There's a tension playing out in boardrooms right now.

CFOs want to see hard numbers proving AI investments are paying off. AI leaders are seeing initial ROI from individual workflows, but they’re also trying to buy some time and belief to prove out the longer-term transformative potential of AI.

This is what Michael Domanic, Head of AI at UserTesting, calls “squishy ROI vs. hard ROI.” Hard ROI is the ROI you can measure in time, dollars, and headcount. Squishy ROI is the organizational value that’s very real, but hard to quantify in dollars on a quarterly basis.

He says you need to do both. Report on hard ROI - so that when someone says, “Is this worth it?” you can prove that it is. But also build the organizational belief in squishy ROI - because without it, you’ll be nickel-and-dimed over every initiative.

Here’s how to do it.

How to measure the “hard” ROI of AI transformation

Measuring hard ROI is a relatively simple process - but it takes discipline.

1. Pick a specific deployment. Not “AI” broadly, but one workflow that your team is planning to augment with AI. Michael gave the example of custom GPTs, which UserTesting built to help the BDR team identify more relevant prospects for discovery calls.

“In this case, you’re not trying to calculate holistic ROI across the company - you’re trying to focus on individual deployments,” Michael says.

2. Identify the business metric that matters. In this case, it’s the conversion rate from  outreach to discovery call - because that first call drives opportunity rate and ultimately revenue. You need to be impacting a solid, revenue- or margins-driving business metric in order to see ROI.

3. Measure the before state. Before building, the UserTesting looked at the typical conversion rate on BDR outreach to set the baseline.

4. Measure the after state. UserTesting then looked at the outreach CVR 3 months after deployment. “In those first 3 months, the GPT grew conversion to discovery call by 60%, and then over the subsequent months, it continued to double,” he says.

5. Connect the improvement to dollars. “The discovery call itself won’t tell you anything about the overall outcome – you need to start putting a value to that,” Michael says. “If we know the average value of a deal, and we know the average conversion rate from discovery call to closed-won, we can project the value that booking that initial call has.”

Michael recommends partnering with business unit leaders to report on the before and after of AI ROI - much more efficient than trying to do it yourself as a head of AI.

How to think about the “squishy ROI” of AI transformation

Hard ROI answers the question: “Is this specific deployment working?” Squishy ROI answers the bigger question: “Why are we doing this transformation at all?”

“We're living through one of the most transformational moments in human history,” Michael says. “I used to say this with a little bit of hesitation. I don't have that anymore.”

At the end of the day, your organization has to believe in the “squishy ROI” of AI in order to be an early winner. If you don't believe in it now, you can wait until it's fully ubiquitous and proven - but you won't get the competitive early wins.

Here's what belief looks like in practice. When UserTesting gave employees the ability to build custom GPTs, they ended up with 800+ of them. Most don't drive measurable business value. Many were just experiments. Some were built for individual use only. But Michael says that’s the point.

“Even though many of those GPTs aren't driving business value, the value they drove is that people experimented, and they had a lot of ‘aha moments’ about how to bring AI into their work," he says. “They became more engaged and more thoughtful about how AI is going to transform their work.”

Squishy ROI is about driving hard toward long-term transformation - sometimes with blinders on. If you get distracted by every naysayer saying “there’s no ROI,” you won’t get there before everyone else.

“Employers who consider themselves a good employer bear a responsibility to lead through this transformation," Michael says. "This is going to impact all of us. I think most people want to work for a company that is providing this transformation, providing the enablement.”

How to use both together

The companies getting this right follow a pattern:

Give widespread access (and be okay with squishy ROI from it). When you're rolling out AI access broadly, you're not making a hard ROI case for giving 5,000 employees ChatGPT licenses. You're making the case that this is infrastructure, that organizational learning requires experimentation, and that early movers will capture disproportionate value.

Measure hard ROI for your top deployments. Then you get ruthlessly focused on measuring your 3-10 highest-value deployments. These become your proof that the transformation is working.

Use hard to reinforce squishy. When leadership asks if all those licenses were worth it, point to your BDR GPT that increased calls by 60% or the marketer who eliminated agency costs. Then explain: “These wins only happened because we gave everyone access to experiment and learn.”