When AI Outperforms
Your Best Expert

Quick Answer:

The Move 37 Moment for knowledge workers is the inflection point where AI produces a non-obvious, creative output that surpasses what a human expert would have generated — named after AlphaGo's legendary 2016 move against Lee Sedol. For marketers and strategists, this moment has already arrived. The organizations thriving are not replacing humans with AI, but redesigning roles around irreplaceable human judgment. In Latin America specifically, the workforce data shows augmentation producing an 8–14% productivity surge — not the displacement narrative that dominates coverage of the region.

The Move That Changed Everything

March 9, 2016. Seoul, South Korea. The world's best Go player, Lee Sedol, sits across from a screen displaying AlphaGo — Google DeepMind's AI system. The first game of a five-match series is underway, and what's at stake is more than a game. It is a test of whether a machine can master the most combinatorially complex board game ever devised by humans, a game that professional players spend decades learning to read through intuition.

On the 37th move of the second game, AlphaGo plays a stone on the fifth line near the upper right — a position so unconventional that the commentators initially assume it is a mistake. One international commentator pauses mid-sentence. The other leans forward and says, quietly, "That's a very strange move." Professional players watching remotely are baffled. In the top levels of competitive Go, that position simply isn't played. It violated decades of received wisdom about the opening game.

It was not a mistake. It was the move of the match. Lee Sedol left the room. He came back fifteen minutes later. He lost.

Move 37 is now the shorthand for a specific kind of moment: when an AI system produces an output so outside the usual frame of expert thinking that the expert's initial reaction is to dismiss it — until the full implications reveal themselves. It is not AI being smart in the way humans are smart. It is AI being smart in a different way entirely.

That moment has now arrived for knowledge workers.

The Knowledge Work Inflection Point

The Move 37 Moment in knowledge work does not look like a dramatic scene in a server room. It looks like a marketing strategist receiving an AI-generated content brief that she would not have written — one that correctly identified an audience segment she had overlooked, used a framing she had never considered, and did so in under thirty seconds. Her first instinct is to revise it. Her second instinct, reading it again, is to ship it.

It looks like a legal analyst running an AI over a contracts database and receiving a risk flag on a clause that every senior attorney had read and approved. The flag was correct.

It looks like a paid media team discovering that their AI-generated ad variations are outperforming the concepts their best creative director spent a week developing, not marginally, but consistently across multiple markets.

These are not science fiction scenarios. They are happening across industries, and the data from the marketing sector in particular is instructive. According to Semrush research tracking AI tool adoption among marketers, the use of AI for content creation, SEO, and market analysis has accelerated sharply — with a significant portion of teams now reporting that AI-generated first drafts require minimal human editing before publication. That shift is not trivial. It means the floor of AI output has risen past the bar many teams set for human output.

But here is where the analogy becomes more complex. After Move 37, Lee Sedol did not disappear from Go. He continued to play. And in game four of that same series, he found Move 78 — a hand that AlphaGo had not anticipated, a move so human and so specific to the psychological moment that it broke the machine's confidence. He won that game. He remains one of a handful of players in history to have beaten AlphaGo in formal competition.

The lesson is not that AI wins. The lesson is that the game changed — and the players who understood how it changed could still compete, and sometimes prevail, by playing differently.

What the Adoption Data Tells Us

If you are a team leader or HR strategist looking at AI tool adoption among knowledge workers in 2025, the picture is not one of mass automation arriving uniformly across every task. It is more granular — and more actionable — than that.

Across marketing teams specifically, the highest AI adoption rates are concentrated in three areas: content production at scale, keyword and topic research, and competitive analysis. These are tasks that require processing large amounts of structured data, identifying patterns, and generating formatted outputs — precisely the categories where AI performs at or above expert-level capability.

What's more revealing is where adoption is lowest: campaign strategy, brand positioning, client communication, and creative direction. Not because AI tools haven't been applied there, but because the outputs generated have required more significant human intervention. The delta between AI output quality and human expert quality is narrower in pattern-based tasks and wider in judgment-based tasks — for now.

That gap is closing. The question teams need to answer is not "will AI close it?" — the trajectory is clear. The question is: "What do we do with our best people while it does?"

The teams answering that question well are reassigning their senior strategists from production tasks to decision architecture — the work of defining what questions AI should even be asked, how results should be evaluated, and what the outputs mean in a specific organizational context. That is work that requires domain expertise, institutional memory, and something AI does not have: accountability to the people in the room.

