The New Code Is Already Being Written: Today’s Thinkers on What AI Transformation Really Requires

By Dana Houston Jackson, Founder and Chief Operating Architect, Change – With a Coach Approach, LLC

The Inherited Operating System for Organizations – Expired Assumptions | Part 7

The ideas we need are out there. Some of the newest ones we were taught in the past few decades are standing in the way.

If you started your career in the 1990s or 2000s, you probably think Taylor and Fayol are ancient history. You learned the modern stuff: reengineering, Six Sigma, the Balanced Scorecard, ranking your people on a curve. Here’s the uncomfortable part. A lot of that “modern” management was the old operating system with a spreadsheet.

If your AI efforts keep hitting a wall, some of the wall is newer than you think. This article looks at today’s professors and big thinkers who are writing the new code, then at the newer ideas that have expired too. And with all my articles, I’m going to ask you questions. Take the second and think. This article ends with five questions and one small step.

The world they're solving for

Today’s thinkers are working on a problem Taylor never had. Information is everywhere, work changes every few months and a growing share of tasks can be done by software that learns. The research is starting to show what that means. In a large study of customer support agents, Erik Brynjolfsson, Danielle Li and Lindsey Raymond found that a generative AI assistant raised individual productivity by 15% on average (using their criteria), with the biggest gains for less experienced workers. In a field experiment with 758 Boston Consulting Group consultants, Fabrizio Dell’Acqua, Ethan Mollick and colleagues found that on tasks inside AI’s capabilities, consultants completed 12.2% more tasks, 25.1% faster, at more than 40% higher quality. On a task outside those capabilities, AI users were less likely to get it right.

That’s the whole challenge in two studies. AI can make people much better, and it can make them confidently wrong. Getting the first without the second is a design problem, not a software problem.

What they're building

Read today’s leading thinkers side by side and a few shared ideas stand out.

• Amy Edmondson, Harvard Business School. The Fearless Organization (2018) and Right Kind of Wrong (2023) show that teams learn only when it’s safe to speak up and fail intelligently. AI adoption is one long series of small experiments, so this is the foundation.
• Ethan Mollick, Wharton. Co-Intelligence (2024) argues for working with AI as a partner, keeping a human in the loop and learning by trying. His “jagged frontier” is a reminder that nobody knows in advance where AI helps and where it fails. You find out by doing.
• Erik Brynjolfsson, Stanford. In “The Turing Trap” (2022), he warns against aiming AI at imitating and replacing people, and argues the bigger gains come from augmenting what people can do.
• Gary Hamel and Michele Zanini. Humanocracy (2020) makes the case that bureaucracy is the real tax on innovation and that authority should move to the people closest to the work.
• Rita McGrath, Columbia Business School. The End of Competitive Advantage (2013) argues that advantages are now temporary, so organizations have to be built to keep moving, not to defend one position.
• Alexey Krivitsky, Craig Larman and Roland Flemm. 10X ORG treats organization design as a lever for capability: not ten times more activity, but ten times more ability to learn and adapt.
• Paul Cheek, MIT Sloan. No One Works Here (2026) describes organizations as networks of people and AI agents, with people moving from doing tasks to setting goals and overseeing systems.
• Frederic Laloux. Reinventing Organizations (2014) documents self-managing organizations, like Buurtzorg, that already work this way at scale.

Put together, the new code looks like this: safety to learn, people in the loop, authority near the work, structures built to adapt and AI aimed at making people better.

The newer code that's also expiring

Each of these ideas solved a real problem in its time. Each one, used the way most of us were taught, now gets in the way of AI.

• 1970: Milton Friedman’s New York Times Magazine essay argues business’s social responsibility is to increase its profits. It held managers accountable for other people’s money. Turned into quarter-to-quarter shareholder primacy, it makes AI a cost-cutting tool first. In 2019, 181 CEOs of the Business Roundtable signed a statement committing to all stakeholders.
• 1980s–2000s: Forced ranking, made famous by Jack Welch’s “vitality curve” at GE, pushed managers to make hard calls. It also taught people to compete with the colleague at the next desk. Microsoft dropped its version, stack ranking, in 2013. AI adoption needs people to share what works, and nobody shares when sharing helps the person they’re ranked against.
• Mid-1980s onward: Six Sigma, born at Motorola and made famous by GE after 1995, cut defects in repeatable work. Applied to everything, it treats variation as the enemy. AI adoption needs variation: many small experiments, and plenty that fail.
• 1990–1993: Michael Hammer’s “Reengineering Work: Don’t Automate, Obliterate” (1990) and Reengineering the Corporation (1993), with James Champy, told companies to redesign processes, not pave over them. Good advice. In practice it was often done to people, from the top, and became shorthand for layoffs.
• 1992: Robert Kaplan and David Norton’s Balanced Scorecard got leaders looking beyond finance. Cascaded down as a control dashboard, it turns into Fayol’s checker with more colors. Today it shows up as “AI usage rate” targets that measure logins, not better work.

