• Everyday AI
  • Posts
  • Ep 725: Measuring AI ROI: Why you’re doing it wrong and the 7 Steps to fix it (Start Here Series Vol 11)

Ep 725: Measuring AI ROI: Why you’re doing it wrong and the 7 Steps to fix it (Start Here Series Vol 11)

Google Unveils Gemini 3.1 Flash-Lite, the Supreme Court Sides With Humans on AI Copyright, OpenAI Scrambles to Amend Its Pentagon Deal, and more

Author

Everyday AI
March 03, 2026

 

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Think AI is delivering ROI at your company? In Episode 11 of our Start Here series, we explain why you’re probably measuring it wrong — and how to fix it. Give today’s show a watch/read/listen to find out.

🕵️‍♂️ Fresh Finds: Claude Rolls Out Memory for Free Users, Qualcomm Unveils a New AI Chip for Wearables, Google Tests New Gemini Enterprise Projects Feature, and more Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Google Unveils Gemini 3.1 Flash-Lite, OpenAI Scrambles to Amend Its Pentagon Deal, Apple’s new AI-powered Macbook and more. Read on for Byte Sized News.

💪 Leverage AI: AI ROI isn’t the problem. Measurement is. Here’s how to finally get receipts. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Claude goes down, NVIDIA makes multiple billion dollar AI bets, Anthropic to sue gov and Supreme Court shoots down AI copyright. Check it here!

Ep 725: Measuring AI ROI: Why you’re doing it wrong and the 7 Steps to fix it (Start Here Series Vol 11)


Why can't most companies show their ROI on GenAI?

Because their implementation is backwards.

If you're using the same digital transformation playbook that you used for the social media and cloud eras, you're in trouble.

On this 'Start Here Series' episode, we break down what your company is doing wrong and the 7 Step process to properly calculate ROI on your AI efforts.

Also on the pod today:

• The seven-step AI ROI fix 🛠️
• MIT “study” debunked: vibes only 📉 
• Employees pocketing AI productivity 🏌️


It’ll be worth your 43 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Krisp Converts others’ accents in meetings for you, Sequirly is the safety net for your AI workflow, Shavely is a Real-time multilingual chat

Claude Memory — Claude just rolled out Memory for free users, letting you easily save, import, and export your AI memories.

HONOR Robot Phone — HONOR’s new Robot Phone actually moves and reacts, plus the Magic V6 foldable packs a huge battery in a super thin frame. Click To Learn More

OpenClaw Upgrades — OpenClaw’s new update brings Telegram live streaming and a native PDF tool. But sleep mode? Still missing.

Qualcomm New AI Chip — Qualcomm just dropped a new chip for AI wearables, promising smarter gadgets with longer battery.

Pocket Sized AI Brain — Researchers made an AI vision model thousands of times smaller with monkey brain tricks. Curious how?

Amazon $40B Investment — Amazon is pouring $40 billion into AI and data centers in Spain. See how this could reshape Europe’s tech game.

Google Testing Projects — Google is quietly testing a projects feature in Gemini Enterprise, hinting at a bigger upgrade soon.

New iPad Air — The new iPad Air packs the M4 chip and powerful AI features for next-level speed and smarts. Preorders start soon—curious what it can do?

Meta Shopping — Meta’s chatbot now suggests products with pics and prices. Is this the future of shopping?

1. Google Unveils Gemini 3.1 Flash-Lite: Speed Meets Savings 🧠

Google has just previewed Gemini 3.1 Flash-Lite, its newest AI model built for lightning-fast responses and wallet-friendly pricing, available now in Google AI Studio and Vertex AI.

Claiming a 45% boost in speed over previous versions, Gemini 3.1 Flash-Lite is designed to handle complex tasks with smarter, dynamic processing while slashing costs for developers. With impressive leaderboard scores and high marks in scientific and multimodal reasoning, it’s positioned as the most efficient Gemini yet.

2. Apple Launches MacBook Air M5 With AI Power Boost 💻

Apple has revealed its new MacBook Air M5, rolling out March 4 with a stronger focus on AI performance, improved wireless connections, and double the base storage, but at a higher $1,099 starting price.

The M5 chip promises up to four times faster handling of AI tasks compared to the previous model, making it a timely upgrade for anyone relying on AI tools. The refreshed laptop keeps its popular design and battery life, while adding faster SSD speeds and support for two external displays.

3. OpenAI Tweaks Pentagon Deal Amid Backlash 🕵️

OpenAI is scrambling to amend its AI supply contract with the US Department of War after CEO Sam Altman admitted the rushed agreement "looked opportunistic and sloppy."

