AI eliminated execution. Now, taste is the only asset left.

There is a phrase that the CEO of a global holding company said to Forrester, without giving his name. The phrase is this: “By 2028, we will double profits and cut people in half.”

He was not making a prediction. He was describing a plan. And the plan has already begun.

Forrester had a 2022 projection saying that 7.5% of agency jobs would be automated by 2030. Thirty-two thousand positions in eight years. It seemed manageable. Something the market would absorb without noticing, the way it absorbed layouts made in Canva, copywriters replaced by templates, photographers replaced by stock image banks. A slow erosion.

Except that at the end of 2025 Forrester threw out the old projection. The new one: 15% of jobs eliminated in 2026 alone. Not by 2030. In twelve months. Forty-seven thousand positions. And that after an 8% cut that had already happened in 2025.

This is not a cyclical crisis. It is a change of model.

The execution layer is being eliminated

I could start this piece by talking about artificial intelligence as if it were something distant, a market trend, a presentation slide with pretty icons and arrows pointing up. But that is not how this story works for those who are inside it. For those who work with brand, with communication, with creation, the thing is simpler and more brutal: the creative execution layer is being eliminated. Not threatened. Eliminated.

Bloomberry analyzed 180 million job postings and found what it called a “bifurcation”: creative execution roles, copywriters, photographers, illustrators, are in secular decline. Not seasonal. Secular. Meanwhile, creative direction, strategy and product design roles keep stable demand. Creative work is splitting in two, and the cut line is exactly where AI can operate: in volume production.

Forrester calls this “workforce inversion.” The old model worked like this: expensive creative directors at the top supervising a large layer of cheap juniors who did the grunt work. The spread between the cost of the junior and the value charged to the client was the profit. AI inverted that equation. You no longer need the junior layer. The new model is a small team of senior people working directly with AI tools. The juniors who did the volume disappear. And the middle managers who supervised the juniors disappear along with them.

Gartner predicts that, by 2026, 20% of organizations will use AI to eliminate more than half of middle management roles. Fast Company documented how Amazon, Dell, Microsoft and Google are already flattening their structures. Al Jazeera reported that companies are using AI agents not only to complete tasks, but to assign tasks to humans, monitor performance and recommend promotions or dismissals. The boss of the future may not be human. But that is not the point.

The point is something else.

95% of AI agents are fake

While companies are eliminating the creative execution layer, they are simultaneously buying, at a premium price, tools that promise to replace that layer. The problem is that most of these tools do not work.

Gartner analyzed more than 3,000 vendors selling “AI agents” and found that only 130, about 4%, deliver real agentic capability. The other 95% are chatbots and old automations rebranded with a new name and a higher price. This got a name: agent washing. It is the greenwashing of artificial intelligence. You take an automation flow that already existed, put “agent” in the name, and charge 10 to 50 times more.

Deloitte was more direct. It said many AI agent implementations are creating what it called “workslop,” poorly designed agents that add work to the process instead of eliminating it. The projection is that more than 40% of agentic AI projects will be canceled by 2027. Not for lack of technology. For lack of clarity about what the technology really does versus what marketing says it does.

Slack, inside Salesforce, predicted that 2026 would be “the year of the lonely agent”: companies will create hundreds of agents per employee, and most will sit idle, like software licenses never used. Impressive, but invisible.

So we have a scenario: companies fired the people who did the work, bought fake tools to replace the people who did the work, and are now producing content at industrial volume with generic quality. And the consumer has already noticed.

The consumer developed antibodies

Gartner’s March 2026 research is the one that should be on the wall of every marketing director on the planet. Half of consumers, 50%, prefer brands that avoid using generative AI in public-facing content. Sixty-eight percent frequently question whether the content they see is real. Sixty-one percent say they frequently doubt the reliability of the information they use to make decisions.

SmythOS reported that 52% of consumers reduce engagement with content they believe is AI-generated. It is not a subtle preference. It is an active rejection. And when brands try to create emotional connection using AI, the effect is worse: research published in the Journal of Business Research documented that AI-generated emotional communication provokes what they called “moral disgust” in the consumer. Not indifference. Disgust. This reduces positive word of mouth and erodes brand loyalty.

Klaviyo surveyed more than 8,000 global consumers and found that even among AI enthusiasts, those who use and trust it, 60% think brands are relying too much on the technology. And 33% encounter low-quality content, the so-called “slop,” several times a week. Trust is not stable. It degrades.

Advertising Week published a panel with branding directors from several global consultancies. The conclusion was unanimous: when everyone uses the same models trained on the same data, distinction evaporates. One of them, from Saffron Brand Consultants, said the risk is not that AI-generated content looks robotic, because often it does not. The risk is that brands start to sound the same. Another, from MarketingProfs, was more precise: generic content does not just perform poorly. It costs money and destroys brand equity.

Adweek, in its March 2026 issue, closed the reasoning: content volume, speed and variation are becoming almost free. That means “good enough” creative collapses in value. What became scarce is taste. Direction. Restraint. Cultural relevance. And the ability to create something that does not look like it came out of the same statistical blender as everything else.

