Everywhere you look, you’ll see claims of AI’s power to allow companies to do more, go faster, and stay competitive. There’s just one snag: Proving those claims isn’t always easy or intuitive.
An August 2025 piece in The New York Times concluded that AI might help some companies improve efficiency and effectiveness, but generally only in very specific use cases.
Does this mean it’s time to rethink how much budget goes toward AI systems or tools? I would argue no. From my experience, the problem isn’t that AI isn’t working. On the contrary, I think AI is helping companies in significant ways. However, many companies simply haven’t figured out how to measure AI investment’s ROI.
AI measurement is tricky
AI is still new to many companies. As a result, they haven’t determined which key performance indicators (KPIs) will pinpoint whether an AI product is delivering on the promise of a sizable ROI.
I saw this firsthand when working with a team that began using agentic AI in an assistive capacity within its customer support department. Traditional metrics like speed to resolution and net promoter scores rose after the team put the agentic AI software into action.
However, other levers also contributed to the improved KPIs, such as more comprehensive team member onboarding, greater empowerment of human agents to solve problems quickly, and better talent sourcing and hiring. As a result, the team struggled to determine exactly how much AI had improved performance.
This doesn’t mean that it’s impossible to put a number on the value of an AI solution. Many AI tools offer tremendous time and cost savings, operational scalability, and overall productivity. But determining an AI tool’s true ROI requires thinking in new ways.
1. Look beyond standard metrics
Conventional KPIs don’t always measure the value of an AI product. That said, you can develop other ways to prove on paper that an AI tool helps your organization. Instead of looking at the output of individual tasks, consider how AI improves workflows. According to AI market adaptation and translation company Smartcat, organizations seeing stronger AI returns use it across connected content operations rather than limiting it to isolated tasks. By accelerating these processes, organizations can respond faster to changing market signals and free employees to focus on higher-value work.
One tangible way to capture this is to measure time saved from draft through deployment. Start by benchmarking a single, recurring pipeline such as regional campaign rollouts or product documentation updates. Log how long it takes to move from initial content drafting through review, localization and deployment. If AI cuts a two-week turnaround to three days, you have a practical measure of its ROI.
2. Measure how long you can go without new headcount
Adding talent to your workforce is expensive. With AI, you may be able to keep a relatively lean headcount while still producing more. One way to measure ROI after implementing AI tools is to track how long it takes you to need more people as you grow.
I’ve seen this play out directly across scaling operations. As TropicApp’s interview with Zapier executive noted: a stable headcount coupled with increased output is one indicator of an AI solution’s ROI.
Just be ready to look beyond how much more time each employee has to get tasks done. Sometimes, AI’s real impact is freeing up workers so they can maximize output (even if they spend less time on the clock).
3. Adopt qualitative metrics
Quantitative metrics make impressive KPIs. However, qualitative ones like employee satisfaction and customer experience can indicate the ROI of an AI investment. For instance, higher employee satisfaction often means lower turnover costs and less burnout, while better customer interactions tend to boost retention and lifetime value.
One caveat: If you go this route, you’ll need a before-AI baseline. To establish one, audit historical performance logs or run a quick benchmark pulse survey before rollout. A starting point illustrates whether incorporating AI into your workflows made your qualitative measures trend up or down. Without a baseline, you may have difficulty explaining the meaning behind your after-AI data.
For teams that already skipped this step, the workaround is simple. Look back at pre-launch performance logs or compare current results against teams that aren’t using the tool yet.
The next time you hear doubts about whether AI is truly paying off, think twice before you believe them. Not every AI tool works wonders, but many give companies a serious competitive edge. The key is figuring out which ROI measurements will help you determine if your favorite AI tool is an asset or a liability.
Chalmers Brown is the cofounder and chief technology officer at Due.