Leading in the Age of Many Right Answers: From Autonomous Transformation to Decision Intelligence

Cassie Kozyrkov captures this tension brilliantly. She points to a striking fact: despite tens of billions invested in generative AI projects, only about 5% of enterprises are seeing measurable returns. The problem isn’t the technology. It’s the mindset.

Traditional AI was like target practice: one input, one correct output. Did the system recognize the cat? Did it predict the next word accurately? Metrics like accuracy and precision worked because there was a single bullseye to hit.

Generative AI is different. It is more like jazz: improvisations, riffs, endless right answers. Ask a GenAI tool to write a product description, and it might produce three equally strong but stylistically different outputs. Success isn’t about correctness. It’s about usefulness, fit, and resonance.

Yet organizations are wired for certainty. Boards, CFOs, and executive committees still ask the familiar question: “What’s the ROI?” They want numbers in a spreadsheet, proof of accuracy, linear cause-and-effect. But GenAI resists that framing. Its value lies in intangible but profound shifts: faster ideation, reduced employee burnout, richer customer experiences, new creative possibilities.

I’ve lived through this dilemma in my own roles. Too often, organizations measure what is easiest, not what is most meaningful. I’ve seen projects celebrated because dashboards were green: sales targets delivered, impressions achieved, content published, while deeper strategic value went unnoticed. This is what Cassie calls “polished inefficiency”: activity that looks impressive but fails to create the capability, resilience, or transformation that truly matters in the long run or as S. Sinek would put it, what truly matters in the ‘infinite game’.

The leadership test, then, is whether we can resist the gravitational pull of easy metrics, and instead take ownership of defining what value really looks like in a world of many right answers.

Brian Evergreen’s Autonomous Transformation offers a profound counterpoint. He argues that transformation is no longer an episodic event. It is continuous reinvention, powered by AI and guided by human purpose.

For decades, transformation was project-based: a digital launch here, a restructuring there, an AI pilot program when budgets allowed. Evergreen points out that this episodic approach no longer suffices. The pace of technological change is simply too fast, and the competitive environment too unforgiving. Transformation must now be treated as a living capability, not a project with a start and end date.

One of his most resonant warnings is against digitizing inefficiencies. It is tempting to apply AI to speed up processes that are familiar but outdated. But doing so simply accelerates waste. Leaders must ask: are we truly reimagining the way value is created, or just making yesterday’s bottlenecks move faster?

Examples abound. Amazon institutionalized “Day 1 thinking” to avoid stagnation, embedding continuous reinvention into its DNA. Over the last two decades, it launched over 20 new consumer-facing businesses; from AWS to Prime Video to Alexa, ensuring that transformation is not episodic but constant. Patagonia, too, demonstrates what Evergreen means by aligning transformation with purpose. From reinvesting 1% of sales into environmental causes to transferring ownership in 2022 to a trust dedicated to fighting climate change, the company treats transformation not as a one-off initiative but as a continuous process of alignment between business and values.

What I admire in Evergreen’s framing is the emphasis on systemic alignment. Continuous reinvention isn’t achieved through a scatter of pilots or one-off innovations. It requires aligning culture, governance, technology, and strategy so the organization can adapt as naturally as a living organism.

This resonates with my own leadership journey. I’ve seen firsthand how organizations that treat transformation as a project struggle to adapt when the environment changes. By contrast, teams that cultivate adaptability as a muscle, making reinvention part of their DNA, are better prepared for what comes next. Success in this era is not crossing a finish line, but sustaining the capacity to keep running, no matter how often the terrain shifts beneath our feet. Falling in love with ambiguity is the best thing a leader could wish for.

If Evergreen gives us the vision, Kozyrkov gives us the practical scaffolding. Her Leader’s Playbook for GenAI Metrics dismantles the illusion of universal KPIs and replaces it with a framework for navigating subjectivity.

Some of her core principles are deceptively simple and simple things are the most overlooked:

• No metric is universal. What works in one domain may be irrelevant in another.
• Metrics must ladder to strategy. They should connect to the purpose of the initiative, not generic AI benchmarks.
• Don’t overmeasure. The overhead of measurement must not outweigh the benefits.
• Metrics are leadership choices. The most important governance question is: Who decides what success looks like?

She outlines a step-by-step playbook:

1. Purpose & Objectives – Why are we doing this? What are our goals? Who owns the definition of success?
2. Intended Use & Scope – Define clearly what GenAI should and should not do, including its constraints.
3. Capability Decomposition – Break down broad qualities like creativity or trust into measurable sub-dimensions (clarity, novelty, ease of use, user confidence).
4. Operationalization – Translate those sub-dimensions into measurable signals (e.g., readability scores, survey responses, task completion times).
5. Aggregation & Trade-offs – Combine signals thoughtfully, while avoiding Goodhart’s Law (the trap of measures becoming targets).
6. Calibration & Golden Datasets – Test alignment with human judgment and business KPIs, keeping evaluation datasets refreshed.
7. Human Raters – Use human evaluators as noisy sensors, not final arbiters, to ground subjective judgments.

Real-world examples reinforce her point. GitHub Copilot has shown that developers complete coding tasks 55% faster with AI assistance, but surveys revealed that what users valued even more was the reduction in cognitive load and the ability to unblock creative work. Similarly, a Forrester study on Microsoft Copilot found productivity improvements equivalent to about 10 hours saved per employee per month, but outcomes varied significantly by role and context. These cases demonstrate Kozyrkov’s central argument: the metric that matters is not just speed or efficiency, but usefulness in the context of strategy.

What I find powerful in her framing is the insistence that metrics don’t reveal universal truths. They are steering signals, created to help leaders navigate ambiguity. They reduce uncertainty in a given context, but they cannot eliminate it.

As someone who has led cross-functional digital projects, I recognize the temptation to seek the perfect metric. Kozyrkov reminds us that no such metric exists. The task of leadership is not to find the answer, but to choose what matters, to align the organization around it, and to own that choice with conviction.

Put Evergreen and Kozyrkov together, and a new model of leadership emerges; one I think of as decision-centric autonomous leadership.

• From Evergreen: the recognition that transformation must be continuous, systemic, and aligned with purpose.
• From Kozyrkov: the realization that metrics are not absolutes but choices, and that in a many-right-answers world, leaders must decide what “usefulness” means.

The synthesis is clear: AI is not simply a technical capability. It is a leadership test.

In practical terms, this changes how we evaluate success. In e-commerce, for example, it is easy to focus on metrics like impressions served, conversion rates, or search rank improvements. But the deeper questions are:

• Did the system reduce decision fatigue for the customer?
• Did it accelerate trust in the brand?
• Did it align with values that drive long-term loyalty and relevance?

These are not simple metrics. They require judgment. They require leaders to move from managing outputs to shaping meaning. And they require organizations to get comfortable with ambiguity, guided by purpose rather than trapped by dashboards.

The implications for leadership are profound:

• Mindset shift: From correctness to usefulness; from efficiency to resilience.
• Metrics shift: From outputs to outcomes; from bullseyes to steering signals.
• Responsibility shift: From delegating measurement to owning the definition of success.

This is not easy work. It demands courage to operate without the comfort of certainty, and humility to accept that clarity comes not from erasing ambiguity but from guiding teams through it. It also requires a new kind of leadership literacy, one that blends technical awareness with strategic foresight and ethical responsibility.