Agentic Commerce: How Human Desire Beats Bot Compliance

As artificial intelligence automates how we shop, the most durable competitive advantage won't be data or efficiency. It will be desire.

Your personal AI is about to restock your fridge. It will negotiate the price of milk with a grocery bot while you sleep. It will find the best-reviewed, most eco-friendly, cheapest laundry detergent on the market. This is the promise of ‘agentic commerce’: a world of frictionless, optimized shopping. And for brands, it’s a potential nightmare

The new mandate for marketers is to prepare for this world. Structure your product data. Expose your APIs. Make yourself “agent-friendly.” The race for “GEO,” or Generative Engine Optimization - essentially SEO 2.0 - is on, and everyone wants to be the first result on the algorithmic shelf.

But this rush to appease the bots ignores a fundamental question: What happens when the messy, inefficient, and deeply human process of choosing is automated away?

The Frictions That Create Desire

Let’s be honest. Not all shopping needs to be a meaningful journey.

Nobody wants to “discover” toilet paper on a Tuesday night. When you’re out of laundry detergent, efficiency is the value proposition. Your Amazon “subscribe and save” purchases are wonderful plumbing.

But agents treat everything like a feature-comparison problem. They don’t distinguish between toothpaste and a winter coat, between soap and a wedding dress.

Think about your last important purchase, whether a piece of furniture, headphones, some clothes for a job interview. Did you efficiently compare specifications and execute the optimal choice? Or did you browse, watch videos, read contradictory reviews, ask friends, change your mind, and finally buy something different than you started looking for?

This process, pretty inefficient, very human, is how we build our taste and preferences. It’s how we learn what we actually want, not just what we think we need. You went looking for running shoes and discovered you care about it being recyclable. Or that minimal design appeals to you. Or that you’re willing to pay more for a brand that sponsors athletes you admire.

Shopping, at its best, is a discovery process. AI agents, in their quest for optimization, threaten to compress that journey directly to checkout. Clean, fast, and hollow.

The Limits of the Curation Engine

But what if the agents aren’t so simple? Maybe they will evolve into sophisticated “taste-makers.” An agent with access to your streaming history, social media, and personal calendar could learn your aesthetic and values. It wouldn’t just find “running shoes”; it would find the specific, sustainably-made shoe from an indie brand that aligns with your identity.

In this vision, the agent doesn’t kill discovery; it elevates it into hyper-personalized curation. However, this optimistic view runs into three fundamental walls.

First, it assumes a level of data sharing that is far from guaranteed. A true taste-making AI requires near-total access to your digital life. Without your complete, uncensored history of thoughts, desires, and behaviors, its promise of perfect curation falls apart. In an age of increasing privacy concerns, the prerequisite for this “smart” agent (our total transparency) is its biggest vulnerability.

Second, even with perfect data, it creates a prison of the past. A taste-making agent works by analyzing your history to predict your future. It can give you more of what you already like, reinforcing your current identity. But true discovery often comes from serendipity : the happy accident that introduces you to something you never knew you could love. The agent creates a perfect filter bubble of one, a beautifully decorated prison for your taste, preventing the random friction that helps you evolve.

Third, it automates meaning. Meaning isn’t a destination; it’s a byproduct of effort. The satisfaction of finding the perfect item is tied to the work you put in, the research, the deliberation, the hunt. We value what we struggle for. By making discovery effortless, the agent devalues the outcome. It gives you the answer without letting you wrestle with the question, stripping the process of its power.

The Commodification Engine

So whether the agent is simple or sophisticated, it pushes brands towards the same economic black hole: perfect competition.

Marketers are being told to provide structured data, real-time pricing APIs, and detailed specifications. In doing so, you are inviting your own commodification. Perfect competition exists when buyers have complete information and products are perfectly comparable. In such a market, prices converge toward marginal cost, and nobody makes money.

“Agent-friendly” commerce removes the buffers that protected you. When your product is reduced to a row in a comparison matrix such as price, features, ESG score, delivery time, you’re no longer a brand. You’re a data point. And data points compete on price.

Consider what happened with search engine optimization. Everyone optimizing for the same keywords. Very costly and hardly a winning playbook. Just a tax to be findable. Now imagine that across every product attribute. Generic “running shoes” fed into an agent returns 50 options ranked by algorithmic score. You might win the transaction by having the best price-to-feature ratio. But you’ve won a race to the bottom.

