Human Input: MAICON 2025's 'Move 37' moment that marketers can’t ignore

When a single move changes the game

Nine years ago, an AI stunned the world of Go with what became known as “Move 37,” a play so unconventional that commentators fell silent, world champion Lee Sedol left the room, and everyone watching knew the ground had shifted. That was AlphaGo’s second game in a five-game series it would eventually win 4-1. The phrase “Move 37” has since become shorthand for a moment when technology forces us to rethink what’s possible and what our role becomes next.

Earlier this month at the Marketing AI Institute's annual MAICON event in Cleveland, Paul Roetzer, founder and CEO of the Institute and SmarterX, put that moment front and center for marketers. As he framed it from the stage: Move 37 is “the moment when you realize AI is better than you at the thing that you do.” It may start as tasks, he warned, but “those tasks will add up and it will start to affect entire jobs.”

I felt a different energy in Cleveland compared to last year: less hype, more straight talk about work, capability, and responsibility. Attendance had jumped signficantly to roughly 1,500, and the hallways were packed with teams comparing notes on their own Move 37 moments. These are the unsettling and clarifying times when the old playbook breaks.

Context: what changed in a year

If you simply walked the expo you could still see two big shifts: First, the models have moved fast. Voice and video systems made the biggest leaps. Demos looked like production tools instead of science fair projects. Second, operator comfort surged. Many more people were building custom GPTs, chaining tools, and automating real workflows, not just tinkering.

In a video interview, Boulder's Kerry Fitzmaurice distilled a pattern I heard everywhere: We are entering B2A and A2A territory, business-to-agent and agent-to-agent. “Eventually our agents are gonna go find something for their human. They are gonna talk to another agent, get the information, and then bring it back or already have the criteria embedded to make the purchase.” That reframes who our content must persuade – and how

Jeremiah Owyang, venture capital investor at Blitzscaling Ventures, dove deeply into this during his keynote on Wednesday, The Future of AI Marketing, A Silicon Valley Insider’s Perspective. “This is the new business model… The future of marketing is when your customer isn’t a human… it may be their agent that’s shopping for them.” His thesis mirrors a broader “agent-to-agent” push I’m seeing in enterprise circles. You can read more about his thoughts on this trend in a recent Workday blog.

The punchline for marketers is uncomfortable but actionable: Content is decoupling from presentation. Agents will ingest, reason over, and reassemble our material for end users and for other agents regardless of our beautiful landing pages. Owyang again: “You will also launch an API that only talks to the agents… decouple your content so it goes where the consumer goes, and that’s in front of the agent.”

Colorado spotlight. Liza Adams, founder and GTM strategist at GrowthPath Partners, brought a people first lens to “virtual employees,” showing how AI teammates slot into real teams without replacing human judgment. Her model is simple. Humans set goals, provide context, and make calls. AI contributes memory, speed, and tireless execution. In a standout case study, she walked through a large enterprise running more than 200 custom GPTs that teams chain for launches and campaigns, supported by a lightweight directory GPT that suggests which teammates to use and who owns them.

The takeaway was not bigger models, but how approachable this has become when you define clear roles, steps, and handoffs, and how quickly value shows up when the culture treats AI as teammates that extend people rather than shortcuts that replace them.

Video star. On Thursday morning, Filmmaker PJ Ace shifted the focus to AI native video production in his keynote, The Rise of the AI Filmmaker: From Concept to 250M Views in 60 Days. His team can compete with million-dollar commercials for just a fraction of the cost, without location shoots, and on week level timelines. The economy of advertising is changing in front of us.

One notable omission across the week: China. As Christopher Penn noted post-MAICON, there was “a biggest thing missing” conversation about geopolitical and competitive dynamics. Whatever your view, ignoring China’s AI trajectory is a strategic mistake for global marketers and U.S. policy thinkers alike.

What I saw and what it means

1 - The Move 37 mindset is now a management task

Roetzer’s keynote was a reminder that model improvements are compounding, and that many knowledge roles will feel “AI better than me” moments on a rolling basis. Leaders need to plan for cumulative task displacement, not just tool adoption. The practical move is to maintain a list of high volume tasks, retest them quarterly against current models, and decide what is human led, AI led, or agent led. Tie those decisions to outcomes like cycle time, quality, and cost.

2 - Roetzer’s reality check on jobs, value, and time

Roetzer put hard numbers behind the anxiety. There are roughly 136 million full time employees in the United States, and estimates suggest 70 to 100 million of them think and create for a living. That is the target surface for AI native builders. Pulling May data from the Bureau of Labor Statistics, he highlighted about 11 trillion dollars in annual wages across occupations. Then he showed why wage pools, not software categories, are drawing capital. Sales and related occupations total 13.3 million jobs with 722 billion dollars in wages. Customer service representatives account for 2.7 million jobs and 123 billion dollars. Market research and marketing specialists represent 860,000 roles and 74 billion dollars. Media and communications add more again.

His point was blunt. The jobs with the largest wage spend are being targeted first. If you think no one would try to automate entire job families, Roetzer pointed to Y Combinator’s recent call for vertical AI agents and its premise of automating work outright. Pitch decks that map a wage pool, claim a percentage can be automated, and outline a path to product are already getting attention.

That is the capital incentive behind what he called an economic Turing test. Before each hire or backfill, leaders will ask a simple question: Should I hire an agent or a set of agents instead of a human?

