Explore the Levels of Change Management

Latest Posts Individuals Projects and Initiatives Enterprise Archives

How Are Companies Measuring AI Success in Organization-Wide Adoption

7 Mins

Published: May 28, 2026

Organizations are moving fast to deploy AI tools, but don’t have the right frameworks and strategies to measure and understand whether their deployments are adding value.

In the absence of a coherent measurement framework, leaders have no reliable way of knowing what’s happening or where they need to support employees. The instinct is to track visible, quantifiable data points, such as system logins, but those metrics measure overall activity, rather than changed behavior.

Prosci research on AI adoption across the enterprise shows that 38% of AI adoption challenges stem from user proficiency issues, making it the greatest barrier between deployment activity and actual adoption. System logins could demonstrate employee use of the system, but actual usage might go overlooked without the right framework to capture these human challenges.

For organizations to truly measure AI success in terms of organization-wide adoption, they need measurement frameworks that account for the human side of change.

The measurement problem most organizations are getting wrong

The AI adoption metrics that are easiest to report and track are those that indicate deployment completion, not behavioral adoption. When enterprise AI initiatives stall, most organizations evaluate the metrics most readily available to them rather than those that reveal the barriers requiring attention.

The measures most organizations rely on during AI rollouts include:

License provisioning and access rates – The number of employees who have access to an AI tool.
Training completion rates – The percentage of employees who complete assigned onboarding and training sessions.
Feature utilization counts – The frequency with which employees are opening or interacting with various aspects of the AI tool.
Self-reported satisfaction scores – Employee ratings of their experience with the tool or AI training.

These measures point toward activity, not the behavioral differences that drive long-term value realization for organizations. Enterprises must distinguish between leading and lagging indicators in AI adoption: access, training, and feature usage can signal early momentum, while behavior change, workflow integration, and business impact reveal whether adoption is leading to transformation. The measures that actually indicate AI adoption success include:

Behavior change – Whether employees change their behavior and adopt AI tools or resist them entirely.
Proficiency over time – Whether employees get faster, more confident, and more independent in their use of AI tools.
Workflow integration – Whether employees embed AI into how they do their work or view it as a parallel, optional step alongside their workflows.
Reliability and trust – Whether AI use holds when employees feel under pressure, or they revert to previous habits when the stakes are high.
Role-specific capability – Whether employees use AI in meaningful ways unique to their specific function.

Organizations that focus on deployment metrics increase their risk of “shelfware.” “Shelfware” risk refers to licensed tools that go unused because they fail to measure actual adoption. This is not a technology failure but rather a measurement issue.

Moving from deployment-only to actual adoption metrics means acknowledging that tool measurement alone isn’t enough. The measures that matter most include whether each individual uses AI and whether they change their behavior as a result.

Start with a shared definition of AI success

Organizations first need to define what AI success looks like at the organizational, team, and individual levels. AI adoption is not a one-size-fits-all journey; it’s a multifaceted process that requires different approaches at each level.

In our AI adoption research, we found that different organizational levels evaluate the impact of AI differently. Executives skew toward strategic and operational outcomes, like cost savings and new growth opportunities. Team leaders balance tactical needs with strategic requirements, caught between delivering results and meeting organizational expectations. And frontline employees are motivated by practical applications, such as improvements to tasks and response times.

This misalignment explains why many AI transformations stall and fail to achieve the results they expect. Leaders drive strategic implementations while frontline workers want practical solutions.

To address this gap, organizations must define success statements at each level before measurement begins, with these focus areas in mind:

At the organizational level, success statements tie to strategic outcomes, including productivity, competitive positioning, cost efficiency, or risk reduction.
At the team level, success statements tie to outputs and workflows, including the quality of the work product, delivery speed, and reduced rework.
At the individual level, success statements tie to task-level capability and confidence, including whether employees can use a tool independently, effectively, and without friction in their jobs.

In practice, this kind of multi-level structure can function like a balanced scorecard, enabling enterprises to evaluate AI success across organizational, team, and individual dimensions rather than relying on a single adoption metric. Without clearly defined success statements at every level, organizations are left wondering where barriers lie and are unable to address them.

