Close Menu
The Financial News 247The Financial News 247
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
What's On
China Has Outspent The U.S. On Research For The First Time. 3,375 American Scientists Are Telling Congress To Pay Attention

China Has Outspent The U.S. On Research For The First Time. 3,375 American Scientists Are Telling Congress To Pay Attention

May 20, 2026
15 Skilled Trades With The Highest Pay And Growth Potential

15 Skilled Trades With The Highest Pay And Growth Potential

May 20, 2026
FTC urged to investigate Roblox for allegedly exposing kids to sex predators, misleading public about safety

FTC urged to investigate Roblox for allegedly exposing kids to sex predators, misleading public about safety

May 20, 2026
Your Company Is Measuring AI Adoption Wrong. Track This Instead.

Your Company Is Measuring AI Adoption Wrong. Track This Instead.

May 20, 2026
Trump-Backed Burt Jones And Healthcare Tycoon Rick Jackson Advance To GOP Runoff In Gov. Race

Trump-Backed Burt Jones And Healthcare Tycoon Rick Jackson Advance To GOP Runoff In Gov. Race

May 20, 2026
Facebook X (Twitter) Instagram
The Financial News 247The Financial News 247
Demo
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
The Financial News 247The Financial News 247
Home » Your Company Is Measuring AI Adoption Wrong. Track This Instead.

Your Company Is Measuring AI Adoption Wrong. Track This Instead.

By News RoomMay 20, 2026No Comments5 Mins Read
Facebook Twitter Pinterest LinkedIn WhatsApp Telegram Reddit Email Tumblr
Your Company Is Measuring AI Adoption Wrong. Track This Instead.
Share
Facebook Twitter LinkedIn Pinterest Email

Abhishek Gandotra is VP of Product at American Express. He is the author of The Sovereign Agent: A Manifesto for the Inference Economy.

To maintain your company’s edge, measure human judgment and decision-making regarding AI outputs.

​Every enterprise dashboard I’ve seen in 2026 tracks the same AI metrics: adoption rate, time saved, tickets automated, cost per inference. These numbers are going up and to the right at almost every company that has deployed agentic AI at scale.

They are also the wrong numbers.

I’ve spent 20 years building platforms in financial services—first at Green Dot, where we powered Apple Pay Cash and served millions of accounts, and now at American Express. In both environments, I’ve watched organizations measure the thing that’s easy to count and miss the thing that actually matters. AI adoption is no different. The metrics that feel like progress are masking a convergence problem that will show up in 18 months as a competitiveness crisis—and by then, the damage will be structural.

Here’s what I mean, and what to do about it.

The Convergence Trap

When an entire team adopts the same AI model with the same default settings, their output starts to look the same, converging not toward excellence, but toward the model’s median. Marketing emails start sounding identical. Engineering design documents propose the same architectures. Strategic analyses reach the same conclusions. This is not a failure of the people. It is the predictable result of a system where 12 professionals are all polishing the same model’s default output.

From the outside, productivity is up. From the inside, differentiation is gone.

I’ve seen this pattern play out in content marketing, legal document preparation, financial reporting and software quality assurance—industries where AI-assisted output scaled the fastest. The firms that maintained competitive differentiation had one thing in common: they measured quality of intent, not just output volume.

Three Metrics That Actually Matter

1. Human-Originated Decision Rate

What percentage of your team’s output was directed by specific human intent before the model was engaged and how much was generated by the model and then edited? There is a material difference between “I specified what I needed and the AI drafted it” and “the AI produced something and I cleaned it up.” The first leverages the technology for intentional reasons. The second is dependence on the tech. Most organizations cannot currently distinguish between the two, and that inability is a strategic blind spot.

2. Output Divergence Across The Team

If five people on the same team produce work on the same topic using the same tools, how different are the results? In a healthy AI-augmented workflow, the outputs should reflect the distinct judgment and domain expertise of each individual. If the outputs are converging—if you could swap the bylines and nobody would notice—the team has collapsed into what I call model-median production. The humans are adding formatting, not value.

3. Refusal Rate

How often does someone on the team reject model output and substitute their own judgment? A refusal rate of zero is not a sign of good AI tools. It is a sign that nobody on the team is exercising judgment. The most valuable AI-augmented workers I’ve observed are the ones who use the tools aggressively and override them selectively. They know when the model’s answer is good enough and when “good enough” isn’t good enough for this particular decision.

The Taxonomy Shift Happening Under Your Metrics

The traditional knowledge worker did one thing: produce output. Write the code, draft the memo, build the model, send the email. AI has compressed that production work so dramatically that paying a human for pure production is becoming economically irrational.