The Risk No One Is Talking About Loudly Enough

There is a version of the AI-in-the-workplace conversation that focuses on job displacement — on AI taking roles, reducing headcount, restructuring departments. That conversation is happening. It is important. But it is not the most immediate risk for most knowledge workers reading this in 2025.

The more immediate risk is this: workers who use AI well are replacing workers who do not use AI at all.

This is not a prediction. It is already visible in hiring patterns across creative, analytical, and strategic roles. The benchmark for "strong output" has shifted in organizations that have integrated AI into their workflows. A strategist who produces AI-augmented briefs, analysis decks, and content outlines in the time it previously took to produce one is not just more efficient — they have raised the comparative standard for the entire role. Everyone around them is implicitly being measured against what AI-assisted work looks like.

This creates a polarization dynamic that organizational leaders are only beginning to grapple with. The gap between the highest-performing knowledge workers and the average-performing knowledge workers is widening — not because the top performers are more talented in the traditional sense, but because they have learned how to operate AI systems as a genuine extension of their own capability. They have developed what might be called AI fluency: the ability to brief AI well, evaluate its output critically, identify its failure modes, and integrate its results into work that a client or stakeholder can trust.

The organizations that will be most disadvantaged are not the ones that chose AI over humans. They are the ones that waited too long to decide what kind of humans they needed in an AI-assisted environment.

The Human + AI Capability Framework

Understanding where the boundary sits today — and how it is shifting — is the starting point for any meaningful role redesign. This is not a permanent map. Review it every six months.

What AI Does Better Now

• Pattern recognition at scale — Identifying trends, anomalies, and correlations across datasets too large for human review within practical time constraints
• Generating high-volume variations — Producing dozens of copy variants, headlines, subject lines, or visual briefs in minutes, enabling rapid testing
• Synthesizing large datasets — Summarizing, categorizing, and extracting key findings from research, reports, and competitive landscapes at a pace no human team can match
• 24/7 execution without fatigue — Running analyses, monitoring outputs, and executing repeatable processes at any hour without degradation in quality
• Cost-efficient first drafts — Producing serviceable initial versions of content, code, briefs, and reports that require editing rather than creation from scratch
• Non-obvious pattern outputs — The Move 37 capability: generating solutions outside conventional expert framing, sometimes outperforming accumulated human expertise

What Humans Do Irreplaceably

• Contextual judgment — Understanding the unstated, the politically sensitive, the emotionally charged context that surrounds every business decision and that never appears in a dataset
• Ethical reasoning in novel situations — Applying values, organizational principles, and moral reasoning to problems that have no precedent and no clear rule to invoke
• Relationship trust — The accumulated credibility, empathy, and reliability that earns a client's confidence or a team's loyalty — built over time and not transferable to a system
• Novel problem framing — Identifying the right question before attempting to answer it: the meta-skill that determines whether all subsequent AI work is pointed at the correct target
• Organizational navigation — Reading and working within the unwritten rules, power dynamics, and cultural norms that determine whether good work actually gets implemented
• Reading the room — Real-time adaptation to human emotion, group dynamics, and the unspoken signals in a meeting, presentation, or negotiation that change the right response instantly

Redesigning Roles Before the Moment Forces You To

The organizations that navigated the Move 37 inflection point best in 2024 and into 2025 share a structural pattern: they defined the new unit of work before they implemented the tools. They started not with the AI capabilities available, but with a forensic look at where their senior people's time was actually going — and how much of that time was pattern-based work that AI could absorb.

For marketing teams, that analysis typically reveals a striking imbalance. A significant share of hours spent by experienced strategists go to production-adjacent tasks: formatting deliverables, generating first-draft content, pulling competitive data, and building reporting structures. These are tasks that accumulated over years of team growth and never got formally reassigned. They are also precisely the tasks where AI is now most capable.

The redesign question is not "can we reduce the team size?" It is: "If we free our best people from these tasks, what would we do with that capacity?" Teams that answer that question specifically — not with vague language about "strategic focus" but with actual new deliverables and decision rights — are the ones that emerge from the transition with stronger output and higher morale. Teams that answer it vaguely are the ones where the AI tools create anxiety rather than momentum.

Three Principles for Role Redesign

Anchor roles to judgment, not production. Redefine job descriptions around the decisions a person needs to make, not the deliverables they need to produce. AI produces deliverables. Humans make the calls that determine whether those deliverables are right.

Make briefing a core skill. The quality of AI output is largely determined by the quality of the brief it receives. The ability to write a precise, contextually rich brief is now a differentiating professional skill — equivalent in importance to what presentation skills were a decade ago. Invest in developing it.