A gray-haired professional man sits at a desk, focused on a computer screen while resting his chin on his hand. Wearing a dark blue shirt, he appears thoughtful and concentrated in a warm office setting, suggesting analytical work, decision-making, or digital productivity.

A four-layer measurement framework for organization-wide AI adoption

Most AI measurement frameworks stop at the surface. They capture what the technology is doing without ever examining whether the people using it have changed how they work and whether the organization has built the conditions for that change to last.

The Four-Layer AI Adoption Measurement Model below organizes AI adoption into four distinct layers, each building on the one before it. Taken together, these layers can serve as a balanced scorecard for enterprise AI adoption, helping leaders track both leading and lagging indicators across activation, behavior change, business impact, and governance.

Layer 1 – Activation and access

Activation metrics are the foundation of The Four-Layer AI Adoption Measurement Model. They answer questions like: Are employees using the AI tool? How many active users are there across the organization? What is our license utilization.

These metrics serve as a baseline. A significant gap between provisioned licenses and active users is an early warning. Consistently low activation rates can indicate an awareness or access problem, as employees may not know the tool is available, may not understand how it applies to their role, or may face technical barriers.

The risk of treating activation metrics as primary success indicators is that they only describe activities, not real value. When activation is the headline metric, organizations routinely declare success when the deployment project closes, long before any evidence of behavior change exists.

Layer 2 – Behavioral change and proficiency

Behavioral change is the layer that determines whether the investment in AI is producing value or producing the appearance of value. It answers questions like: Are employees working differently? Do they use AI tools effectively in their real workflows?

Prosci’s study, Keys to Unlocking AI Adoption, revealed that user proficiency is the single largest challenge in AI adoption, cited by 38% of respondents, more than double the rate of technical integration issues (16%) and organizational adoption challenges (15%). Organizations are not primarily failing at AI because the tools don't work. They are failing because employees lack the proficiency to use them effectively, and enterprises rarely measure that gap with enough specificity to act on.

Top Challenges of Enterprise AI Adoption

Organizations can’t infer proficiency from training completion alone. Enterprises need to define proficiency by role, track it over time, and assess it against the actual tasks that matter for each function. Doing so helps organizations overcome significant barriers through the lens of Prosci’s ADKAR® Model: Knowledge and Ability barriers.

Layer 3 – Business and operational impact

Business and operational impact metrics are the outcomes that most organizations care about and that AI investments aim to deliver: quality improvements, cycle time, throughput, and stakeholder satisfaction.

Prosci research on how team leaders measure AI success paints a clear picture: AI success indicators are dominated by performance improvements (54%), stakeholder satisfaction (13%), and time savings (12%). The challenge is that organizations often set targets for these outcomes at the enterprise level without tracing them back to the role-level behaviors that drive them. A cycle time improvement happens because a specific group of employees changed how they work, consistently and competently enough to move a measurable outcome.

Ultimately, these metrics depend on achieving success in the first two layers.

Layer 4 – Culture, governance, and trust

The fourth and final layer is the most underserved in most measurement frameworks, and increasingly, the one with the highest strategic stakes.

Prosci research draws a clear line between organizations that are succeeding in AI implementation and those that are struggling: the differentiators are not primarily technical. Prosci identifies leadership commitment and clarity as the #1 differentiator between successful and unsuccessful AI adoption, with a +1.65 difference. High-performing organizations score +1.29 in transparency and clarity around AI strategy, while low performers lag at -0.54, creating a communication vacuum that undermines trust.

Understanding employee trust in AI outputs, clarity around AI usage and prohibited use cases, governance compliance, and Shadow AI visibility (unauthorized tool usage) tell leaders whether the conditions for sustained, organization-wide AI adoption are in place.

Segment your measurement by role, not just by tool

Prosci's research across 1,107 participants found measurable differences in how AI is perceived, trusted, and used across organizational levels. Executives report higher trust (+1.09) and ease of use (+1.19) than frontline workers (+0.33 trust; 0.78 ease of use). One-size-fits-all deployments and measurement frameworks mask friction if you don’t measure them meaningfully.

