Essays

EssayStatus
Why Agent Relations, and why now
Why the work needs a name, and why the moment is now. Read it below.
Live
The six domains, explained
Discover to Measure, one at a time, with an example of what good looks like in each.
Draft
Agents are not search engines
Why optimizing for an agent that acts is a different job from ranking in a list of links.
Draft
The failed agent task is your most honest feedback
Every task an agent could not finish with you is a customer you did not know you lost.
Coming soon

Why Agent Relations, and why now

Published November 2026

Ask a friend where to eat tonight and you get one answer. Ask an agent and you get one answer too, but it is built from everything the agent could read, ask and do in a few seconds. The restaurants it could not read, or could not book, never made the list.

This is already how a growing share of people research, compare, book and buy. They hand the first pass to an agent. The agent decides what to read, whom to ask and what to recommend. For organizations, that is a new audience, and it behaves differently from every audience before it.

Organizations have met this kind of shift before. When the press mattered, companies built public relations. When investors needed a steady, honest line to the business, they built investor relations. When developers became the buyers of software, they built developer relations. Each was a discipline with its own craft, its own owners and its own measures. Agent Relations is the same idea for the agents that act for people.

Agent Relations is everything an organization does to help external AI agents find, trust, and do business with it. It runs in both directions: agents finding and presenting you, and you answering and transacting with them.

Today nobody owns it. Marketing controls the story but not the endpoints. Engineering builds the endpoints but not the story. Commerce owns the prices, service owns the answers, and legal owns the risk. Each holds a piece, and an agent meets all of them at once. When the pieces disagree, the agent repeats the wrong one to a customer.

The tools are arriving faster than the practice. Crawl rules for AI agents, llms.txt, schema.org markup, the Model Context Protocol, agent-to-agent cards, commerce protocols and signed-agent identity all exist, and each is at a different stage. A busy leader cannot be expected to track them. A practitioner can.

That is the case for a field. It needs a shared vocabulary, so a marketer and an engineer mean the same thing by the same word. It needs a framework, so an organization can see where it stands and what to do next. It needs Standards Mapped in plain language, a way to prove skill, a place for practitioners to meet and honest research on how far organizations have come.

It is also early. That is a good reason to start now. The habits organizations form in the next few years will set how agents treat them for a long time. The work is not large at the start: let agents in, state your facts clearly, and watch what they try to do. The rest follows from what you learn.

The Institute exists to define that practice and keep it honest. We are writing the Framework in the open, and we will say plainly where we do not know yet.

Read new essays as they come