Bullet Thesis: how will we advertise to AI agents?

Hi, it’s Michael from firstminute capital. We are a $500m seed fund backed by 130 unicorn founders. In a previous life I was a foreign correspondent for the Financial Times for 10 years, and the founding editor of Sifted. Along with my brilliant colleagues, I invest in European pre-seed and seed software companies like Vocca, Scalera, Granola and ai.work.

—

Here’s the big question I can’t stop thinking about: When AI agents become the primary users of the internet... how will we advertise to them?

We are at foothills of a major shift. AI agents like OpenAI’s Operator can already perform complex online tasks on our behalf—booking restaurants, buying clothes, finding hotels.

According to Meta’s Yann LeCun, this is just the beginning. When I was in Paris recently, he said: “In the future, every single intersection in the world will be mediated through AI systems. We will all have a staff of AIs.”

Imagine that: 5 billion internet users, each managing, say, 10 agents. That would be 50 billion agents roaming the internet—evaluating options, making decisions, and executing transactions on our behalf.

In this world, agents—not humans—become the dominant users of the internet. And that begs a lot of questions. But the one I am thinking about is:

How will we market to them?

A Multi-Trillion Dollar Advertising Industry—Up for Grabs?

Advertising powers the biggest internet businesses in the world—Google, Meta, Amazon, ByteDance. Over $1 trillion will be spent globally this year alone.

Yet these ads are designed for humans: visual imagery, emotion-laden copy, attention hacking.

But an AI agent doesn’t scroll Instagram. It doesn’t “click.” It doesn’t feel FOMO. It simply evaluates data and makes decisions based on rules, probabilities, and performance history.

So what will advertising look like in a world where the end consumer isn’t a person—but an autonomous software agent?

What MarTech Is Talking About Today

To be clear, this is not what most people in marketing tech are discussing. Let me quickly break down what they are talking about, which is:

1. AI-Powered Content Creation Tools, which enable marketers to produce high-quality assets efficiently (that target humans!). These are tools for text generation (like Jasper and Copy.ai), video creation (like HeyGen and Synthesia) and image generation (like DALL·E and Midjourney).​

2. AI-Driven Customer Segmentation and Personalization Tools, which analyze customer data to identify distinct segments and deliver personalised marketing messages to humans.​ These are tools like Segment from Twilio, Dynamic Yield and Optimove.

3. AI-Powered Customer Support and Chatbots, like Ada or Sendbird, which uses AI to engage website visitors in real-time conversations, qualifying leads and answering queries.

There are also lots of people excited (see this from A16Z) about what happens with this all comes together and moves from AI tools that helps marketing teams to whole agentic systems that ARE marketing teams. This is a more classic software-replacing-work and venture-focused way of thinking about the disruption coming in the marketing stack.

What I am talking about is different though.

Because ultimately this is all serving a market still oriented around humans. None of this being discussed is about influencing AI decision-makers. It’s all about empowering humans to do marketing more efficiently.

Where the Future Is Beginning: Otterly and Profound

That said, there are a few early pioneers who have seen how AI tools like Chat GPT are being more and more used for search (e.g asking Chat GPT or Perplexity for recommendations for the best restaurant in Paris).

These companies are creating a brand new field, which is SEO for LLMs. They are trying to ask the question: how do I make sure my product appears in Chat GPT and not my rivals?

Otterly is one of these building tools for SEO for LLMs. In a world where customers ask AI chatbots for answers, brands want to be the result returned. Otterly helps them understand how to structure content so it’s more easily found by AI tools.

Profound (backed by Khosla Ventures) is attacking the same challenge, essentially an AI-native SEO analytics and optimisation platform. It also helps brands understand how they appear in AI search experiences—tracking queries, surfacing gaps, and optimising messaging for LLM interfaces.

But ultimately, as Thomas Peham, the founder of Otterly, says, we will over the coming decade see a shift of AI “from an acquisition channel to a consumption channel” with both consuming on his platform. This is part of a wider shift of the agents becoming customers themselves.

In short, we are just entering the foothills of something huge brewing in the world of marketing. Because we will soon be marketing to robots not humans.