AI agents are taking over purchasing decisions – and your brand may not even be visible
A new warning from a Harvard researcher should make every Swedish business leader stop and take notice. The rules of e-commerce are being rewritten – not by consumers, but by AI agents acting on their behalf. Swedish companies that fail to prepare risk becoming invisible at precisely the moment it matters most.
A woman in Shanghai spoke nine words to her phone: “Order my usual lunch, deliver it 20 minutes later.” That was all. Meituan’s agent “Xiaomei” took over – found the right restaurant, applied her preferences, paid, tracked the courier and adjusted the delivery time automatically. The woman never clicked once.
That is the scene Howard Yu, researcher at Harvard and IMD, uses to illustrate what he calls the end of the click economy. For 25 years, the entire internet has been built on a single assumption: that people click. Google made money from search, Amazon from comparison, Meta from scrolling. Performance marketing via search engines and social media accounts for $600 billion of a global advertising market worth over $1 trillion per year. Now those clicks are slowly disappearing, as AI agents read reviews, process returns and actively choose not to show ads.
Yu draws on economist Ronald Coase’s Nobel Prize-winning theory from 1937: companies exist because markets are expensive to use. AI agents are now demolishing those transaction costs, on both the supply and demand sides. IMD professor Mark Greeven has coined the term “agent shelf” – the short list an AI agent selects from before you are even shown to the consumer. It is a machine making choices, and it chooses based on what it can measure and verify: delivery reliability, clarity of return policy, speed of error resolution and whether product data is sufficiently well-structured to be trusted. Yu calls it “machine-readable trust” – brands no longer tell stories to people, they submit specifications to agents. Your return policy becomes a data field. Your delivery precision becomes a percentile.
What does this mean for Swedish companies?
Sweden is one of Europe’s most digitally mature countries, but having a fast website and good Google rankings is no longer enough. When AI agents such as ChatGPT, Gemini and Perplexity begin acting as purchasing assistants – and soon as genuine decision-makers – the entire logic of how you are found, recommended and chosen changes.
China is 2–3 years ahead of the Western world in this development, with infrastructure such as Alipay, WeChat Pay and densely connected logistics networks. As many as 83 percent of Chinese people say they trust AI, compared to only 39 percent in the US (Stanford 2025 AI Index). But the development is moving quickly westward. ByteDance recently launched a built-in agent in a mobile phone – the first 30,000 units sold out immediately. The battle over who controls the agent layer has already begun, and Swedish e-commerce businesses and brands are right in the middle of it.
The consequences of failing to act
If your brand is not on the AI agent’s short list, it does not matter how good your ads are or how optimised your SEO is. You simply do not exist in that purchasing flow.
Walmart was forced to pull back from its collaboration with OpenAI on automated checkout precisely because delivery estimates were wrong and shipping costs incorrectly stated – the agent could not trust the data. That is a warning sign: gaps in product data, delivery precision or return processes are no longer just customer service problems. They are the reason AI agents choose to exclude you.
For Swedish companies, this means that past achievements and strong brand identities do not automatically carry over into the new reality. A well-known Swedish retail chain with unstructured product data or unclear return information can lose out to an unknown competitor that happens to have cleaner data fields.
How to turn this into an advantage
The good news about “machine-readable trust” is that it is measurable – and therefore improvable. Yu identifies four concrete key metrics that determine whether an AI agent chooses to recommend you:
- Delivery precision – what proportion of orders are delivered on time
- Refund speed – how quickly a return results in a refunded amount
- Time to resolve discrepancies – how long it takes to handle a problem in an order
- Error rate in product data – how clean and accurate your product information is
Yu is clear: “If it takes more than a day to produce these figures, you have your answer. To an agent, your brand is not yet verifiable.”
Swedish companies that address these four metrics now are positioning themselves to appear on agents’ short lists while competitors are still trying to understand what is happening.
Kaistone helps you become visible to AI agents
At Kaistone, we are Sweden’s leading experts in AEO (AI Engine Optimization) – optimisation for how brands are perceived, described and recommended by AI-driven search engines and agents such as ChatGPT, Gemini and Perplexity.
Our core product, Kaistone AI Audit, gives you a complete picture of how your brand appears in the AI search landscape – and what you need to do to improve your position. The audit consists of three parts:
- AI Search Analysis Report – a thorough examination of how your brand is mentioned in AI responses, what tone and positioning is used, and how technically accessible your information is to AI agents
- Prioritised action plan – concrete measures ranked by impact, with time estimates for each step
- Implementation cost estimate – so that you know exactly what it costs to act
In a world where “machine-readable trust” determines whether you are even visible, Kaistone AI Audit is the starting point for taking control of your brand’s AI visibility.
Summary
AI agents are taking over the role of purchasing assistants – and soon decision-makers – in a global e-commerce market worth trillions of dollars. What determines whether your brand appears on the agent’s list is no longer reach or advertising budget, but machine-readable credibility: clean product data, fast returns and reliable deliveries. Swedish companies that act now can gain a significant competitive advantage. Those who wait risk becoming invisible.
Do you want to know how your brand is seen – or not seen – by AI agents today?