For contrast, picture a business with no inherited code to unlearn. The (Un)Common Business, founded by Callan Faulkner, didn’t bolt AI onto an old operating system. It was built on AI from day one. She told the class she started it in April 2024, after a mastermind retreat with “five other incredible women” where she was “doing this ChatGPT demo for them.” They told her to go all in, so she shut down her real estate investing and tech consulting businesses. In the bootcamp I attended, she described starting on her own, hoping 10 people would come to her first AI masterclass. The REtipster Podcast reported she built a business of about $4 million in under 18 months. By September 2026, she told us, the company had roughly 60 to 70 people working alongside more than 150 AI employees and, “very conservatively,” would do about $40 million in revenue this year. What stood out to me wasn’t the tools. Every AI employee has a human boss, and quarterly reviews ask people what they automated. And, by her count, nobody has been fired for automating their own work: “We have just promoted them.” That’s the opposite of a ranking curve. Those are her figures, not audited ones, and a young company doesn’t carry the baggage a 100-year-old one does. But it shows what’s possible when nobody has to unlearn the old code first.

Where it still shows up

• AI business cases built only on headcount savings.
• Performance systems that rank people against each other while asking them to collaborate on AI.
• Process improvement teams that want a defined, repeatable process before anyone may experiment.
• Scorecards with “% of employees using AI” as the headline metric. That’s counting cattle, not checking whether they’re fed.
• Transformation programs run from a steering committee and announced to everyone else.

What it costs us now

Here’s the irony. Hammer’s best line, “don’t automate, obliterate,” is exactly the right instinct for AI: redesign the work instead of bolting AI onto the old process. But because reengineering was done to people, many of them hear “redesign” and brace for a layoff. So they hide what AI can do for their job. The newer code taught people that improvement is a threat. Now we need them to lead it.

The bigger picture: Taylor, Weber and Fayol still shape the default operating system most organizations fall into: specialization, hierarchy, rules, measurement and management as control. Most of the management ideas of the last 30 years didn’t replace that system. They upgraded it: better measurement, tighter control, sharper competition between specialists. As AI agents take over specialized tasks, the people whose specialty is changing need the opposite of a ranking curve. They need room to learn, a say in the redesign and a reason to believe there’s a place for them on the other side.

What to do instead

• Start with safety. Before the tool, make it safe to experiment and to say “this didn’t work.” Edmondson’s research is the place to begin.
• Augment first. Use Brynjolfsson’s test: does this use of AI make our people more capable, or does it only replace them?
• Map the jagged frontier together. Mollick’s advice: let teams try AI on their real work and share where it helps and where it fails.
• Move authority to the work. Hamel and Zanini’s question: what decisions could the people closest to the work make if we let them?
• Measure capability, not activity. Replace usage counts with outcomes: faster decisions, better work, fewer handoffs.
• Retire the curve. If people are ranked against each other, don’t expect them to teach each other.

Unlearn to make room

These may be the ideas you were promoted for. Here’s what’s worth unlearning.

How today’s thinkers fit with the others: Follett, McGregor and the adaptive leadership thinkers in Part 6 were right early. Today’s thinkers are proving it with data. Edmondson is McGregor’s Theory Y with a research lab. Hamel and Zanini are Weber’s critics with a business case. Mollick and Brynjolfsson are adding the new piece: how people and AI work best together.

Where AI meets the newer code

Where it clashes: Forced ranking makes people hoard what they learn about AI. Six Sigma thinking wants a stable process before anyone experiments. Shareholder-value pressure aims AI at cutting costs before anyone has learned what it can create. Scorecards measure usage, so people log in and change nothing.

Where it helps: Some of the newer code is worth keeping. Reengineering’s instinct to redesign, not just automate, is the right one. Six Sigma’s discipline belongs in the parts of the work that are proven and repeatable, and AI can help there. A scorecard that balances people, customers and learning, not just money, is a good idea when it’s used to learn, not to control.

What has to be redesigned: Performance systems that reward sharing, not ranking. Improvement methods that make room for experiments. Business cases that count capability, not just cost. And a transformation approach built with people, so “redesign” stops sounding like a threat.

Check your own organization

Picture a video post-production studio in 2025 that bills clients by the editor-hour. An assistant editor uses AI transcription and search to log and organize 60 hours of documentary interviews in an afternoon instead of three days, then spends the time she saved building a stronger first cut. The next week, the utilization report flags her as underperforming. She billed 20 fewer hours. Her supervisor asks what happened. She says she was “catching up on organizing,” and she stops mentioning the AI. The editor across the hall, ranked against her on the same report, never asks how she did it. Nobody meant to punish her. The studio’s operating system did it automatically. The fix isn’t a memo encouraging AI use. It’s rethinking what the studio measures and how it prices the value of its work, with the editors in the room.

• Which of your metrics would punish someone for using AI well?
• Are people ranked in a way that makes them compete instead of share?
• Does your improvement method allow experiments, or only proven processes?
• Is your AI business case built only on cutting costs?
• Which of today’s thinkers would recognize your organization, and which would be worried by it?

One small first step: Pick one metric your team reports every month. Ask the team, “If someone used AI to do this work better and faster, would this number make them look better or worse?” If the answer is “worse,” change the metric before you ask anyone to change how they work.

Next in the series: the way forward, and how to choose our operating system on purpose.

Sources