Twelve months from now, the reports on ROI from artificial intelligence investments will be quite different.
The business world will have reached the other side of the hype cycle horseshoe. Ambitious boards of directors and even the most eager of evangelists will come to terms with what builders of the technology have known for a while now: AI is not a standalone technology initiative, and therefore it can’t be measured as X dollars spent on AI, Y dollars gained.
That’s not to suggest that the promise of AI is anything less than era defining. Across industries, executives are investing heavily in AI projects, eager to unlock the promise of higher-level work and transformative outcomes. Global AI optimism is growing.1 Yet, despite the surge in spending, executives continue scrambling to demonstrate clear, measurable return on investment. The 2025 Kyndryl Readiness Report found that 61% of business and IT leaders feel increased pressure to prove ROI on AI, and barely half (54%) are actually seeing positive returns.
Why? I think it’s because most organizations look at AI through the wrong lens.
For decades, technology investments have been measured by a simple equation: implement a solution, track the dollars generated and look for opportunities to rinse and repeat. Traditional ROI is backward-looking; it measures what has already been realized. The approach works great for discrete products and services: ‘When my company launched a new product we generated $X in revenue.’ But it falls short when applied to AI, where ROI is more about how fast the organization learns, adapts and scales value through new capabilities. There’s no direct line from “we used AI” to “we made X dollars” — mainly because that’s not what AI does.
As the Chief Technology & Product Officer of a pioneering healthcare company, I spend significant time explaining that AI is not a product. At its core, it’s just another tool in the business toolbox. (However era-defining.) The true value of AI is it gives humans time to do more of what humans do best: connect, decide and create. Those outcomes are not going to show up on the annual 10k report.
I’m not suggesting that AI is not a revenue generator, because it is. But a year from now, the ones to report ROI from their investments will not be the ones drawing a direct line from amounts invested in AI to sales growth, revenue generated or other traditional ROI metrics. They’ll be the ones looking behind the numbers to understand how AI was applied to maximize the human-digital capacity at scale.
Four ways to apply AI
When I talk about gauging the ROI of AI investments, I start with the basics: What business problem are we trying to solve, and what are the associated KPIs?
We can then look at the problem through one or more of the four ways to apply AI:
- Augmenting human capacity: Can we help employees work faster, with more information at their fingertips?
- Enhancing human judgment: Can we use AI to analyze information rapidly, draw insights and help the humans in the loop make better decisions?
- Extending human reach: Can we deploy AI to enable self-service options or automate routine interactions so humans can focus on the higher-value, high-touch work at the top of their license?
The true value of AI is it gives humans time to do more of what humans do best: connect, decide and create. Those outcomes are not going to show up on the annual 10k report.
Examples in healthcare
My work is in value-based healthcare in the U.S., and I have been continuously looking at ways we can use technology to enhance patient outcomes. Viewed in macro, it’s not only a space that cries for a new model, from sick care to healthcare (a different topic for a different article), but healthcare also desperately needs more humans who have more time to create more personalized and more effective patient care experiences.
Clinicians can use AI tools to prepare for patient appointments. AI can help them quickly refresh on patients’ histories and research potential approaches to care recommendations.
Imagine a clinician doing back-to-back 15-minute telehealth visits. AI can automatically generate a concise, structured snapshot of each patient before the visit, query current clinical guidelines, relevant literature, and population outcomes to surface potential next-step options for consideration by the clinician. The use of AI doesn’t replace the expert judgment these medical professionals have refined through years of practice, but it enhances it.
Ambient AI tools augment their capacity by enabling them to spend less time on data gathering and more time interacting with and observing patients, ultimately resulting in better patient experiences and improved health results. Over the course of a 15-month trial, California-based Kaiser Permanente found that such tools not only saved physicians roughly 16,000 hours in documentation time, but also improved patient interactions and job satisfaction.2
AI is also transforming back-office operations by creating a scalable digital human workforce that automates routine tasks and amplifies human impact. It assists with accurate medical coding, claim submissions, denial prediction, automated appeals and error detection — reducing administrative burden and improving financial performance. AI also optimizes scheduling, minimizes no-shows, and balances clinician load, freeing humans to focus on what they do best: connecting with patients to drive better outcomes.
Whether the context is an exam room or the back office, we’re not asking: “Did this AI project generate revenue?”
Instead, it’s, “How did AI help move the KPIs that already drive our business?” Have we met medical adherence targets? Have we increased patient engagement? Ultimately, have we scaled capacity without sacrificing quality?
These are outcomes that emerge when you create more human capacity, enable better judgment, extend reach and inspire more creativity in problem solving. But this is more than a problem of attribution.
Gartner forecasts spending on AI will surpass USD$2 trillion by the end of 2026.3 Boards of directors, investors and other stakeholders will continue to demand a way of accounting for the investments. The pressure to prove ROI is not going away.
So how do we translate this framework for thinking about AI applications to an alternative dashboard for demonstrating ROI?
I think the answer may be in relabeling the I.
Return on Intelligence
We may consider Return on Intelligence to measure how effectively AI enhances an organization’s ability to learn, decide and adapt.