The deal, which replaced Anthropic as the Pentagon's preferred AI provider, quickly sparked fears about domestic surveillance and AI-powered warfare, fueling a "delete ChatGPT" campaign and an employee revolt. OpenAI now says its technology will be explicitly barred from surveillance and autonomous weapons use, but skeptics are questioning whether these promises hold up under government pressure.

4. Alibaba Unveils Qwen 3.5 Small Model Series 🖥️

Alibaba just dropped its new Qwen 3.5 Small Model lineup, promising "more intelligence, less compute" for AI fans and developers.

The models range from ultra-lightweight versions for edge devices to a surprisingly capable 9B model that punches above its size. This release marks a shift toward making high-performance AI tools more accessible and efficient for real-world use, research, and experimentation.

5. ChatGPT Faces Backlash After Pentagon Deal 🤝

ChatGPT saw a dramatic spike in US app uninstalls and negative reviews this weekend after news broke of its partnership with the Pentagon to deploy advanced AI in classified settings.

The fallout was swift: downloads tumbled, 1-star ratings soared, and competitor Anthropic’s Claude app surged to the top of the charts by refusing military use, citing safety concerns. Consumer sentiment swung sharply, favoring Claude’s stance and pushing it ahead in rankings across multiple countries.

1. GDPVal Just Closed The Case 🔥

Here's that GDPVal breakdown we promised.

OpenAI's GDPVal benchmark tests today's top AI models against seasoned professionals across 44 occupations in nine GDP sectors. Expert judges with an average of 14 years of experience grade those outputs completely blind.

AI wins those comparisons 70% of the time. At 100 times the speed.

So when a viral MIT study dropped in August 2025 claiming 95% of enterprise AI pilots deliver zero ROI, markets panicked hard.

NVIDIA shed three and a half points. Palantir dropped nearly nine percent.

That 95%? Came from 52 qualitative interviews with zero quantitative data. (It’s marketing masquerading as a vibe study.) 

The researchers called their own findings directional. MIT's NANDA lab, by the way, is out there actively building and selling its own agentic AI product.

Sus.

Real quantitative data from thousands of actual business leaders tells a wildly different story. IDC found a $3.70 return for every dollar invested. Wharton, Google Cloud, and Deloitte all found between 74% and 84% of enterprises reporting real ROI gains.

The debate is cooked, fam.

Try This

Next time someone cites the "95% failure rate" in your boardroom, ask one question: how many companies did that study survey with quantifiable data for that claim? (The answer is zero. lolz) 

Pull IDC, Wharton, Google Cloud, and Deloitte into your next budget conversation. Thousands of respondents, real receipts, wildly different picture.

Your AI investment decisions are too important to run on vibes.

2. Your ROI Is On Someone's Golf Course ⚡

Here's the dirty little secret nobody in leadership wants to say out loud.

Your AI ROI prolly exists. It's just sitting in your employees' pockets, especially those who are remote or hybrid workers. 

What were they doing with all that reclaimed time? Chilling. Golfing. Taking a walk. 

Can't even be mad. If nobody changed the output expectations, why would anyone volunteer the extra capacity?

Workday reports 89% of organizations haven't updated half of their job roles to reflect the AI age. Workers are getting dramatically more productive inside job structures built a decade ago, still measured against the same pre-AI output bar.

That invisible productivity ain't showing up in your ROI dashboard. It's showing up on someone's Tuesday tee time.

Try This

This week, ask five remote or hybrid employees one honest question: what are you actually doing with the time AI saves you?

Their answers will straight up tell you where your ROI went.

Pull a few job descriptions from your highest AI-usage teams. If expected outputs haven't changed since 2022, the role structure is leaking your returns.

Pick one role and rebuild output expectations for an AI-native world.

3. Seven Steps To Finally Get Receipts 🚀

Here's the real reason most companies can't prove AI ROI. It ain't because AI doesn't work.

It's because they never measured anything before AI showed up.

That's the BASE. Baseline Assessment of Standard Execution.

You gotta run it before AI ever touches a single workflow.

Have multiple employees complete the exact task with zero AI. Record average time, error rate, rework cycles, and cost per completed task.

Because once AI is already woven into steps two and six of your 10-step process, that baseline is gone forever.

The seven-step sprint: define your success rubric, measure the human baseline, build 20 to 40 real messy test cases, configure the exact production workspace, run each case three times with memory off, grade outputs blind, then do the math. Time saved multiplied by fully loaded hourly rates, minus AI subscription costs.

Retest monthly. Under-the-hood model updates can quietly tank your numbers without a single announcement from any AI lab.

Try This

Pick one workflow and start the BASE this week. Two or three employees, zero AI, every step timed and documented.

You cannot collect that baseline retroactively once AI is already in the mix.

Run the AI version three times with memory off and grade blind against the human version. That's how you stop having the ROI conversation and start winning it.

 

 

Reply

or to participate.