Agentic shopping is creating a crisis for marketers. Here’s how to fix it
Understanding your customer is the foundation of marketing. What drives decisions? What does the person in front of you respond to? What makes them choose you? Those questions sit at the back of every marketer’s mind when designing a campaign. So what is a marketer supposed to do when the customer turns out to be an AI agent?
For more than two decades, we built the internet to filter bots out, and we were right to. Now many of the ones arriving at a website are not really bots at all. They are customers in a new form. Tens of millions of Americans already trust a personal AI assistant, such as Meta’s Muse or Instinct, to book, buy, and arrange aspects of their lives for them. Meta told investors on its last earnings call that more than 10 million people in the United States and Canada had set up Muse agents within six weeks, that a quarter of them use it daily, and that the average purchase runs above $200. Those people are not browsing. They are delegating.
Websites are reporting record numbers of visitors as these artificial intelligence assistants move across the internet at machine speed, working through a task someone gave them. The marketing reflex has been to filter harder. These are not real visitors, the thinking goes, and we need a way to tell AI agents apart from customers.
Telling them apart does matter, though not for the reason most people assume. If a business keeps blocking AI assistants, it loses in two different ways.
The first is immediate. AI systems are built to finish the task they were given. An assistant that runs into a wall on your site does not file a complaint or abandon the errand. It moves to the next site that enables them to complete their task, which is very likely a competitor’s. The customer who sent it never finds out you were an option.
The second is slower and harder to reverse. These agents generate feedback as they work, and that feedback shapes what the next generation of models knows about the web. A business an assistant could never complete anything with does not sit at the top of that context. It is remembered, if at all, as somewhere not to bother trying.
The floor, not the goal
Letting agents in is the floor rather than the goal. The more interesting work is happening at companies that are not simply opening the door but deciding, deliberately, what an assistant may do once it is inside, where a human has to approve a step, and at which point in a process a person must take over. My company scanned 2,053 organizations to test what an AI assistant sent by a customer could actually accomplish on their site. Just 2% have built a way for a customer to authorize an assistant with access they can limit and withdraw. None have reached the top level of readiness, where agents are a supported channel across web, interface, and voice.
Consumers, meanwhile, have already made their decision. They are handing assistants real money and real tasks, and the usage numbers are not ambiguous. The next competitive ground is going to be at the agent level, and in the agent language, as businesses compete for the attention and the ease of use of software, rather than people. A new digital channel is opening, and marketers will have to build a strategy for appealing to something with no eyes.
That is not a figure of speech. Every digital channel that marketers have learned in 30 years assumed a person at the other end. Someone looks at the page, reads the headline, feels something about the imagery, and decides. Search, social, display, and email all rest on that assumption, but it has stopped being reliably true.
The gap this leaves is enormous. Of the organizations we measured on whether they have built something an assistant can actually act upon, the average score is 2.5 out of 100. Just 2% percent publish a product catalog a machine can read. Just 7% have published any position at all on whether assistants are welcome. The companies doing best are overwhelmingly software businesses, which score 57.7 on average against a field average of 38.8, not because software is clever, but because it is already built for machines reading machines.
There is a risk in doing nothing, and it is not the one most marketers would guess. When a page became outdated, a promotion ended, or a referral program expired, the practice was to archive it or take it out of the navigation and move on. An assistant does not navigate. It reads what is there, and a page hidden from a menu is as findable as the homepage. Some 23% of the organizations we scanned had a file published telling an assistant what is current and what it may use. For the rest, an assistant can return to its customer with an offer that expired two years ago, presented as though it stands today.
A question of intent
Set against all of this is the part that should make marketers genuinely interested rather than merely anxious. An assistant arrives carrying intent. Not a referrer and a bounce rate, but a direct and specific statement of what a customer wants, delivered before the customer gets there. No channel in the history of marketing has handed anyone that. A business able to hold a conversation with an assistant learns exactly what the person behind it is looking for. Only 7% of the organizations we scored can connect to an assistant at all, and it is reasonable to assume those 7% are pulling ahead in a competition before their rivals even realize it exists.
None of the work required here is exotic. Publish a product catalog a machine can read. Publish a file that tells an assistant where things live and what it may do. Give customers a way to authorize an assistant for a narrow task, with access they can withdraw. Decide, deliberately and in writing, whether agents are welcome, rather than leaving a 20-year-old configuration to answer on the brand’s behalf. And then, treat the result as a channel, which means tracking why these visitors arrived, what happened when they did, and how many converted into customers.
The reason to treat this as urgent rather than interesting is that positions in this channel are being allocated right now, quietly, according to which businesses an assistant can actually finish a task with. Marketers spent the mobile transition arguing about whether the site needed to be responsive while competitors built things that worked only on a phone. This shift is further along than it looks, and the companies winning it are not winning with better creative. They are winning because an assistant could read them.
内容来自 Fast Company,版权归原作者所有。
阅读原文 ↗