors for task execution, new employees for onboarding. Each group needs different things from the same tool.
How does geographic cohort analysis help enterprise AI leaders?
Geographic cohort analysis surfaces adoption patterns that are invisible at the company level. An AI agent that performs well for English-language queries but inconsistently for other languages will show declining adoption in specific regions. Without geographic segmentation, this averages out and goes undetected. With it, the issue is identifiable and addressable. In regulated industries, geographic cohorts also help surface compliance and governance variation across jurisdictions.
What does adoption trajectory data reveal about AI value?
Adoption trajectory tracks how usage evolves over time for a specific cohort, not just whether the tool is being used. Departments whose session depth grows, whose return rates stay high, and whose users apply the AI to progressively more complex tasks are generating genuine productivity gains. Departments whose engagement is plateauing or declining after an initial period are showing a signal that the workflow fit or the tool configuration needs attention. Trajectory data catches this divergence early, while there is still time to intervene effectively.
How does cohort-level measurement improve AI ROI reporting to leadership?
Company-wide figures answer deployment questions: is usage growing, is satisfaction positive. Cohort-level measurement answers the business questions boards and CFOs actually need answered: which functions are generating measurable return, how deeply has AI changed how those teams work, where is the next investment most likely to produce results. McKinsey found that AI high performers, those attributing more than 5% of EBIT to AI, are significantly more likely to have systematic measurement practices at the function level. The measurement approach is part of what makes them high performers.
Stay up to date on what we're learning, building, and seeing as enterprise teams deploy and measure AI agents in production.