Build critical evaluation into the workflow, not the review. AI fluency is not just about using tools. It is about knowing when the output is wrong in ways that aren't immediately obvious. That skill requires domain expertise, and it is the reason experienced knowledge workers remain essential even as AI handles more of the production surface area.

What Lee Sedol Understood That Most Organizations Don't

When Lee Sedol retired from professional Go in 2019, he cited AlphaGo as a key factor. He did not say the game was ruined. He said the ceiling had been removed — that there was now an entity that could not be defeated, and that this changed what it meant to reach the top of the profession. He chose to step back from the competitive form of the game, but he did not stop playing. He continued to study, to teach, and to write about Go.

What he understood was that his value as a Go expert was not reducible to whether he could beat AlphaGo. It was located in what he brought to the human experience of the game: the ability to explain it, to make it meaningful, to connect it to something beyond a sequence of optimal moves. That is not a consolation prize. That is a genuine and irreplaceable form of expertise.

The knowledge workers thriving in 2025 are the ones who have made the same reorientation. They are not competing with AI on AI's terms — high-speed pattern processing, tireless output, encyclopedic synthesis. They are doing what only humans do: deciding what matters, for whom, and why. Holding other people accountable to that. Carrying the moral weight of decisions that affect real people.

The Move 37 Moment has arrived in knowledge work. The question is not whether it has happened. It is whether your organization redesigned the role around what happened next. At the operational level, autonomous AI agents are already executing multi-step workflows that previously required knowledge worker judgment at each step.

Open question: If AlphaGo had been available to Lee Sedol as a training partner from the beginning of his career — not as an opponent, but as a collaborator — would he have become a better player faster, or would he have developed differently in ways we cannot predict? The answer might determine whether the most important AI skill a knowledge worker can develop is not how to use AI, but how to decide when not to.

Frequently Asked Questions

What is the Move 37 Moment in knowledge work?

The Move 37 Moment in knowledge work is the inflection point where AI produces a non-obvious, creative output that surpasses what a human expert would have generated — named after AlphaGo's legendary Move 37 played against Lee Sedol in 2016. For knowledge workers, this moment describes the specific instant when an AI system produces an output that a skilled professional would not have reached independently, or would have taken significantly longer to produce. It signals not the end of expertise, but a fundamental shift in where human expertise should be applied. The framing matters because it is not about AI being generally smarter — it is about AI being capable of non-obvious outputs in specific, well-defined domains, which is a more precise and more actionable claim.

Which tasks is AI now better at than human experts?

AI currently outperforms human experts at several categories of knowledge work: pattern recognition across large datasets, generating high volumes of content variations at speed, synthesizing information from thousands of sources simultaneously, maintaining consistent execution around the clock without fatigue, producing cost-efficient first drafts across formats (copy, code, briefs, reports), and identifying non-obvious correlations in structured data. These are not trivial tasks — they represent a significant portion of what knowledge workers spend their time on today. The crucial distinction is between tasks that are pattern-based (where AI excels) and tasks that are judgment-based (where humans retain the advantage). That line is not fixed, and it is shifting toward AI as models improve and accumulate more domain-specific training data.

How should marketing teams adapt to AI capabilities in 2025?

Marketing teams should restructure workflows so that AI handles high-volume, pattern-based execution while humans focus on judgment-intensive decisions. This means repositioning roles around briefing quality, strategic framing, stakeholder communication, ethical oversight, and creative direction rather than content production volume. In practice: audit which tasks currently consume the most hours, identify which of those tasks are pattern-based versus judgment-based, and redesign the workflow accordingly. The goal is not to reduce headcount but to radically raise the ceiling on what a team can produce and decide. Teams that complete this redesign with specificity — actual new deliverables and decision rights, not vague "strategic focus" language — report stronger output and higher morale from the transition.

What is the irreplaceable human advantage over AI?

The irreplaceable human advantages over AI are contextual judgment (understanding the unstated, the political, and the emotionally charged), ethical reasoning applied to novel situations, relationship trust that is built over time and cannot be replicated by a system, novel problem framing (identifying the right question before answering it), navigating organizational dynamics and unwritten rules, and reading the room in real-time human interaction. These capabilities are not just hard for AI to replicate — they are foundational to why organizations function at all. A company is not a data-processing system. It is a human institution that runs on trust, authority, culture, and judgment. AI can serve that institution powerfully. It cannot be that institution.