A simple, role-based segmentation model for executives, team leaders, and the frontline can address some of these gaps. Here’s an example:

Executives – AI is informing strategic decisions. Behavioral indicators include: using AI-generated analysis as input to planning and resource allocation; leaders actively model AI use in visible ways; and governance decisions reflecting direct familiarity with how employees use AI tools across the organization.
Team Leaders – Teams embed AI into their work delivery. Behavioral indicators include: AI outputs are being reviewed, applied, and quality-checked as part of the standard workflow; the team leader is actively coaching proficiency rather than assuming training has handled it; and AI use is consistent across the team rather than concentrated in a few enthusiastic individuals.
Frontline – AI use is confident, independent, and provides task-based value in the specific tasks that define the role. Behavioral indicators include: the employee can use the tool without prompting or assistance for workflows, proficiency is improving over time, and the employee can evaluate AI outputs critically.

Segmenting measurement by role does not require a separate framework for each tier. It requires that the success definitions established at the outset (organizational, team, and individual) are reflected in how the organization collects, reports, and acts on data.

What change management has to do with AI success

The organizations measuring AI success most rigorously are treating adoption as a change management challenge, not just a technology deployment. That distinction has measurable consequences. Prosci's Best Practices in Change Management research found that projects with excellent change management achieve their objectives 93% of the time, compared to 15% for projects with poor change management. That’s the difference between crossing the deployment finish line and creating lasting behavioral change.

Prosci’s ADKAR Model — Awareness, Desire, Knowledge, Ability, and Reinforcement — offers a useful diagnostic for practitioners to map measurements to individual adoption stages and distinct interventions to overcome barriers. The specific failure most AI initiatives share is the gap between Knowledge and Ability. Training completion indicates that employees received information, not whether they can apply it independently, under real conditions, in the workflows that matter most. Understanding how to spot these gaps and deploy the right change management strategies to close them sets organizations that succeed apart from those that don’t.

Prosci’s ADKAR Model

ADKAR model outlining five stages: Awareness, Desire, Knowledge, Ability, and Reinforcement to support and sustain change

Organization-wide AI adoption requires measuring the human side of change

The organizations that will get the most from AI are not necessarily the ones that moved fastest or spent the most money on high-quality tools. They are the ones building the discipline to know — with specificity, at every level of the organization — whether their people are actually changing how they work, and what to do when they aren't. That’s the power of change done right.

Prosci

As the global leader in change management, Prosci helps organizations turn complex change into something people understand—so they can act with confidence and deliver results. Built on more than 30 years of research, Prosci partners with enterprises to scale change, enable adoption, and realize outcomes across complex transformations, including ERP and AI. Our work brings clarity and structure to change, helping leaders move from strategy to action and ensure results endure. That’s what change done right looks like.

See all posts from Prosci

Shared Services Center Drechtsteden Improved Customer Satisfaction through Integrated Service Center

