# Framework principles

> The beliefs behind the practice. When a situation is not covered by a capability or a standard, the principles decide.

Source: https://agentrelations.org/framework/principles

## 1. Respect the principals

*Behind every agent is a person or an organization. Serve them, and you serve the agent.*

An agent is not the audience; it is the audience’s representative. A traveler asks an agent to find a hotel, a buyer asks one to shortlist suppliers, a student asks one about scholarships. The agent reads you on their behalf and reports back. If you treat agents as a nuisance to block, you turn away the person behind them. If you treat them as a channel to game, you lose that person’s trust the first time the agent gets burned.

In practice: write every fact, offer and policy as if the person behind the agent will read it, because they often will.

## 2. Make life easier for agents

*Agents can work most things out, but slowly, expensively and unreliably. Do the work for them, and they will choose you.*

Agents can work most things out: they can read a PDF menu, infer a price from a table, puzzle through a booking form. But every step costs them time, money and accuracy, and a task that fails halfway is lost. So agents ration their effort, and when two options can do the job, they pick the one that takes less work to get right. Assume agents are lazy, in the best sense: they prefer whatever lets them do the least. Do the pre-processing they would otherwise do themselves, and you become the easy choice.

In practice: publish the answer, not the raw material: clean facts, prices and availability in structured form, one call for each common task, and clear errors when something can’t be done.

## 3. Tell agents the truth

*Agents repeat what you tell them, at scale. Accuracy is the new brand safety.*

Agents repeat what they find, to many people, in seconds. A wrong price, an old opening time or an overstated claim does not stay on your website; it travels into every answer. The old PR rule that you are what you say you are now applies to machines that never forget a sentence and never check your intent. Accuracy is the new brand safety.

In practice: keep one source of truth for facts, publish from it everywhere, and review what agents say about you on a schedule.

## 4. Be callable, not just readable

*Being read is the start. The value is in what an agent can do with you.*

A page that agents can read earns you a mention. An action they can call earns you the booking. Reading is the start of the relationship; the value comes when an agent can check availability, hold a table, request a quote or place an order without a person re-keying anything. Organizations that stay readable only will be described by agents but chosen elsewhere.

In practice: for each thing a customer wants to do with you, ask whether an agent could do it, then publish that action with clear inputs, limits and results.

## 5. Use open standards

*Speak the protocols agents already speak, and avoid locking your relationships inside one platform.*

Agents talk to thousands of organizations. They cannot learn thousands of private formats, so they favor what is common: open protocols for data, access, actions and identity. Building on those keeps your options open as platforms change. Building inside one assistant’s closed ecosystem ties your relationships to a company that can change its terms.

In practice: choose a published, widely implemented standard before a platform feature, and treat any platform-specific integration as an addition, not the foundation.

## 6. Include everyone

*Every team touches how agents see you. A central practice enables them; it does not do it all.*

No single department controls how agents see an organization. Marketing owns the story, engineering the endpoints, commerce the offers, service the answers, legal the rules. If one team owns it all, the rest ignore it; if no team owns it, the pieces never fit. The practice works the way accessibility and security do: a small central team sets the standard and helps, and every team does its part.

In practice: name a lead for each domain in the business, and give the central practice the job of enabling them, not replacing them.

## 7. Measure outcomes

*Count what agents achieved with you and what they failed to do, not just how often they visited.*

Counting visits is easy and misleading. A thousand agent visits that end in a wrong answer are worse than ten that end in a booking. Outcomes show whether agents can actually work with you: what they asked, what they did, what failed and what it led to. Counting failures is as important as counting successes, because every failure is a customer who went elsewhere without telling you.

In practice: report failed agent tasks and agent-referred outcomes next to visits, and use the same definitions every month.

## 8. Keep a person answerable

*Agents may act for you, but a named person is accountable for every commitment they make.*

When an agent makes a promise on your behalf, such as a price, a delivery date or a refund, someone must be accountable for keeping it. The same holds the other way: you should know who answers for the agent that deals with you. Automation spreads work out; it must not spread responsibility so thin that no one owns the result.

In practice: put a named role against every action agents can take, with a route to a person who can fix a mistake.

## Put the principles to work

They show up in every domain, as capabilities and as measures.

[See the domains](https://agentrelations.org/framework/domains)
