Adoption of artificial intelligence seems to be happening at, well, the speed of AI. Instant everything doesn’t always translate to instant return on investment. Jumping on the AI bandwagon without knowing what outcomes you want and how you plan to measure them makes for much noise without any real traction.
So, how can you tell if all the new AI tools you’ve added or want to amount to sound investment? We asked members of the Fast Company Impact Council how they measure ROI on AI initiatives. Here, 20 of them share what matters when deciding whether AI is worth the buzz or has fallen flat.
1. OUTCOMES FOR SPECIFIC PROBLEMS AT A SUSTAINABLE COST
Bigger frontier models don’t automatically create bigger business value. The real benchmark for AI is whether it delivers measurable outcomes for specific problems at a sustainable cost. One area where we really see this playing out for our customers is in internal applications and dashboards. Many Software-as-a-Service solutions are being replaced by custom AI-built tooling. It’s cheaper, faster, and adaptable. — Eric Simons, Bolt.new
2. PROGRESSION OF WHERE YOU NEED TO GO
I break our initiatives into advanced, maturing, emerging, and nascent. For advanced programs, they have clear outcomes like higher throughput and workflow velocity and are scaling. Mature efforts show clear business outcomes but are not yet reaching full potential even at their current scale. Emerging initiatives have a clear outcome in mind, but no measurable impact yet. Nascent effort is largely individual experiments. AI usage and token consumption are not proof of ROI and often confuse what you want to impact. I recognize some efforts are in the early stage, but you must show the progression of where we need to go. — Thomas Scott, Wrike
3. STANDARD BUSINESS OUTCOMES
Our approach to AI has always been about creating concrete value. It’s pretty simple: We measure our standard business outcomes, not AI usage. The idea is to focus on high impact use cases and evaluate whether AI is augmenting teams, speeding up innovation, and tightening security. — Khozema Shipchandler, Twilio
4. QUALITY AND WHAT THE NEXT TEAM CAN REUSE
Usage volume is the easiest AI metric to pull, but it tells you almost nothing. Same with speed. Anyone can make something faster now. What I look at is two things: quality and what the next team can reuse. Does the work still meet the quality bar we’re known for? Can the next team use what we learned to reach a better result faster? We recently designed, built, and launched a mobile app in 72 days, compared with six months or more using the old approach. That speed matters, but the ROI is delivering quality work through a process our team continues to evolve every day. — Peter Smart, Fantasy
5. AGAINST SAFETY AND PRIVACY
The easy part is measuring AI against our biggest objectives and its impact on children. We have years of benchmarks without AI. The harder part is valuing what AI makes possible that wasn’t before, while anticipating risks. For us, that means weighing every use of AI against children’s safety and privacy, not just our reach. As the technology evolves, so must our approach to measuring its value. — Michele Walsh, UNICEF USA
6. ENGAGEMENT AND TIME SAVED IN REAL TIME
For AI to help you improve business results, you need the discipline to match the right tool to the right problem, and to scale what works. Start by measuring in real time and focus on outcomes that matter most for your business, like engagement and time saved. The real value, though, comes from what your people do with that time. If every efficiency gain becomes productivity pressure, burnout increases. That’s why you should redirect capacity intentionally. Shifting people from routine work into strategy, coaching, and innovation can become your competitive advantage. — Jacqui Canney, ServiceNow
7. RELATIONSHIP BETWEEN REVENUE AND OPERATING COSTS
I care less about one ROI number and more about the relationship between revenue and operating costs. If revenue is growing faster than operating expenses that tells me AI is working. The best results I’ve seen come from AI handling the repetitive parts of a job so people can spend their time on the work that requires judgment. I worry about boards pushing for quick headcount cuts and calling that an AI strategy. That’s not transformation. That’s just cost-cutting with better public relations. — Michael Weening, Calix
8. NOT ON TOKENS
Focus on the productivity gained and how much more “good” work can be completed in the same time frame. Never rely on AI alone—but leveraging AI tools to get to 60% to 80% productivity increase can be an outstanding change in output. — Alexis Crowell, Axelera AI
9. ON THE BUSINESS OUTCOMES AGREED ON IMPROVING
We measure ROI on AI by agreeing on the business outcome we’re trying to improve and the metrics we’ll use to evaluate success. We then track those results over time. It sounds simple, but it keeps the focus where it belongs and gives us a clear view of what’s working and where we need to adjust. We’ve found that the biggest gains often come from small, continuous improvements that help people make better decisions and create a better experience for our customers. — Dennis Anderson, ArcBest
10. RELATIVE IMPROVEMENT OF OUTCOMES
I think we need to think about AI ROI a little differently. It’s a mistake to look for ROI on a project-by-project basis because the impact of AI is often indirect, compounded, and sometimes unexpected. Instead, measure the improvement in the outcomes that matter to your business, whether that’s revenue, brand lift, sales pipeline, or productivity, against the cost of your team and the AI tools they use. The relative improvement matters more than the absolute dollar value and that’s where you start to see whether AI is creating value for the business. — Pierre-Loic Assayag, Traackr
11. HOW CONSUMERS INTERACT
We help brands use AI to create experiences that personalize shopping journeys and engage their customers. To measure ROI, we look at how consumers interact with those experiences and how that interaction influences purchasing decisions. We look for boosts in customer engagement, web or app traffic, conversion, basket size, and overall sales, as well as other signals of business and revenue impact. — Alice Chang, Perfect Corp.
