Why this AI-powered video company measures impact rather than token usage

AI systems are infiltrating workflows all across the world of business, likely compelling companies and their HR teams to drive employee AI adoption and usage thoughtfully, without blowing through entire AI budgets. Measuring what constitutes success is becoming a growing question for business leaders and people pros.

“Everything is being redefined through AI, but in particular, how every company runs in terms of [an] operating model,” Laura Gonzalez Florez, AI-powered video platform Synthesia’s chief of staff and head of people, said. “For me, it’s not a systems-only problem, it’s a systems, it’s a people, and it’s a workflow problem.”

Pricing for the new technology is evolving into a pay-per-use service, with AI tokens serving as units that vendors use to represent the work being processed by large language models (LLMs). In recent months, some companies—and unsanctioned employees—designed leaderboards and trackers that monitor token use, encouraging some to maximize their token use, known as tokenmaxxing.

For Synthesia, tokenmaxxing didn’t stick. Like, at all.

Leaders at the company instead looked to spotlight “some of these learnings and the great things that are changing” as a result of new AI processes during its company-wide all-hands and in communications channels like Slack. “Once we pick those, we then try to deploy them more consistently around the organization,” Gonzalez Florez said.

Tokenmaxxing “became a hot topic around ‘this is what matters,’” she said. “For us, we believe in giving people the tooling—and we’re very lucky to have the ability to fund those things—but actually it was more about that. That’s not what matters in terms of measuring. What matters is impact, so we completely moved away from those vanity metrics of maximizing tokens.”

Tokenmaxxing simply relies on usage metrics: Those with high AI token use can showcase how much or how often they can turn to AI, but not whether or not they can discern how to use AI in order to drive impact.

“One of our principles is effectively high ownership, high impact,” she said. “We don’t care about you telling me what process you’re going to follow, what plan you have, what you’re going to do. No one cares. Show me the results, show me the impact.”

For Gonzalez Florez, it’s more useful to the organization to identify “the people that are finding crazy smart ways to fundamentally change a process” and recognize those efforts and implement those key use cases more widely across the org.

“That is very embedded into how we think about our talent, and so that drives, obviously, recognition in terms of whether it’s celebration, [a message in a] Slack channel; whether it’s a spotlight on an all-hands.”

Gonzalez Florez told HR Brew the company explored measuring AI-adoption 12–18 months ago, and found AI usage was not an issue.

“We don’t have a problem of actually activation and trying to push [adoption],” she said. “We have an advantage, because our employees are biased towards using it. We’re like, ‘okay, we’ll never measure that again, and we’re now looking at how is it used.’”

Gonzalez Florez told HR Brew that tokenmaxxing came up most in recruitment when candidates asked about the company’s approach to tokenmaxxing.

“Some candidates were asking about it, and that’s when we thought we had to have a very clear stance on it,” she said. “For us, it’s a nuanced differentiation between being in a great position to allow experimentation and giving everyone the tools they need to get stuff done and ship things, which really matters to a lot of these incredible individuals that we want to hire.”

It comes back to the AI strategy at Synthesia: identifying the systems, people, and workflows that require some transformation to best deliver impact in an AI-enabled world.

“What matters is the judgment and the quality of decision-making,” Gonzalez Florez said. “The only thing that matters at the end is: are we making the right decisions to move fast and deliver what we want to deliver?”