This is the central trap. You don’t want to be just in the agent’s comparison matrix. You want to be the specific thing a person asks for by name before the agent even starts its search.

The Involvement Ladder is a Matter of Context

Every purchase sits somewhere on what we might call an “Involvement Ladder”—from low (automatic, efficiency-driven) to high (emotional, identity-driven). The default setting of an AI shopping agent is to yank every product down to the bottom rung, applying cold logic of feature comparison to everything.

But crucially, a product’s position on this ladder isn’t fixed; it’s determined by context and, more importantly, by brand.

Toilet paper is a low-involvement restock. But if you’re hosting a dinner party to impress, suddenly you may be buying the luxury, quilted, lotion-infused version. The context turned it into a high-involvement choice.

Jewelry seems like a classic high-involvement purchase. But for an ultra-wealthy individual buying a simple gift, it might be a quick, low-involvement task delegated to an assistant or an agent.

Coffee can be a low-involvement caffeine delivery system. Or it can be a high-involvement ritual of craft, community, and lifestyle. The difference between an espresso at next coffee machine and beans from a roaster you follow on Instagram.

The new battle for marketers is to win the context and move your brand up the ladder through meaning, not features.

The Great Paradox: More Automation Requires More Humanity

This brings us to the paradox at the heart of agentic commerce. There is this idea that AI would make marketing more technical, more data-driven, more algorithmic.

The opposite is true.

When the transaction layer becomes automated, the only durable competitive advantage is winning the preference before the agent is even activated. In other words, branding.

Agents can execute purchases. They cannot create desire.

There are two distinct layers:

Preference Formation (Pre-Agent Space): This is where brands win or lose. It’s cultural, emotional, and identity-driven. It’s the layer where someone decides “I want Adidas” before they ever ask their agent to buy shoes.

Transaction Execution (Agent Space): This is where agents operate. It’s important as you need to be findable and functional. But if you’re only competing here, you’re competing in the commodification matrix.

Look at Liquid Death. They took water—the ultimate commodity—and made it identity-driven through irreverent branding and cultural positioning. Now people ask for Liquid Death specifically. The product is water. The brand is meaning.

Or consider Patagonia. An agent comparing technical specifications would find a dozen jackets with similar warmth-to-weight ratios at lower prices. But Patagonia customers are buying into environmental values, durability as philosophy, and a community of like-minded people, not specs. The brand exists in the pre-agent space, built through storytelling, activism, and shared identity.

The more automated commerce becomes, the more brands will need to invest in the least automatable thing: human connection and meaning. To compete in the age of algorithms, brands must become more human, not less.

The New Brand Roadmap: Dual-Track Strategy

Now brands are competing on two tracks simultaneously. And can’t afford to ignore either one.

Track 1: Agent Optimization (Table Stakes)

This is about not losing. You must be shoppable by agents:

• Structure your product data properly

• Provide real-time inventory and pricing

• Ensure your APIs work flawlessly

• Make checkout frictionless

If you fail here, you’re invisible. The agent won’t even consider you. But being perfect on Track 1 doesn’t make you win. It just keeps you in the game.

Track 2: Desire Creation (The Only Moat)

This is where you actually win. You must be specifically requested by name:

• Build cultural relevance through creators and influencers

• Invest in community and belonging

• Create experiences that generate attachment

• Tell stories that make your brand mean something

• Cultivate taste-makers who advocate for you

Track 1 determines whether the agent can find you.

Track 2 determines whether anyone cares to ask for you specifically.

If your product is genuinely low-involvement and you cannot elevate it, accept it. Optimize everything for agents. Automate your supply chain. Compete on price and distribution. Prepare for thin margins. This is a viable strategy—if you go all in and run the tightest ship possible. But be honest: if you’re a commodity, you’re in a race to the bottom. The winner will be whoever can operate most efficiently at the thinnest margins. Can you be that company?

The War Ahead

The death of shopping isn’t the death of marketing. On the contrary.

Agents will automate transactions. But they cannot automate desire. They cannot manufacture meaning. They cannot create the cultural context that makes someone want a specific brand.

The irony is perfect. Structured data won’t save you. Perfect APIs won’t save you. Being “agent-friendly” alone won’t save you.

Being wanted will.

When someone cares enough to say “get me those Adidas.” instead of “find me running shoes.”

That one word, your brand, is the only moat that matters.

Everything else is just plumbing.

Optimize for agents to stay visible. Build meaning for humans to stay valuable.