The salesperson now has a sponsor

And then, in February 2026, OpenAI put ads inside ChatGPT.

It is not a technical detail. It is a breach of contract. ChatGPT was built on an implicit promise: I give you objective answers, you trust me. When the answer starts coming alongside a paid ad, contextualized by your conversation, the question changes. It is no longer “what does the AI recommend to me?” It is “who paid the AI to recommend this to me?”

Adweek warned: marketing will be the function held accountable when consumers ask what is organic, what is paid, and whether the AI is serving their interests or the brand’s. Rithum surveyed 1,046 consumers and found that 58% blame the brand when an AI recommendation contains wrong information. They do not blame the AI. They blame the brand. And 64% of consumers between 18 and 27 years old buy based on AI recommendation without checking anywhere else.

AI became the salesperson. But no one asked whether the salesperson is trustworthy. And now the salesperson has a sponsor.

Google DeepMind, by contrast, declared at Davos that it has no plans for ads in Gemini, framing advertising in AI systems as a trust risk. Anthropic made a joke about OpenAI’s decision during the Super Bowl commercials. These are strategic positions, of course. But the fact that “having no ads” became a competitive differentiator for an AI says something about the moment we are in.

The two wrong poles

There is something I want to say clearly, because the dominant discourse swings between two equally useless poles.

The first pole is blind optimism: AI will solve everything, everyone will be an “agent director,” the future is one of creative abundance. That is the talk of someone who sells tools, not of someone who does the work.

The second pole is nostalgic pessimism: we need to go back to the “human touch,” reject AI, embrace imperfection as resistance. Adweek demolished this pole in one sentence: brands that lean on humanity as a deliberate contrast to synthetic sameness are also getting it wrong, because they will not take advantage of the AI capabilities that could benefit their teams.

Both poles are wrong because they are debating the tool instead of debating what the tool revealed.

What AI revealed is simple: most of the brand content that existed before AI was already generic. It was already produced in volume, optimized for short-term metrics, emptied of point of view. AI only accelerated the production of the mediocre to a point where mediocrity became impossible to ignore. The “sea of sameness” was not created by AI. It just raised the water level.

And this, paradoxically, is the best news a creative director could receive.

Production became a commodity, curation became the asset

Forrester says originality is the most significant factor that reduces a role’s automation potential. Bloomberry confirms: roles that involve complex decision-making, client interaction, user research and strategic thinking are holding up. What is falling are the repetitive execution roles.

HBR published in February 2026 an article about the emergence of a new role: the “agent manager.” Fortune documented the same trend with different language: creative workers will not be replaced by AI, they will become “directors” who manage AI. The question is no longer “how do I do this?” but “what do I want done, and how do I delegate it?”

Vercel CEO Guillermo Rauch told Business Insider that the manager of 2026 is a manager of agents. Not a people manager who uses AI as a tool. An orchestrator of autonomous systems. And he added: companies that treat AI agents as tools you set and forget will get burned. The ones that invest in the human layer, in the agent managers who understand both the capabilities and the failure modes, will win.

This is the complete inversion. In the old model, value was in the ability to produce. In the new model, value is in the ability to decide what is worth producing. Production became a commodity. Curation became the asset.

The way out is editorial

I have been writing about brand and culture for years, and I have never seen a moment when so much data converged on the same conclusion: those who have a point of view will thrive. Not those who have more tools, more automations, more agents, more volume. Those who have editorial vision. Those who know what not to do. Those who understand that restraint is a form of power.

The market is accelerating in the wrong direction. It is investing in production speed at the exact moment the consumer is developing antibodies against synthetic content. Companies bought fake automation to produce content the public already identifies and rejects, while firing exactly the professionals who could fix the problem.

It is a self-destructive loop. And the way out is not technological. It is editorial.

The way out is to have something to say.

Sources cited: Forrester Predictions 2026: Marketing Agencies. Gartner Marketing Survey, March 2026. Gartner Agentic AI Analysis 2025-2026. Bloomberry, analysis of 180 million job postings. Deloitte Insights, Agentic AI Strategy. Adweek, March 2026, 10 AI Marketing Trends. Advertising Week, Why Branding Still Needs a Human Touch in 2026. SmythOS, The AI Content Trust Gap. Klaviyo, AI Consumer Personas Playbook. Rithum Consumer Survey 2026. MarketingProfs, Elastic Marketing. Harvard Business Review, February 2026, To Thrive in the AI Era Companies Need Agent Managers. Fortune, Brainstorm AI 2025. Business Insider, Vercel CEO. OpenAI, Testing Ads in ChatGPT, February 2026. TechCrunch, ChatGPT Rolls Out Ads. Globant, Adobe Summit 2026 BXOS. NIM and Journal of Business Research, AI-Authorship Effect. Fast Company, AI and Middle Management. Al Jazeera, AI Agents Reshaping Hierarchies. BCG, Global Consumer Radar 2025.