Why AI Agents Can't Own Serendipity

I set out one evening to buy a blue blazer. The cut was clear in my head. The fabric, the shade of blue (not too light, not too dark), the designers I'd consider, the budget. Everything was decided.

Then a pink skirt showed up on my screen.

It was styled with an outfit I would never have assembled. A grey linen jacket that looked like the twin of one already hanging in my closet sat next to it. The color was wrong for me. Or what I thought was wrong for me.

I wanted the skirt more than the blazer I had come for. The budget held room for one. The skirt won. The blue blazer remains unbought.

That happened because something in my brain broke past what I had planned. The want did not exist until the instant I saw the skirt. The seeing was the purchase. It created the demand I then satisfied with my purchase.

Now pit that experience against the agentic commerce system everyone is chasing.

The promise of the shopping agent is clean. You say what you need. The agent finds it. It never wastes your time on things you did not ask for. The agent, handed my instruction, would have found blazers. It would have compared three of them, checked my size, my loyalty points, the available promo codes, recommended the best option and closed the sale. It would have done its job perfectly and sent me home with a blazer I would have bought if the pink skirt had never existed.

That is the feature. It is also the blind spot.

I have written before that AI agents are becoming the department stores of the future. The competition in commerce is shifting from who has the cleanest interface to who owns the intent. What we have today is, under the conversational veneer, an very smart search bar. Better than Google. Able to strip friction out of the path to a purchase. All of that holds.

Search is the path to the thing you already know you want. It is not the tool for changing what you want. Changing what you want has always been a big part of what shopping is.

An agent can introduce you to a new brand of blue blazer. It can surface a merchant you never considered. What it cannot do is show you the pink skirt that was never in the consideration set at all.

The implication baked into agent architecture is that intent equals all of demand. It does not.

Intent is the part you can already name. Feed an agent a decade of my purchases and it learns that I reach for French blue and never coral. It knows I buy the knee-length cut and skip the mini. That profile is real. The agents will get extraordinarily good at it.

That profile, no matter how complete, is a record of what I have already wanted. It is not the same as the want that does not yet exist.

Robert Merton, the sociologist who gave us the concept of serendipity, said it takes two things at once. A chance encounter. A mind prepared to seize it. Not luck alone. Luck plus a person who sees the unsought thing and grasps its value in the same instant.

My mind was prepared for the skirt. In the half-second I saw it, I knew the color worked. I knew it matched a jacket already in my closet. Someone else scrolls past the same image and feels nothing. Same encounter. No prepared mind. No serendipity.

The encounter was unplanned. The prepared mind was mine. The agent, by design, removes the encounter and stands in for the mind. It kills both halves of Merton's formula at once.

There is no chance in an agentic world because chance is inefficiency. There is no prepared mind in the loop because the whole point of the agent is that you do not have to bring one.

Some will say this is fixable. Better data. Learn the customer well enough and the agent surfaces the skirt too. That misses the nature of the problem.

We instruct an agent in language. Typed or spoken. The want for the skirt did not arrive as language. It arrived through my eyes, before any words, in the half-second before I had even been introduced to the idea of a pink skirt. Sight formed the want. By the time there were words to type, the moment that created it was already gone.

The psychologist Robert Zajonc made the case in 1980 that feeling comes before thinking. The pull toward something hits first. The reasoning arrives after. It happens fast. A 2014 study from Mary Potter's lab at MIT found the brain can recognize an image flashed for 13 milliseconds. Far too quick for any second thought.

By the time I could have turned the skirt into the words "pink skirt," my brain had already seen it, already wanted it, already finished wanting it. The words come last.

Agents can see well. Hand a modern model a photo and it reads it back in detail. A pink midi. Grey jacket. Here are six like it. The model turned the picture into words. That is the only thing it can do with an image. Describe it. Reason over the description.

It is not what the skirt did to me. A model can read ten thousand skirts and build a finer description of each. It never does the wanting. It can see the skirt in perfect detail and want none of it. The whole design of the agent is that I no longer have to be the one looking. That means I am not there for the skirt to happen to.

TikTok proves the point. A majority of TikTok users say they have made an impulse purchase on the app, according to a Bizrate Insights 2024 survey. Among TikTok Shop users, the share who report buying something they discovered there runs past 70%. They did not search for it. They saw it. A creator held it up. It was styled. It moved. The want formed on contact.

Social commerce runs on the opposite principle from the agent. In 2023, Americans spent an estimated $71 billion on purchases driven by social media. Things they never set out to buy. Things they saw their way into wanting.

