Which metrics do you rely on to measure the business impact of an AI initiative, and why?
Most answers focus on user engagement, ROI, and efficiency metrics, though some emphasize workflow improvements, cost savings, or direct business outcomes like revenue per employee.
11 answers
Answers

Marissa Mayer
CEO & Founder · Dazzle Ai
Retention
Marissa Mayer says retention over time is probably the most important metric, and that the ratio of daily to longer-term active users matters because useful technologies get used nearly every day.
Semafor · Watch on YouTube

Patrick Moorhead
Founder, CEO and Chief Analyst · Moor Insights & Strategy
ROI
Patrick Moorhead says he looks for proof of life in end users, meaning enterprises seeing big ROI swings, because the only way to do this is to reduce expenses or increase top-line revenue.
The answer starts at 11:51.
The AI Forecast: Data and AI in the Cloud Era · Listen on Apple Podcasts

Sagnik Nandy
Chief Technology Officer · Docusign
Cost and Token Efficiency
Sagnik Nandy says his team measures cost per token and tokens per task, saying both are very important because they focus on ROI and sustainability, and reducing tokens per task and cost per token buys more headroom to try things.
The answer starts at 36:23.
The Engineering Leadership Podcast · Listen on Apple Podcasts

Brian McMullin
Senior Vice President and Head of Product · Network Solutions
Retention
Brian McMullin says he relies on retention, though he calls it a lagging indicator and says frequency and breadth of product usage are predictive of customers sticking around.
The answer starts at 26:56.
LaunchPod · Listen on Apple Podcasts
Jonathan Evens
Product Lead · Google Deepmind
Side-by-Side Win Rates
Jonathan Evens says he looks for side-by-side win rates, to continuously evaluate model outputs and see whether they are better than a previous version, a competing product, or a baseline.
Mind the Product · Watch on YouTube

Supratik Shankar, CFA
Building Dview · Dview
Time Saved
Supratik Shankar says AI automation will free up bandwidth and save time for existing teams to focus on other tasks, which he says creates meaningful impact for the organization.
Analytics Insight · Analytics Insight · Listen at 22:53

Jeff Westcott
Director, Practice Technology, Innovation & AI · Akin Gump Strauss Hauer & Feld LLP
Adoption and Usage
Jeff Westcott says he relies on adoption and usage metrics, such as whether people come back and use a tool daily or weekly and how much time they spend, because adoption is a good benchmark.
ILTA Voices · ILTA Voices · Listen at 31:33

mike yu
cofounder, ceo · vesta
Human Cost to Originate
Mike Yu says the numbers they look at are the fully loaded human cost to originate, since AI agents are meant to directly replace operational labor; he adds head count, loans per underwriter monthly, and cycle time.
Chrisman Commentary · Watch on YouTube

Pete Johnson
Field CTO, Artificial Intelligence · MongoDB
Workflow Performance Metrics
Pete Johnson says he relies on existing workflow metrics—like those used to bonus and evaluate employees—because when AI is added to a workflow and those metrics jump, he can attribute the jump to the agent and compute ROI.
Super Data Science: ML & AI Podcast with Jon Krohn · Watch on YouTube

Srinivasan Raghavan
Chief Product Officer · Freshworks
Usage and Satisfaction
Srinivasan Raghavan says he relies on a combination of usage metrics and satisfaction metrics, measured partly through in-product surveys of 5–10% of users monthly, because these signals make the functionality more effective.
DataCamp · Watch on YouTube

Neil Patel
Co-Founder · Np Digital
Revenue per Employee
Neil Patel says he finds measuring revenue per employee a healthy way to look at it as of today, citing someone whose revenue per employee rose from $600,000 to $1.2 million as showing efficiencies.
Marketing School - Daily Marketing Tips · Watch on YouTube