On Thursday, I sat down in front of 20,000 people and talked for 7 hours about the question on every single leader’s mind: Where the hell is the AI ROI I’ve been promised?
Here’s the lesson I took away from it: There is plenty of AI ROI. The successful Heads of AI are finding hundreds of use cases that save time, free up capacity, or drive growth.
But it takes a lot of work, and the work is not glamorous. It’s a lot easier to NOT do the work, then say, “The ROI doesn’t exist.”
During the session, one attendee (I won’t name names) kept saying, “When are we going to start talking about ROI?”
Dude – we are. We have been. Building GPTs, running lunch and learns, sharing use cases, winning hearts and minds, mandating proficiency – THIS is the hard work of driving ROI. But if you want a magic wand, keep looking. You won’t find one here.
Here are my favorite lessons about the work it takes to drive AI ROI – and what you can expect if you do. Get the full playbook of yesterday’s insights here.
1. You’re not ready for sophisticated workflows if you don’t have basic proficiency.
We often talk to clients who want to jump right to sophisticated workflow redesign – and meanwhile they haven’t rolled out ChatGPT Enterprise and no one at the company has even one AI use case.
Leo Casado from Autodesk said there are three levels of AI integration and workflow sophistication:
Level 1: Individual workflows built on top of your existing LLM
Level 2: Team workflows involving multiple steps and data manipulation
Level 3: Enterprise orchestration using automated systems with engineering integration
He said most companies try to jump straight to Level 3 (enterprise automation) without building proficiency at Levels 1 and 2 first. But there’s plenty of ROI in the “non-glamorous” workflows built with custom GPTs.
He gave the example of a QBR (quarterly business review) standardization GPT, which automates the first draft of a QBR by pulling and formatting performance data and expressing it via slides. It still needs human review, but it saves the team 10+ hours a month in creating the content. Not glamorous – but meaningful.
2. You must insist on knowing, “Did this AI project save time or money?”
Most people concentrate on getting buy-in for AI investments – but don’t follow up on the ROI on the back end. Eric Porres from Logitech mandates “outtake forms” for every AI pilot, which detail the project strategy and the time or dollars saved.
He says in one case, an SQL validation automation cut manual review time by 90%, saving the team nine hours per week. Another ops assistant saved 250 hours a week across teams. He says when you start tracking this way, you’ll find pockets of AI ROI everywhere in your organization (and you’ll know what didn’t work so you can learn from it).
3. Know the difference between “squishy ROI” and “hard ROI.”
I loved this distinction from UserTesting’s Michael Domanic. Hard ROI is your time and cost savings, expressed in numbers (aka – what the CFO is looking for). Squishy ROI is the potential for your organization, which you just have to believe in.
For hard ROI, lean on your business units to report their progress. These are the people in your org who really know what’s going on on the ground floor, so get them to participate in measurement. Michael’s marketing team used to have to build dozens of microsites by hand, until a marketer vibe-coded inside GPT to automate microsite creation.
For squishy ROI, it’s more about belief. You need to REALLY believe two things: One, that AI will transform your company, and two, that you bear a responsibility to your employees to lead them through it.
4. You can’t skip change management (even though it takes so much time).
Dr. Alicia Abella from Novo Nordisk is a great example of how hard and time-intensive change management is. When she joined the company, there was zero AI infrastructure. In the last 10 months, she’s developed intake processes, approved workflows, built and trained 3+ dozen AI champions, run lunch and learns, booked external speakers, and created detailed adoption plans for each AI product.
She described driving AI adoption as “like marketing the new iPhone.” One company All Hands or memo is not enough – you have to talk about AI until people are sick of it (believe me, I’ve done this at Section).
Reality check
The reality is that AI ROI isn't a destination – it's a discipline. The companies seeing real returns aren't the ones waiting for AI to magically transform their business overnight. They’re the ones who have appointed someone to lead the effort, invested in good tools, painstakingly tracked and measured value, and run a million office hours to help people use AI.
If you're frustrated that your AI investments aren't paying off yet, ask yourself: Are you doing the work? Because the ROI is there – but only if you're willing to roll up your sleeves and build it.