From eight separate service counters for the 11 affiliated customer organizations to one central service office. The Drechtsteden Shared Services Center realized their ambition. In the past, each department of the Drechtsteden Shared Services Center had its own service desk (online, call and counter) with its own service levels, tools and processes. These service desks were occupied in turn by employees of the department. Many employees did this work in addition to their ‘regular’ work. As a result, a telephone call was sometimes felt as a disruptive interruption of the daily work. However, this was not the case in the IT and Finance departments, because they had already teams that focused specifically on client services, and that worked very well. The IT service desk, for example, used an advanced service management process and tool, a knowledge management system and the recording of incoming calls. It was therefore decided to introduce this model for the entire services center in order to improve central accessibility for clients (one-stop-shopping and integrated services) and the desire to increase service levels for the entire organization. The immediate ‘why’ for launching this change project was a reorganization of the Facility Services department and a decreasing customer satisfaction. This led to the merger of al service desks into one central and integrated service office. How did you end up at Prosci? “One of their consultants had previously done a job for Drechtsteden, regarding services to the citizen that was a successful change. Nevertheless, there were also two other service providers on our shortlist. We were looking for an adviser who could think in terms of customer impact and guiding the organization though a transition. There was a match with this project/change manager. You would think: a process specialist only focuses on processes, but that turned out to be incorrect. The project manager focused on all relevant aspects of the change; he mastered the entire scope. Both technical and people side. “ "We did not get an answer to our question, but a solution to our problem." — Marvin Suijker, Manager IT What kind resistance have you met in the organization? “We had to ensure that the departments were willing to hand over their first-line client services to us. The starting point was and is the cooperation with these departments, where the best possible service to the customer was our shared focus. Of course there was some hesitation at the beginning, because they assumed they would lose control of quality and their work. Prosci has helped us very well with this. We split the change into smaller projects for all involved departments to chart customer questions, processes and systems. There were of course some people who had their doubts about these developments. We invested most time and energy in the large majority who was willing to embrace the plans without forgetting the others. Everyone undergoes a change in his or her own pace. Those who need longer time can indulge in the exemplary behavior of others who are already further in their individual change process. The change phases are the same, but the pace and needed approach is different per group and individual.“ "The project/change manager has not done a job for us but with us." — Annemarie van de Ven, Direct Supervisor SCD Services How did you get buy-in from the impacted employees? “There was an important role for the Prosci ADKAR® model (Awareness, Desire, Knowledge, Ability, Reinforcement) that was introduced. It all starts with awareness of the need for change. Then the next crucial question raised: ‘what’s in it for you at an individual level?’ That was a moment when sometimes frustration came up. We have noticed that the ADKAR® model is very useful to get all employees step by step on board and make a succesful individual transition. Even though the steps are sometimes small or challenging, the model always works.“ "It all starts with awareness of the need for change. Then the next crucial question raised: what’s in it for you at an individual level?" — Michiel Vrijhof, Customer Services Advisor What happened in practice? “The project/change manager always said: if you want a change to really work for employees, you have to keep talking about the Why and WIIFM (five to seven times). The project/change manager supported on the communication messages and approach for each impacted department and group of employees. We succeeded in taking away the idea that something was taken away from them and on the contrary delivered the insight that the services to our clients would improve. Moreover, he convinced us of the usefulness of sponsorship at executive and direct supervisor level. They openly expressed their support. That helped people to understand the necessity of the new way of working.“ How is the organization doing now? “There are still people who have difficulty with this change. But we recently examined to what grade employees are satisfied with their job and the new way of working. This led to an average score of 7.9. Of course there are people who find it difficult to let go of tasks or take up new activities, but in general people are in the right place and delivering great work.“ "The project team and steering committee has made this change a success by supporting this change with much enthusiasm from their own specific role during this change. Marvin, as a visible and active primary sponsor and building a good sponsorship coalition at board level. Annemarie, by supporting the employees as direct supervisor during the change, which was sometimes perceived and experienced different for each impacted employee. Michiel, by developing the services of the new team step by step together with the different departments and by guiding employees in their journey. An excellent collaboration! It was a great pleasure to be part of this team." — Andy Schoonbroodt, Project / Change Manager When you look back? “Our assignment to the project/change manager was fairly clinical in nature: organize first-line services in a customer-oriented way of working. We now always say: "We did get an answer to our question, but a solution to our problem. And we can also look ahead. At one point the project/change manager consciously stayed in the background and handed us the tools to do it ourselves. The moment he gets out, the services and improvement must of course continue. The project/change manager has not done a job for us but with us. By the way, we still have a help line, because it is our ambition to bring client services to the next level. ” About Shared Services Center Drechtsteden The Drechtsteden Shared Services Center is the shared service center for seven municipalities and four related organisations that together form the Drechtsteden. For these organizations, the SCC provides services in the area of IT, HR, Finance, Purchasing, Information services, Communication, Facility Services and Legal Affairs. To improve the service for their clients (employees of their customer organizations), the SCC decided to set up one central entry point for all customer questions and one integrated services office.

