This agency wouldn't reward AI use until it could prove the payoff was real
At Coalition Technologies, no one gets a bonus for using AI. They get one for proving, with data, that it actually worked.
That kind of proof is possible because the company has 16 years of performance data on nearly every role, collected long before generative AI entered the picture. Jordan Brannon, president and co-founder of Coalition Technologies, a Los Angeles-based e-commerce marketing and web design agency, said that history has shaped how the company approaches AI today.
"We predate AI healthily and had pretty rigorous processes, policies and tools well before AI became relevant to a lot of our workflows," Brannon said. "That's created a lot of opportunity for disruption and change and evolution, as AI has both been hyped and become more capable."
Measuring what actually changes
Coalition's pre-AI performance data now works as a built-in benchmark, position by position, for judging AI's actual impact.
"We have 16 years of data for each position. That sets us up for a healthier evaluation of whether AI contributes or not than a lot of organizations would otherwise have," Brannon said.
That means Coalition can compare how long a task took before AI to how long it takes now, and use that difference as the basis for a bonus.
"Based on the measured time savings or output improvement, that's how we establish the bonuses we give out for AI implementations," Brannon said. "We have a demonstrated test that says a task that took nine hours can now be done in three with AI. We're able to say you're saving six hours across a team of 35 people every month, and we create a bonus attached to that win."
That kind of proof is rare. A widely cited "GenAI Divide" report from MIT found that approximately 95% of corporate generative AI pilots show no measurable financial return, based on a review of more than 300 AI deployments. Brannon said Coalition rewards AI use only after the numbers back it up.
The bonuses aren't only about time saved. Brannon said Coalition has worked to position itself publicly, to employees and clients alike, as a place that uses AI well, not one that uses AI to replace people.
"Our goal as a company is to be positioned as the best users of AI in our respective spaces, rather than the place you go to hire an AI to replace human labor. We've been committed to that messaging publicly and with our clients, and that's helped reassure team members that we're committed to it," Brannon said.
How the training actually works
Every Coalition employee, all roughly 250 of them, completes AI coursework through the company's internal learning system, and it isn't optional. That kind of blanket requirement is unusual. Elsewhere, training remains limited even as AI use grows, according to a recent industry survey.
"We do have formalized required training. Our raises and internal promotions are tied to successful completion of courses and quizzes, and some hands-on demonstrations of your knowledge. That includes the AI coursework, so it's directly tied to advances in pay," Brannon said.
There's no single person running the program, either. Coalition doesn't employ a dedicated AI upskiller; instead, responsibility sits with division heads.
"We've made that a priority for them, and they're reporting to senior leadership on AI testing and use weekly, given the pace of movement," Brannon said.
That weekly reporting cadence is meant to catch shifts quickly, before a plan goes stale. HRD America has reported that many companies' AI training plans are already outdated by the time they roll out.
The risk of doing AI for AI's sake
Brannon said one of the earliest lessons Coalition taught employees wasn't how to use AI, but when not to bother.
"A really common example we see is people using AI where they run up high token counts and maybe aren't even doing something better," Brannon said.
He described that as a "one shot" approach, where someone asks AI to complete an entire task from start to finish on its own, rather than pairing it with a more structured tool, like a spreadsheet with built-in formulas, that gives a clearer and more accurate result.
"Understanding when to use AI has been a big piece of ensuring we're not using AI just for AI's sake," Brannon said.
More companies are facing that same choice as adoption grows nationally. Recent U.S. Census Bureau survey data shows overall AI use among American businesses climbing to roughly 17% to 20% between December 2025 and May 2026, reaching approximately 37% among firms with 250 or more employees. That's the kind of employer HRD America has reported risks scaling mistakes rather than results when training can't keep pace with deployment.
The same test, applied everywhere
Brannon said understanding a tool's actual operational benefit, rather than assuming a new AI feature is automatically an improvement, has been one of Coalition's biggest ongoing challenges.
"One of the biggest challenges is understanding the actual operational benefits of adding all these AI tools in. Are we just adding in extra software and novelty through AI, or are we actually doing something that provides a meaningful improvement to existing workflows," Brannon said.
That's part of why Brannon's advice for other leaders is to start small.
"I'd emphasize starting with bite sized executions rather than large pivots, the same way you'd build confidence with a new hire, then build toward bigger things," Brannon said.
At Coalition, AI earns its keep the same way any other investment would. It gets tested, measured and paid for only once it proves itself.
