Culture is your operating system

Say "human-centered" in an executive meeting and half the room hears soft. They picture a culture that tolerates coasting, and they reach for what feels rigorous instead: mandates, adoption targets, monitoring dashboards.

I understand the instinct. I spent 25 years leading teams through hard change, at Google, at Slack, and as a startup CEO, and the pull toward control when the board is nervous and the quarter is closing is real. I have felt it. It is also the most expensive mistake on the menu. But human-centered doesn’t mean soft: it means doing the hard work of both holding people accountable to performance and supporting them to do their best work.

The fastest way to correct the misconception is with a good case study, like Eric Severson's numbers.

What "demanding" actually looks like

When Severson rolled out a results-driven, high flexibility approach at Gap and later at Neiman Marcus, he gave people more control over when, where, and how they worked, and paired that freedom with strict accountability for results. Within six months of each rollout, unwanted voluntary turnover dropped by half or more.

Involuntary turnover doubled, and in some cases tripled. Roughly 2% of the workforce became 6%. The kind of results that CEOs and CFOs love to see: retain great workers, remove the slackers.

That is the part leaders skip when they borrow the flexibility and leave the accountability behind. Once managers stopped evaluating work ethic and started evaluating outcomes, Severson told me, "all these people seemed to pop up who hadn't really been contributing for years, but they were always showing up." The people who had built careers on looking busy ran out of places to hide.

The employees who stayed loved it. Your high performers have spent their careers watching the meeting marathoners and the office politicians collect the rewards while they carried the work. This wasn’t just an office worker change. Frontline workers gained more control over shifts, for example, and Neiman Marcus store associate retention hit 75%, the inverse of the industry average. That translates to productivity and lower replacement costs.

Demanding and supportive at the same time. High standards, real backing to meet them, and leaders held to those standards first. Nothing about that is soft.

AI is running the same test, faster

The old proxy for performance was looking busy. The new proxy is output volume, and AI has made that one nearly free to fake; even Amazon and Meta have had to back away from volume metrics.

Which lands the whole thing on managers. Three things I keep finding in nearly every research study and conversation with leaders getting real ROI out of AI:

Manage to goals, not output. Indeed's Hannah Calhoon describes tracking a "portfolio of metrics." Her team measures how fast engineers ship code, and also how often that code has to be revised. Her warning is the argument in one sentence: "If you focus just on that acceleration and productivity win, you create a system that will dump tons and tons and tons of workslop into everybody else's experience in a way that's really bad for the business." An 80% adoption target is exactly that system. It measures activity, and activity is the one thing AI can now generate infinitely.

Buy your teams time to learn together. I pulled together leaders from four companies advanced in AI leverage to compare, blind, their lessons learned. The most central was that teams are the center of healthy AI adoption. Setting standards for quality, understanding norms (“Is it OK to use AI for my performance appraisal?”) and learning together that drives a focus on team goals, not individual usage. One champion on a team who figures it out and builds skills for others is worth 10 top-down mandates..

Adoption travels sideways, through peers, or it does not travel. That makes managers the load-bearing wall. Pinterest's Doniel Sutton says it better than I can: managers are "player-coaches who actively shape the culture of our company by creating space for teams to learn, grow and develop." And then the line I would tape to every CEO's monitor: "Some people think AI is making managers optional. I'd argue it makes the good ones matter more than ever."

Get your hands in the work. BCG’s Gabriella Rosen Kellerman, who helped coin the term "workslop", offers the most useful management practice I have heard all year. For one week, keep two folders: AI-assisted work that impressed you, and AI-assisted work that did not. Then share both with your team. "You're their manager and ultimately the work products are for you." That is what a quality bar looks like in practice, and you cannot delegate it to a policy or a training module.

You also cannot do it from inside back-to-back meetings, or while managing 12-15 direct reports and a slew of agents. The no-so-quiet flaws in cutting your management layers and thinking they can all player-coach their way through the biggest change to hit the world of work in a century.

Culture is the operating system

Culture is the connective layer that aligns your people, decisions, and performance. Just as an operating system determines how software interacts with hardware, your culture determines how your teams interact with your business strategy. If the distance between your stated culture and the reality of your employees experience is sizable, friction is the result—and your AI investments will only amplify the friction. But if it’s aligned, AI acts as a force multiplier for growth.

That’s true of any technology. At Slack, every new channel was public by default. That was a deliberate choice. We believed companies move faster when information is shared openly, so we built the belief into the product. Microsoft Teams defaults to private, which reflects a different and defensible belief about how information should travel. Two tools in the same category, each one quietly training the behavior it assumes.

AI does the same thing, at far greater speed and scale. Point it at a culture built on monitoring and you get faster monitoring and more polished compliance. Point it at a culture built on growth and you get growth.

The data holds up. People-centric organizations are seven times more likely to be further along in AI adoption, twelve times more likely to have C-level executives directly engaged, and six times more likely to be investing in training. The conditions that make AI work are the conditions that make people work: clear goals, psychological safety, honest feedback, and decisions made close to the work.

Where to start

• Audit your metrics for output theater. If you are tracking adoption rates, you are measuring the wrong thing. Zapier's Brandon Sammut is blunt about it: "97% adoption rates mean almost nothing."
• Deal with your low performers. Your best people are watching to see whether you will tackle the politically tied but not-really-competent underperformers.
• Protect learning time, and stop taxing your early adopters. The reward for figuring it out cannot be more work. One place to start: focus on closing the gaps on cross-functional alignment and connection.
• Get in the work yourself. Say out loud what good looks like, then say it again next quarter when the tools have changed underneath you. Show your bumbling along the way.

Walmart and Amazon bought the same technology. What separates them is whether they’re thinking human-led, AI-enabled or the opposite. That decision is your culture, and AI is about to broadcast it to everyone who works for you.

Get more insights from Brian and watch his session from Culture First Forum North America.