12. WHETHER IT HELPS PEOPLE CREATE MORE VALUE
AI ROI isn’t measured by how much work AI does. It’s measured by whether it helps people create more value. We absolutely track speed, productivity, and cost savings, but those are the starting point, not the finish line. As AI makes execution cheaper, judgment becomes more valuable. AI gives our teams more time to think, create, and tell stories that feel more human and authentic. The real ROI comes from combining the math of AI, data, and speed with the magic of creativity, storytelling, and human connection. The goal isn’t to produce more marketing. It’s to create better marketing that drives stronger business outcomes. — Vineet Mehra, Chime
13. AGAINST THE PROCESS IT’S MEANT TO IMPROVE
I measure the ROI of AI against the process it’s meant to improve and not the novelty of the model. I establish a baseline for time, accuracy, cost, and user outcomes, then determine whether AI reduces administrative work, improves service delivery, and helps people reach better decisions faster. For workforce and benefits systems, outcomes must be balanced against security, fairness, and human oversight. — Paul Toomey, Geographic Solutions
14. WHETHER IT HELPS PEOPLE DO WORK PREVIOUSLY OUT OF REACH
Efficiency and hours saved are the easiest things to measure and the least interesting. The better questions are whether AI improved the quality and speed of decisions, shortened the learning curve on something difficult, or helped people do work that was previously out of reach. That last one matters most, because it changes what an organization is capable of, not just how efficiently it operates. Define the intended outcome before you deploy and measure against that. As the tools become ubiquitous, differentiation will come from what people are willing to attempt with them. — Patrick Frend, Delve
15. BY OUTCOMES, NOT ACTIVITY
I measure ROI on AI initiatives the same way I measure any business investment: by outcomes, not activity. The key is tying AI efforts to the business metrics that matter most. To measure impact accurately, you need a trusted system of record and a shared source of truth across the organization. Without that foundation, it’s difficult to separate meaningful results from noise. With it, you can clearly evaluate whether AI is improving customer experiences, increasing operational efficiency, and driving measurable business value. — Melissa Puls, Ivanti
16. BY EFFICIENCY AND EFFECTIVENESS
We encourage clients to measure AI ROI two ways: Efficiency—the labor hours it saves, and effectiveness—the sales it recovers. Prepping for a weekly business review used to take four to six hours, now it takes under five minutes. But the bigger opportunity is structural, like closing a shelf-space gap that’s preventing your best-selling items from meeting demand. That’s where the real dollars show up, not in a shorter to-do list, but in sales you’d have otherwise left on the table. — Are Traasdahl, Crisp
17. WHETHER AI TOOLS EXPAND WHAT THE BUSINESS IS CAPABLE OF DOING
AI ROI should be measured by outcomes, not adoption. I look at whether it increases revenue, lowers costs, improves decision making, or allows the team to accomplish more without adding headcount. If an AI initiative cannot be tied to measurable business improvement, it may be interesting technology, but it is not yet a successful investment. The best AI tools do not simply make existing processes faster. They expand what the business is capable of doing. — George Kailas, Prospero.ai
18. ON WORKFLOW OUTCOMES
ROI on AI initiatives is as simple as measuring spending declines due to productivity gains and/or increases in revenue. But it takes time for those needles to move. Near term impact is more easily seen by focusing on specific key performance indicators and outcomes at the workflow level like reduced time to answer an employee question or resolve a customer issue. So, measure and optimize workflow outcomes and overall ROI will be seen in time. — Randi Lee, Lucas Advisory
19. THROUGH QUANTITATIVE AND QUALITATIVE SIGNALS
We measure AI ROI through quantitative and qualitative signals, tied to outcomes investors care about: revenue growth, operating leverage, and competitive advantage. Quantitatively, we track adoption, efficiency gains, and development speed, then check whether those gains translate into faster releases, lower costs, stronger margins, or scaling without adding head count. Qualitatively, teams demonstrate new AI capabilities at company-wide meetings, showing better decisions and stronger customer experiences. AI delivers ROI when it strengthens the economic engine of the business, not when more people use it. — Cody Barbo, Trust & Will
20. IN DATA INTEGRITY AND DEFENSIBILITY
Companies often mistakenly prioritize cost savings and speed when measuring AI ROI, ignoring the foundational importance of data quality. If AI is trained on narrow or questionable datasets, efficiency gains cannot compensate for biased or legally indefensible outputs. True ROI must prioritize data integrity and defensibility. A fast answer that lacks ethical or legal standing is not a real return. — Denas Grybauskas, Oxylabs