The feed shows you things. The agent fetches what you named. Both are digital. Only one of them puts a person in front of the thing while the person is still looking.

Consider a different case. You tell the agent to find terracotta pots for your garden. The agent returns pots in assorted sizes and prices. Along with them come recommendations on potting soil, garden gloves, slow-release fertilizer. Every one of those recommendations is directly related to the thing you asked for. Pots imply soil and fertilizer. Planting implies dirty hands, which imply gloves. The agent did not invent a want. It unpacked the wants already riding alongside the one you stated.

Let us call that adjacency.

The pink skirt was not adjacent to anything. The gloves were reachable from the pots by a chain of need, and the agent walked that chain in milliseconds. There was no chain from a blue blazer to a pink skirt. The skirt did not serve the blazer. It did not complete the outfit. It did not follow from the prompt that drove my search. No amount of reasoning over my stated intent arrives at it.

Adjacency extends what you already wanted. Serendipity replaces it. The agent is built to do the first. It structurally cannot do the second.

I organized this into a framework. Two forces drive commerce. The first is intent. How clearly you already know what you want. The second is influence. How much being shown something formed the want.

On two axes, four quadrants emerge.

High intent, low influence. You know what you want and no one had to show you. Garden hoses. The same moisturizer you have bought for six years. This is the agent's home ground. It does this faster and better than you ever could. Frictionless is a pure gift here. Agents rule.

High intent, high influence. You can name the want now, you could not have a month ago. Something put it there. A friend's air fryer. A set of reviews. By the time it reaches the agent it has hardened into a product name. The agent looks like it served your intent. It satisfied a want that something else already built.

Low intent, high influence. The pink skirt. You were not shopping for it. You would never have typed it. The want was made whole in the instant of seeing. This is the corner the agent cannot enter. The agent honors what you asked for and strips out everything you did not. This corner is made of nothing but the things you did not ask for.

Low intent, low influence. No goal. Nothing has grabbed you. The aimless wander where a lot of discovery used to start. The agent has nothing to do here. So in an agentic flow it barely happens. It is the browsing the agent mutes just by existing.

The agent lives in the column where influence is low and intent does the work. The skirt lives alone in the opposite corner. All influence. No intent.

One version of the story ends there. Serendipity cannot live in the agent layer. It cannot live in the machine-to-machine layer being built now, where two agents reconcile a stated intent against a catalog and there is no eye in the loop. Nothing that sees and then wants. Serendipity survives only where a human eye is still doing the looking.

The skirt did not materialize from thin air. Someone made it in that color. A buyer chose to stock it. A merchandiser styled it with that grey jacket and not some other one. None of that was an accident. Stores have always planned inventory to capture this exact spend. They read styles, trends, weather, the shape of who their customer is. They place bets on the unsought item. The one nobody walked in for that a certain kind of person walks out with.

Serendipity at the level of the rack is manufactured. It always has been.

What the store does not have is the context. It does not know that this woman, on this night, was looking for a blue blazer with budget for one thing. That the jacket styled with the skirt was the cousin of something already in her closet, which is why the outfit caught and held her.

The aggregate bet on the pink skirt is theirs. The specific moment I saw it and bought it was mine.

Serendipity has always lived in the gap between those two. Between the want a system can predict across thousands of people and the want it can trigger in one of them, here, now, at the instant of seeing.

That gap is the question facing agentic commerce. It is a question of data and timing, not luck. A system that held enough of me, my eye, my closet, the colors I reach for and the ones I never would, the budget I set. It could in principle close the distance the store or site cannot. It could put the unsought thing in front of me at the moment I was ready to be taken by it.

Louis Pasteur said it best in 1854. In the fields of observation, chance favors the prepared mind.

The store or the site supplies the encounter. What it cannot supply is the prepared mind. A system that knew me well enough could.

The question is not whether it can be built. It can. The question is who would build it.

The answer may not be the people racing to win agentic commerce right now. Not the shopping agents. Not the players sprinting toward frictionless. The thing they would have to restore, the unrequested visual encounter aimed at a specific prepared eye, is the exact friction their whole business exists to remove.

They are optimizing the column where intent does the work. The skirt lives in the corner where intent is absent and being seen is everything. You do not reach that corner by getting better and better at returning only what was asked.

You reach it by putting something unasked-for in front of a human eye. At the moment that eye is ready. With enough knowledge of the person to know that it is.