4 Ways to Build Better Change Outcomes With Research and Data

One of my secret weapons to improve organizational outcomes as a Change Advisor is data. As part of our role as change practitioners, we regularly assess, measure and track progress through various data points. But for me, the need for data starts earlier than that as I build commitment and support for change management on projects, help others understand change management, and much more. The Value of Data in Change Management Prosci’s biennial Best Practices in Change Management benchmarking research shares what others in the discipline are doing, the challenges they’re facing, and the successes they’re achieving. The rigor behind the data also helps make change management less “soft,” enables change practitioners to frame the work we do more concretely, and supports adoption and usage in business terms. As an Engagement Leader and member of the Prosci Advisory Services team helping clients and organizations develop change capabilities, I leverage the Best Practices in Change Management data and insights in four distinct ways: 1. Building commitment and support Pulling information directly from the report to influence key stakeholders, either within the project team or with senior leaders, is a common way I use the data. I often share the correlation between highly effective change management and projects coming in on-time, on-budget, and meeting or exceeding objectives. This graph is especially persuasive when you're trying to gain support for a dedicated resource to serve in the change manager role. Even more importantly, this data can be a driving force for building support for change management as an enterprise capability. When talking to senior leaders, this addresses what they most care about in the language that resonates with them. 2. Building understanding and capability in others Driving successful change in individuals and organizations requires new thinking, new models and new tools. It takes education to build both understanding and capability. As noted in the Best Practices in Change Management – 12th Edition, misunderstanding about change management—what it is and isn’t—is an ongoing challenge for the discipline. To address this issue, the first thing I try to do on every project is ensure a level set with certain groups that will support change. And I turn to the research to make sure we’re all on the same page. For example, with project managers, I often talk through the integration of project and change management, as well as the steps and advice for alignment and roles. Executive sponsors and the sponsorship coalition often need to build their own change understanding and capability as well. The research includes great insights on what it takes to be an effective sponsor of change, including traits of effective sponsors, the most critical activities, and common mistakes to avoid. Among research respondents, those with a very effective sponsor were more than twice as likely to meet or exceed project objectives than those who had very ineffective sponsors. The data shows that is clearly a group to support in building their change capability. 3. Building client solutions When working on transformation initiatives, I almost always share the Best Practices in Change Management research to guide clients and help them build out solutions. If you’ve been doing change management even a little while, you know it is common for companies to overestimate how easy it is for individuals to embrace change and underestimate the amount of time, energy and support it takes to get the return on investment from change initiatives. The report’s findings on resources and team structure are particularly helpful in these areas, offering insights to teams struggling to gauge how much support they need and determine what the change reporting support structure should look like. Organizations can also underestimate the level of skill required to effectively manage change. The benchmarking research includes very specific best practices and criteria for identifying who within a company might be good candidates to develop into change practitioners. I often share this data with key stakeholders to help shape their foundation on projects. I also frequently see change practitioners who have zero or very limited access to the sponsor. When that’s the case, I refer to them to the research on sponsor access and change management effectiveness. 4. Building inspiration and personal capability I often find inspiration and ways to build my own change capability within the pages of the Best Practices in Change Management reports. One thing that revved up my creative juices in the latest report is the chapter on sponsorship. This chapter can be used in so many ways. The information on sponsor challenges offers particularly helpful data and insights about possible adaptations for remediation when your sponsor is at the wrong level (too high or too low) to be effective or lacks time and resources, or even when sponsors delegate their role and responsibilities to someone else. To find your inspiration and identify areas for your capability building, I recommend starting with a deep dive into the report—really take time to absorb the information, make notes, and bookmark specific numbers and graphs you might want to keep at your fingertips. I promise you will be inspired to think critically about how you are engaging your sponsors and supporting them in their role with very specific tactics. As for my own growth, I’m most inspired by Part 4: Adapting and Aligning Change Management. This section covers everything from culture considerations to portfolio management. One particularly useful chapter offers actionable insights for navigating specific project types that make change management complex, such as change that spans wide geographic areas or change with little or no WIIFM. As I delve into this content, I am outlining where I want to spend some experimenting and developing my own skills. Best Practices in Change Management Prosci’s research comprises the largest body of knowledge about change management in the industry. Building upon that solid foundation, Best Practices in Change Management – 12th Edition adds modern insights and data to your change management toolkit, so you can help people in your organization understand the importance of change management, the roles they must play, and how to build effective change capabilities. The research also equips and enables you as a practitioner, enhancing your growing capabilities as a leader of change.