Return on Intelligence captures learning and insight generated by AI systems that make the organization smarter over time.
Back to the example of the clinician whose patient’s condition may not be responding to the prescribed course of care, such measures would provide evidence that AI investments have enabled teams to make better decisions, quicker. Key dimensions may include:
- Decision velocity: how quickly insights turn into action
- Decision quality: the consistency and accuracy of data-informed choices
- Organizational learning: how systems continuously improve through feedback
- Operational agility: the ability to forecast, staff, and scale efficiently
- Engagement intelligence: personalization that drives adherence and retention
- Risk intelligence: the early detection of anomalies or compliance gaps
A provider organization could use AI to power an operations analytics engine that ingests claims, scheduling, and EMR data to identify no-show risk, referral leakage and care gaps. Doing so would build a predictive intelligence that continuously improves operational and clinical decision-making.
Traditional ROI would measure cost savings or revenue from those programs. Return on Intelligence measures how much smarter and more efficient the system becomes because of the data and feedback it collects. Example metrics:
- Percentage reduction in no-shows
- Percentage faster scheduling optimization
- Models retrained weekly using new signals
Together, the Return on Intelligence metrics capture how AI compounds institutional knowledge and human capability.
Return on Interaction
We may consider Return on Interaction, to measure how effectively AI enhances the quality, frequency, and outcome of interactions between patients, clinicians and digital systems.
Return on Interaction measures how each meaningful human-AI touchpoint drives engagement, efficiency and satisfaction.
Such measures would show how teams have been freed up for higher-value interactions with customers. Are employees working at the top of their license? Instead of spending time on the phone helping customers reschedule an appointment or answer a question about a deductible, a patient care coordinator could spend the same time helping patients solve a technical challenge inhibiting their telehealth experience. Can we add valuable minutes into patient visits?
Key indicators include:
- Engagement rate: how often patients or clinicians interact with AI-driven tools
- Interaction quality: measured through satisfaction, comprehension, or adherence
- Conversion to action: how many interactions lead to desired behaviors such as completing exercises or follow-up visits
- Relationship depth: longitudinal engagement and trust
In practice, this might mean higher patient adherence through personalized coaching, more meaningful clinician time because AI handles documentation, or better patient retention through conversational interfaces. Ultimately, Return on Interaction captures how AI transforms every touchpoint into a moment that drives connection and behavior change, which ultimately is what healthcare is about.
A clinician might use an AI documentation assistant that listens to visits, drafts structured notes, and summarizes patient context. This frees cognitive load and time for the clinician that can be redirected to focus on patient interactions while improving accuracy and patient presence.
Traditional ROI would only measure if that engagement led to a billable visit or downstream revenue.
Return on Interaction instead measures the value generated per meaningful patient touchpoint clinically, behaviorally, or operationally.
Example metrics may include:
- Percentage increase in face time with patients
- Percentage decrease in documentation time
- Increase in clinician satisfaction scores
Return on Impact
We may also consider Return on Impact to measure how AI translates intelligence and interaction into tangible outcomes that matter for patients, providers, shareholders and communities. It looks beyond cost savings or efficiency to assess how effectively AI improves health outcomes, equity, and system performance.
Are we scaling existing human capacity? Can we add valuable minutes onto patient visits? Does the timing of upticks in performance against targets for customer satisfaction or retention reflect the introduction of AI into service delivery or other workflows. Key dimensions include:
- Clinical impact: improved recovery rates, reduced readmissions, fewer errors
- Human impact: better clinician experience, reduced burnout, higher satisfaction
- Economic impact: lower cost of care, better risk-based performance
- Societal impact: access, equity, sustainability
If an AI-driven rehab program increases adherence by 30% and reduces re-injury rates it directly raises Return on Impact, proving that the technology not only works, but it also makes a difference.
Return on Impact measures the total value of AI beyond automation, lives saved, errors prevented, time restored to clinicians and trust built with patients. It captures how intelligence amplifies human capability to create measurable, lasting change.
Traditional ROI would measure program cost vs. revenue or cost savings.
Return on Impact measures the magnitude of change in people’s lives and the system — how health, access and experience improved because of the program.
Return on Impact is the ultimate measure of AI maturity — when intelligence and interaction converge to deliver outcomes that are measurable, meaningful and scalable.
These three “returns” together reframe how leaders evaluate AI success:
- Interaction → Are we creating meaningful engagement between humans and technology?
- Intelligence → Are we getting smarter and faster with every data point?
- Impact → Are we improving lives, outcomes, and trust in measurable ways?
Return on Impact measures the total value of AI beyond automation, lives saved, errors prevented, time restored to clinicians and trust built with patients.
Moving the needle with AI
I will close with the same point as I started: AI is not a technology initiative. There’s no more credit for showing proof of use on a shareholder presentation slide. The existence of a tool and the application of the tool are vastly different propositions.
AI is an organizational change initiative. If we create more human capacity. If we enable better human judgement. If we extend human reach and increase creativity. If we reframe the ROI conversation around the sum of these outcomes, then we almost certainly will move the needle.
Have questions or want to learn more?
