AI
Why Organizations Resist Change
Craig Larman’s law holds that organizational systems are self-perpetuating. The feedback loops inside any company are designed, consciously or not, to reinforce whatever the current way of working already is. Holub traces this to MBA programs that train managers to run software teams like Taylorist factories, timing output like items on an assembly line.
The third clause captures the stalling pattern: anyone proposing change must “prove it will work,” while nobody applies that standard to the status quo. Holub’s counter: prove that what you’re doing now works. Nobody can. “Works” is always relative to something.
Why Logic Fails as a Change Strategy
“rational thinking does not work” — Allen Holub
Engineers default to building cogent arguments, expecting logic to move people. It doesn’t. Dunning-Kruger explains why: people calibrate “normal” against their immediate social group. If every manager around you demands estimates and runs sprints, that looks correct, not because anyone reasoned their way to it, but because the peer group reinforces it. An outsider presenting evidence is asking a manager to adopt a belief their peers don’t hold, triggering social ostracism, not reflection. The EJB era is the same dynamic in architecture: it took spectacular failures before the Java community abandoned a stack nobody had actually thought through from first principles.
Speaking the Language Managers Actually Understand
Managers respond to three things: risk, money, and status. Any change proposal must address all three. Lower the risk of the new approach, show it saves real money, and make the manager look good to their peers. Convincing one manager is not enough; social pressure will pull them back unless the peer group shifts too.
Holub’s consulting rule follows: one-on-one sessions work with CEOs and CTOs because positional weight lets them override the group. Middle managers need group interventions, not individual conversations, because change only sticks when the social group around a person changes with them.
The Improvement Kata: Small Experiments Over Transformation
Mike Rother’s Toyota Kata formalizes the approach: start with a distant goal, assess the current situation, then run the smallest possible experiment toward that goal. If it works, run another. If it fails, adapt or try something different. No single experiment carries meaningful risk, which removes the accountability fear that kills large change initiatives.
Holub’s go-to first experiment is reducing work in progress to one item per team. Every time he has done this, throughput doubled or tripled. Run it with two teams for a month. If the results are good, competent management will ask why the other teams aren’t doing it too.
Time Value of Money as a Technical Argument
The time value of money is a concept every MBA already knows. €1,000 received today at 10% interest becomes €1,100 in a year. €1,000 received a year from now stays €1,000. Holub uses this to reframe delivery speed: delaying a release is a financial loss, not a schedule slip.
“We are expensive.” — Allen Holub
Load factor sharpens the point. A €100,000 salary costs the company closer to €200,000 fully loaded. In San Francisco, that means $1 million buys two programmers. Teams releasing weekly stop getting asked for estimates, because working software provides the assurance estimates were supposed to give.
Costing Ceremonies to Build the Business Case
“What matters is money, risk, and status.” — Allen Holub
A six-person team at $100 an hour spends $600 per standup. With overhead and post-standup conversations, a “15-minute” meeting reliably takes an hour. Add the end-of-day integration meeting. Across 10 teams and 200 working days, that is $2.4 million a year on coordination. Ensemble programming eliminates both: the team integrates every few minutes as they work, so there is nothing to scatter or gather.
Management objects that paying five people to watch one person type is wasteful. The actual structure is five people thinking while one records in code, rotating every 15 minutes. The work never stops when an individual steps away.
Notable Quotes
rational thinking does not work Allen Holub · ▶ 15:23
We are expensive. Allen Holub · ▶ 44:06
What matters is money, risk, and status. Allen Holub · ▶ 55:53
Key Takeaways
- Managers respond to risk reduction, cost savings, and status, not agile principles or sprint ceremonies.
- Change spreads through social groups, so targeting one manager rarely works without shifting the whole peer group.
- Small experiments framed as low-risk trials remove the accountability fear that blocks every large change initiative.