What’s emerging is a three-tier structure. At the base, production-level work—the doer role—is being absorbed by AI at an accelerating rate. In the middle, judgment-on-output work—what I call the editor role—is where most knowledge workers have landed, often without realizing it. They review AI drafts, they refine AI code, they approve AI analyses. They are more productive than ever and less differentiated than ever.

At the top, a third role is expanding: the person who decides what work should be done at all. Not the person who writes the design document, but the person who decides whether the design document should exist. Not the person who builds the feature, but the person who kills the feature. This is the architect role, and it is defined not by technical skill but by the willingness to make decisions with real consequences—and to own the outcome when those decisions are wrong.

The companies that will lead in the next two years are the ones whose leadership teams understand this taxonomy and are actively managing the migration from doer to editor to architect across their workforce.

What This Means For Leadership

If you are running a team or a business unit, here is the practical implication: stop celebrating adoption metrics. Your competitors have the same tools, the same models, the same APIs. Adoption is table stakes.

Start measuring judgment. Start tracking who on your team specifies intent before engaging the model versus who reacts to model output. Instead of asking “Did we use AI?” in every project review, ask: “Where did a human override the AI, and why?”

The organizations that will maintain competitive advantage in the inference economy are not the ones that adopted AI fastest. They are the ones where humans retained the capacity—and the organizational permission—to say, “The model is wrong, and here’s what we should do instead.”

That capacity is a muscle. If you don’t measure it, you can’t train it. And if you don’t train it, you’re building your company’s future on rented land.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Abhishek Gandotra
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email

Related News

China Has Outspent The U.S. On Research For The First Time. 3,375 American Scientists Are Telling Congress To Pay Attention

China Has Outspent The U.S. On Research For The First Time. 3,375 American Scientists Are Telling Congress To Pay Attention

May 20, 2026
People Are Really Angry At AI Content Even If It Turns Out That AI Didn’t Produce It And The Content Was Actually Human Made

People Are Really Angry At AI Content Even If It Turns Out That AI Didn’t Produce It And The Content Was Actually Human Made

May 20, 2026
How To Play Before The Release Date

How To Play Before The Release Date

May 20, 2026
Why AI Literacy Has Become A Boardroom And Investor Priority

Why AI Literacy Has Become A Boardroom And Investor Priority

May 20, 2026
AI Data Center Build Out Faces Infrastructure And Political Head Winds

AI Data Center Build Out Faces Infrastructure And Political Head Winds

May 20, 2026
Ronda Rousey’s 17-Second Win Drew Staggering Netflix Viewership

Ronda Rousey’s 17-Second Win Drew Staggering Netflix Viewership

May 20, 2026
Add A Comment
Leave A Reply Cancel Reply

Don't Miss
15 Skilled Trades With The Highest Pay And Growth Potential

15 Skilled Trades With The Highest Pay And Growth Potential

News May 20, 2026

The highest-paid trade skills now offer a compelling alternative for professionals seeking a recession-proof pivot…

FTC urged to investigate Roblox for allegedly exposing kids to sex predators, misleading public about safety

FTC urged to investigate Roblox for allegedly exposing kids to sex predators, misleading public about safety

May 20, 2026
Your Company Is Measuring AI Adoption Wrong. Track This Instead.

Your Company Is Measuring AI Adoption Wrong. Track This Instead.

May 20, 2026
Trump-Backed Burt Jones And Healthcare Tycoon Rick Jackson Advance To GOP Runoff In Gov. Race

Trump-Backed Burt Jones And Healthcare Tycoon Rick Jackson Advance To GOP Runoff In Gov. Race

May 20, 2026
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo
Our Picks
People Are Really Angry At AI Content Even If It Turns Out That AI Didn’t Produce It And The Content Was Actually Human Made

People Are Really Angry At AI Content Even If It Turns Out That AI Didn’t Produce It And The Content Was Actually Human Made

May 20, 2026
Trump Ally Seeks .7 Million Payout From .8 Billion ‘Anti-Weaponization’ Fund

Trump Ally Seeks $2.7 Million Payout From $1.8 Billion ‘Anti-Weaponization’ Fund

May 20, 2026
How To Play Before The Release Date

How To Play Before The Release Date

May 20, 2026
3 Key Takeaways From Trump’s IRS Settlement & Tax Immunity Controversy

3 Key Takeaways From Trump’s IRS Settlement & Tax Immunity Controversy

May 20, 2026
The Financial News 247
Facebook X (Twitter) Instagram Pinterest
  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact us
© 2026 The Financial 247. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.