Alex Kantrowitz, founder of Big Technology and host of the Big Technology Podcast, followed with Backstage with Big Tech: AI Truths from the Frontlines of Innovation, offering the counterweight to Roetzer’s wage math with a case for AI literacy as the path to career longevity and new opportunity. He argued the next phase is less about ever bigger models and more about orchestration and reasoning, which raises the premium on literacy for individuals and teams. In his words, we are entering the age of the individual, but only for those who want it. If you live in the tools and learn how to apply them, you can do more than ever before. There is no single right playbook, so get your hands dirty.

The synthesis is not doom, but a fork. Some firms will race to automate entire functions. Others will pair automation with redesign and retraining so people move to higher value work. The people who lean in will develop real leverage and outperform. Start the shift now, because next October will feel as different from today as today felt from last year.

3 - The cost curve of video has collapsed

PJ Ace's case studies show what AI native production can already do. Concept to launch falls from weeks and tens of thousands of dollars to just days and hundreds or low thousands. That does not mean spray and pray with synthetic content. It means creative teams can run more controlled experiments, with higher learning velocity and better brand control, because the cost of iteration has crashed.

4 - Agent ecosystems are crossing from slides to systems

Agent talk was everywhere, from consumer primary agents to enterprise workflow agents and early agent crews that connect to thousands of apps. The near term pattern is clear: Manual becomes hybrid, and then automated. Agents educate buyers, compare products, and transact. Often they do this without a human visit to your site. That does not kill narrative craft, but it changes the audience. You are persuading humans and informing software that negotiates on their behalf.

5 - The community is more practiced and more open

Cathy McPhillips, CMO at the Institute, put numbers to what I felt in the sessions. Last year saw about 1,100 people. This year there were about 1,500. Attendees came with more stories and a stronger grasp of how AI applies to their businesses.

Andy Jolls, a veteran CMO and first time attendee, echoed that sentiment. Agents were very popular, but it still feels like this is the early innings, with more workflow wins than full autonomy. That is a useful planning signal. Design for supervision and handoffs now, and don't wait for perfect automation.

A quick primer on Move 37

For readers who were not glued to the 2016 match, AlphaGo’s Move 37 in Game Two was a shoulder hit so unexpected that pros called it creative and unique, and Lee Sedol spent more than twelve minutes responding. AlphaGo won the game and the series, while Lee struck back with his own improbable “Move 78” in Game Four, sometimes called the God move, that briefly exposed AlphaGo’s weakness.

The lesson is not who is the hero or villain. The lesson is the shock of the new, followed by human recovery and adaptation. That is the rhythm of the next few years in marketing.

Roetzer’s challenge to the attendees was simple. Every knowledge worker will have a Move 37 moment, but they won't all happen at once. The leaders who navigate it best will pair pragmatic adoption with ethical guardrails and talent development. Replacement pressure is real, and so is the chance to redesign work around what humans do best.

Top 8 takeaways from MAICON 2025

1 - Custom GPT literacy is the new baseline. This felt like the year “I am experimenting” turned into “I am shipping.” Know how to scope, prompt, and govern custom GPTs, and how to chain them into simple workflows. You do not need to be an engineer, and you do need to think like a product owner. Define inputs, guardrails, and success criteria.

2 - Learn automation so you can lead agents. Agent builders will absorb many no-code automations. Learn them anyway. Triggers, actions, data mappings, and exception handling teach the mental model you will use to describe work to agent platforms. If you can diagram a current Zap or Make flow, you can explain an agent’s step-by-step policy.

3 - Write the play, not just the prompt. Agents behave better when you provide a playbook, including roles, objectives, constraints, escalation rules, and stop conditions. Think in clear and auditable steps that a trained teammate would follow. That clarity lets an agent builder – or your future self – implement and supervise it. Vague goals produce vague results.

4 - Get video literate, fast. AI accelerated production has collapsed cost and time. Learn the language of treatments, boards, shots, rights, and disclosures so you can move at the new speed without sacrificing standards. Video is becoming the default medium for testing ideas, and not only for telling stories. Pair speed with clear safety and disclosure policies. Put authenticity and trust at the front of the process.

5 - Trust is the product. The strongest moat will not be model access, it will be credibility. Treat synthetic and agent assisted media with the same care you apply to financial claims. Disclose methods, watermark when appropriate, and set internal no-go lines before the first render. Marketers are ambassadors of trust, and if you break that trust by playing carelessly with new toys, you may not get it back.

6 - Ethics means agency, not just compliance. Move the conversation from “What is allowed?” to “What is the impact on people and outcomes?” Be explicit about where AI speeds humans, where it replaces tasks, and what retraining or role redesign will follow. We are also ambassadors of the brand’s values. Your culture will remember whether you used transformation to invest in people – or only to extract from them.

7 - Prepare for human-to-agent-to-agent buyers. Your next customer may arrive as an agent. Structure facts, policies, and inventory so software can verify and act on them. Publish machine readable truth behind stable endpoints. Narrative craft still matters, but it now shares the stage with content that agents can parse and trust.

8 - Stay a student. Models, video, and agents will not slow down for your calendar. Choose a cadence that fits your team. Monthly skill sprints, quarterly capability audits, and short post mortems work well. Keep updating your personal playbook. The marketers who thrive will be the most adaptable and the most transparent.

Closing reflection: transform on purpose

MAICON’s 2025 theme, Transform, landed because it was not about magic buttons, it was about choices. As Kerry Fitzmaurice said, you do not have to reinvent your tech stack to see real gains. Start with easier applications that return ten times the output without a complete rebuild. That is the door in. Walk through it, then do the harder work. Clarify your agent strategy. Build your Move 37 playbook. Set expectations with your team about how work will change.

I am curious how Roetzer’s forecasts track over the next year. My bet is that by MAICON 2026, human-to-agent-to-agent will feel like table stakes for how marketing gets done.