AI ROI Is Bigger Than Efficiency

AI has entered a more demanding phase in the enterprise. Enthusiasm and experimentation alone isn’t enough anymore; leaders must show measurable returns.

AlphaSense data reflects that shift, with mentions of AI ROI in earnings transcripts up 116% year over year and references to successful AI deployments rising 83%. Even major players are making hard choices about which capabilities are worth sustaining. OpenAI’s recent decision to shut down its video generator Sora is one example. Sora demonstrated what the technology could do and delivered experimental value, but also highlighted the gap between technical progress and business practicality.

Yet as ROI expectations rise, many companies are still using a narrow definition of value centered on time savings, workflow speed, and task automation. That lens misses where AI is creating the most value. In research, strategy, and other forms of high-value knowledge work, AI often improves outcomes by helping people think more clearly, spot gaps, and make better decisions, all of which can’t necessarily be demonstrated by speed.

To understand AI’s true enterprise impact, companies need to move beyond efficiency alone and measure how AI improves judgment and decision quality.

Traditional ROI metrics are too narrow for the current AI era

To date, enterprise technology ROI has been measured through metrics that are easy to quantify, such as hours saved, costs reduced, and processes accelerated. Those metrics still matter, especially in automation-heavy use cases where AI is replacing repetitive or stable tasks.

But generative AI creates value differently. In many cases, AI contributes not by reducing the amount of work, or the time it takes to get done, but by improving the quality of the outcome. AI can challenge assumptions, reveal blind spots, or surface a risk that materially improves a decision. It may do all of that without cutting task time.

Though these outcomes are harder to quantify than time saved, they are more closely tied to real business value. If companies define AI ROI too narrowly, they risk undervaluing the use cases where AI has the strongest strategic upside.

Enterprises need more sophisticated ways to measure impact

AI is not one-size-fits-all support. Rather, it is increasingly personalized to individual goals, interests, workflows, and behavior. Two people can use the same AI system, receive different outputs, and both still have valuable experiences.

One example of this shift is the rise of customizable AI agents. At AlphaSense, these systems allow users to build tailored agents that monitor specific industries, topics, and themes and receive information in formats aligned with how each person works. Around 40% of AlphaSense users who created custom agents engaged with the scheduled alert, suggesting meaningful adoption and recurring value. That usage also demonstrates that people are willing to invest time upfront in configuring AI around their needs when the value is clear.

As AI becomes more proactive, companies need to ask more substantive questions: Did the system surface something the user was not already considering? Did it improve the direction or quality of analysis? Did it increase confidence in a decision? Did it influence a course correction before a mistake was made?

In structured environments, those signals may be easier to track. In areas like research, strategy, and market intelligence, value may emerge more gradually and less visibly. That makes AI ROI harder to measure, but also critically important to define correctly.

As personalized and proactive AI becomes more central to daily work, enterprises need ROI frameworks that capture ongoing impact on decision-making in tandem with discrete efficiency gains.

Also Read: AiThority Interview with Glenn Jocher, Founder & CEO, Ultralytics

The best AI systems expand perspective

At the same time, as systems become more tailored, they can also become more narrow.

AI that optimizes only for relevance based on a user’s past behavior can reinforce existing views while filtering out unfamiliar but important information. The system may feel effective, but it’s actually reinforcing blind spots and filtering out unfamiliar, important information.

That is especially risky in functions where strong decisions depend on broad perspective, signal detection, and exposure to dissenting views. The best advisors help people find what they’re looking for and, in turn, help them question if they are looking for the right thing. This level of thinking requires AI measurement to go beyond relevance to include breadth.

Companies should start considering metrics such as source diversity, perspective expansion, and influence on decision-making. Are users seeing a healthy range of sources and viewpoints over time? Is AI surfacing overlooked risks or counterpoints? Did the output lead to a new watchlist, a new line of research, or a new strategic question? Did it materially sharpen the user’s thinking? Are teams adjusting strategy based on AI input?

This is a critical shift for companies to unlock AI’s full business value. The most effective  AI systems are engineered to make decision-making more informed, more rigorous, and less vulnerable to blind spots.

Measuring What Matters Most

Enterprise AI has reached an inflection point where value must be proven more clearly. That requires companies to think beyond ROI frameworks centered around time savings.

As AI continues to become more personalized and proactive, leaders will need to measure speed alongside improved judgment and decision-making. This evolution will sharpen thinking, broaden perspective, and drive better outcomes.

Also Read: ​​The Infrastructure War Behind the AI Boom