I000°
They start with the constraint, not the tool
The question is never “where can we apply AI?” but “what’s actually limiting our performance?”
Accelerating a step that isn’t the constraint just piles work up downstream: AI-drafted contracts flood the legal reviewers; ten-times-faster content buries compliance. The book calls this the bottleneck paradox, and it’s why so many “wins” quietly damage overall throughput. AI Native leaders start with WHY before they take a leap with technology... and sometimes the fix is connecting two systems, writing a clearer policy, or having an honest conversation.
Chapters 1 & 3 · Promises Versus Reality · The Bottleneck Paradox
II030°
They choose the simplest solution for the right lifespan
Every model and agent you deploy is a liability with a run-rate. Deterministic rules are less costly to maintain.
The hierarchy is explicit: connectivity before intelligence, deterministic solutions before stochastic, and automation before goal-seeking agents. A twelve-rule expense policy doesn’t need a stochastic agent that miscategorizes meals as travel and demands perpetual "prompt-tending"... just a script that can runs for years at zero marginal cost. Stochastic solutions are reserved for problems that genuinely require judgment and pathfinding. Simplicity compounds. Complexity you don’t need is risk you choose.
Chapter 2 · Why Less Is Now More
III060°
They invert the headcount question
Not “how many people does this work require?” but “how few people can own it end to end?”
Brooks’s Law never went away. Coordination costs still grow faster than the capacity people add, but AI makes large execution-only teams a liability. Every person you don’t add saves a salary plus the coordination tax everyone else would pay. The ideal to learn from is the vertically integrated team of one; the practical answer is the crew: a small group so aligned on mission, roles, and standards that it approaches the coherence of a single mind.
Chapters 2, 5 & 6 · The Coordination Tax · Crews
IV090°
They build learning loops, not roadmaps
Learning (not tokens used or agents deployed) is what drives improvement.
Adopting tools doesn’t improve organizations; learning does. AI Native leaders run both loops: adjusting prompts and inputs when output is poor (single-loop) and stepping back to ask whether they’re solving the right problem at all (double-loop). Time to improve teaming, focus on goals over tasks, opportunities for social learning and shared practice, and visible results that help you recognize what "good" means are an operating system. Without it, adoption follows a predictable arc toward silent abandonment.
Chapter 1 · Learning Loops · Argyris’s Double-Loop Learning
V120°
They measure outcomes, not activity (& lock baselines)
Vendor ROI projections are marketing. Seats, logins, and prompt counts are theater.
Independent evaluations put real, measurable AI value capture at roughly one organization in twenty. The winners decide how they’ll measure impact before they fund exploration, lock baselines before launch, and track enablement in metrics the business already trusts... like defect rates, revision cycles, or time to value, disaggregated by AI use. They also read self-reported enthusiasm with caution: social penalties distort every AI pulse survey, so they reward honest disclosure instead of measuring the facade.
Chapters 1, 6 & 9 · From Adoption to Enablement · The ROI Firewall
VI150°
They anticipate the power shifts in advance
People don’t resist tools. They resist what tools do to their security, standing, autonomy, and relationships.
Since Markus’s 1983 study, the pattern has held: the losers of a power shift rarely object out loud. They just become self-protective, and adoption quietly dies. Employees fear losing autonomy almost as much as losing their jobs, and identity disruption is neurological, not attitudinal. A threat to a hard-won professional identity reads as a threat to survival to the nervous system. Power maps predict adoption better than any feature list, so AI Native leaders start with the org chart and culture, negotiate honestly with those who lose, and treat the displaced honorably.
Chapters 1, 6, 8 & 11 · The People Problem · Win With the Humans
VII180°
They demand a dollar figure and a ten-second purpose
If an AI idea or initiative can't name the revenue it increases or the cost it reduces, it is a hobby.
Every initiative answers to a purpose statement that can be fully understood in under ten seconds, twenty-five words or less, including what AI is not allowed to erode. Ideas are judged by financial impact and effort, projects that don’t serve a top-three organizational or departmental goal don’t get pitched, and the bottom half of the list gets killed... publicly. Focus is a leadership act.
Chapters 9 & 11 · The Coffee Test · The Profitability Compass
VIII210°
They define “good” before they automate
A process may only go dark once it has earned the dark.
The AI Quality Gap is the distance between how good AI looks in a demo and what it delivers over time... and the demo is the best it will ever look. Before withdrawing human judgment from any workflow, AI Native leaders ask three questions: what does “good” look like (expressed precisely enough for a machine to check it)? Is the cost of an undetected error bounded and recoverable? Which actual human stays accountable? Dimensions you can safely ignore while a human approves every output may become mandatory the moment that human steps away.
Chapters 4, 7 & 12 · The AI Quality Gap · Know What “Good” Means
IX240°
They architect accountability
AI compresses labor, but does not absorb accountability.
The parts of a business that scale effortlessly are the parts that never touch the physical world; the parts that break are real products, real bodies, real regulators, real money. A human rubber-stamping AI output at volume isn’t oversight but "liability theater", and regulators are learning to see through it. Real accountability is a named owner with the expertise to judge the outcome, the authority to change the system, and a genuine stake in getting it right.
Chapter 6 · Unicorns & The People Problem
X270°
They turn shared context into an AI asset
Every AI capability performs exactly as well as the shared understanding it draws from.
A governed context layer (the processes, roles, standards, and decision rules a quality management system already documents) is the highest-leverage AI infrastructure a company can build, because it makes every assistant and agent perform better instantaneously. Feed AI your “digital exhaust” instead, and you’re paying inference costs to have your own confusion summarized back to you. Artifacts from Baldrige and ISO 9001 are now AI assets.
Chapters 4 & 7 · Quality Management Systems as the Foundation
XI300°
They hunt the eighth waste: confusion
Plausible-but-wrong AI outputs are more expensive than no outputs at all.
Confusion is waste, and AI-enabled work can manufacture it at scale. People act on polished-but-inaccurate AI results with unwarranted confidence; agents take actions without leaving breadcrumbs; all while real effort duplicates and trust erodes. The countermeasure is clarity as a design principle: provenance, logs, and explanations, so every AI output that feeds a decision is traceable to what produced it and who owns it.
Chapter 2 · The Eight Wastes of AI-Enabled Work
XII330°
They retrain collective attention towards company wins
The opportunities you’re missing aren’t hidden. They’re invisible... which is worse.
Like the gorilla in the famous attention experiment, breakthrough value sits in plain view of people trained, hired, and rewarded to watch the basketballs of efficiency metrics. Individual productivity gains evaporate into absorbed time and extra meetings, while the returns that justify AI investment live at the collective intelligence level. Transformation means training people to see opportunities at the team level rather than just individual improvements, a forever discipline.
Chapters 3, 5 & 13 · The Gen AI Value Creation Pyramid
