A 550-Person Media Company Hit 65% Daily AI Use by Hiring for Curiosity Over Process Knowledge

65% of the people at Good Karma Brands use AI on a working day. The company has more than 550 teammates, gave out north of 400 licenses, and got there in under a year. On some days the number goes to 75%.

That’s the headline figure from episode 227 of The Artificial Intelligence Show, where Ty Bauschek, the company’s senior director of sales development and innovation, walked through how a Milwaukee sports media company that runs ESPN Radio’s national operation put AI in front of everyone and made it stay there.

The more useful part is what happened after adoption worked. Good Karma created a full-time role called innovation specialist, and staffed it from inside with generalists.

“people who are AI hungry and curious and generalists are actually able to pick up our workflows easier than conversely teaching people how to use AI who are in the workflows every day.”

Read that as a hiring criterion, because that’s what it is. The company tested two theories against each other and picked one. Teaching the process to a curious outsider beat teaching the tool to the person who already owned the process.

How they got the number

The rollout was deliberately slow for something described as urgent. CEO Craig Karmazin, who founded the company in 1997, did his own comparison of ChatGPT, Claude, and Copilot before committing. The published show notes date the companywide distribution to Labor Day. Around Labor Day last year the company hired two senior software engineers and gave every full-time teammate a ChatGPT Enterprise account.

Then it withheld most of the accounts. A pilot group of about 50 started the Tuesday after Labor Day, with weekly check-ins and a Microsoft Teams group chat for problems. Seven more Tuesdays, 50 people each, so the later groups inherited what the earlier ones had already worked out.

The habit is maintained the way a media company maintains anything: on a schedule. There’s a newsletter every other Tuesday, AI training on the alternating Tuesdays, and a standing AI segment in every companywide, sales, marketing, and content meeting where a teammate shows what they did that week.

Bauschek frames the resistance he’s designing against as a sentence.

“we’ve always done it this way”

He calls those the most expensive words in business.

Two results worth stealing

The first came from sales support. Marketing consultants need to know which ad copy is missing before a deadline, and with 19 consultants in Milwaukee somebody was pulling that list by hand for each one. A teammate automated it into emails that arrive with 48 hours of warning. Bauschek puts the saving at 10 to 15 hours a week.

The second is the one that created the job. A marketing coordinator was spending two to three hours building a single campaign deck, manually pulling slides and logos before the daily four o’clock call, at three or four campaigns a day in a busy week. She built a tool the company calls CAM, a campaign assistant manager. The build dropped to 10 minutes. When she flew to Chicago and helped that office build its own version, the pattern became a proposal: make this somebody’s actual job.

That’s the mechanism behind the new role, and it’s the opposite of how these stories usually go. We’ve written about the vanishing entry-level rung and the experience debt that follows. Here the generalist didn’t get cut. The generalist got promoted into a function that didn’t exist, on the strength of having built something.

Ideas now go through an innovation scorecard graded out of 30. The specialist scores it, leadership scores it, and a GPT scores it as an unbiased third pass before anything gets built. Bauschek describes his own job in the middle of that as absorbing enthusiasm so the specialists can work.

“right now I’m acting as the strategic filter”

The part that should temper the number

Adoption at this level has a bill attached, and Bauschek says so without being asked.

“I think like a lot of other companies in the world right now we’re running out of credits”

That’s the sentence to sit with. A 65% daily usage rate is a cost curve as much as a culture win, and the company’s response is to coach people on using the tool more efficiently. Usage went up, and now consumption is the thing being managed. It’s the same meter we described in the maintenance tax on marketing agents, showing up at a different point in the stack.

Two other cautions. This is one company’s account of itself on a podcast series presented by Google Cloud, with no independent audit of the numbers. And 65% daily usage measures logins, not outcomes, which Bauschek is candid about: the company deliberately shifted from tracking usage to tracking impact, because the first number was easy and the second one matters.

What a 5-to-30-person team can take from it

Most of this doesn’t need 550 people or two staff engineers.

1. Roll out in batches, not all at once. 50 a week meant every group after the first learned from a real internal precedent instead of documentation. A 5-person team can do this with two people and two weeks.
2. Put AI on a recurring agenda item, not a training day. A standing slot in an existing meeting where someone shows one thing they did is cheaper than a workshop and it’s what kept usage from decaying.
3. Watch who’s actually building. The person who automates something is telling you what your next role should be. At Good Karma that signal came from a marketing coordinator, not from the org chart.
4. Price the credits before you scale the habit. The cost lands after adoption succeeds, not during the pilot, and it arrives as a surprise if nobody modeled it.

The reason this story matters isn’t the 65%. It’s that a company looked at its own generalists, decided curiosity was the scarcer input than process knowledge, and wrote a job description around that. Most companies are still doing the reverse.