Selling to Machines
ai, techtitle: Selling to Machines created_at: 2025-11-12T14:52:16.823Z updated_at: 2025-11-12T14:52:16.823Z tldr: Agentic to Customers (A2C) and Direct to Agentic (D2A) is_published: 1 star: true category: ai, tech share_type: add
Have you asked Alexa to restock fruits? Do you know where it was sourced from? Which vendor supplied it? For now, it's all manual in the backend. A person goes to a vendor, does the quality inspection and orders. As the AI is most extensively used, we are switching to getting opinions from an AI and buying from there... a machine is acting as a middleman. In the near future, this AI can make calls to schedule your booking/reservations/appointments matching your requirements.
Now reverse this interaction on the web. You are the vendor. In the first example, there is a person approaching your shop. Based on the interaction and the liking of the interaction, there is a purchase. In the second, where the AI is placing the order, it comes with the "cutoff facts". What's this? Say a machine is looking for an A-grade product with the budget of 10$, if your shop is selling at 10.5$, the machine is not going to stop here, it moves on to the next one, or if it's looking for a barber appointment, it's not going to ask if the budget matches if the store-front (ie, the booking platform for now) shows the reservations are full, even if your visited for more than 4 times in a row. Unless you tell it to try to book it at that spot itself.
The challenge begins when the storefront you are offering for the online customers is a machine, and there are no factors other than facts to sell it......
credits to @luckycat Katja Forbes on the wonderful keynote on "Machines are customers too".
The AI revolution has introduced A into the selling lifecycle - Agentic to customers (A2C) and Direct to Agentic (D2A). A2C is where AI is selling on behalf of a storefront to the customer and D2A is the storefront selling to AI. Right now, this is a "30 trillion dollars" untapped market and disruption.
Yup. As always, the code is the real rock star. Now all we need to do is focus our interfaces to machines - like XML, YAML, JSON as a consumable to the AI for the value proposition. Or maybe an MCP connect interface where it can talk to our LLM to negotiate (trying this out... will update).
Further reading
Having thoughts ?
Add yours