Agent Interaction guidelines (AIG) – Linear developers

Agent Interaction guidelines (AIG) – Linear developers

1 minute, 57 seconds Read

Agents change how software is planned, built, assessed and implemented. Because agents produce work in abundance, roles and workflows are reformed. The value shifts to the orchestrate of input, context technology and the assessment of output.

This shift requires a new contract for interaction between people and competitor. The agent interaction guidelines (AIG) are the fundamental, evolving principles and practices for designing agent interactions that integrate more naturally into human workflows.

An agent must always reveal that it is an agent

When people and agents work side by side, people immediately need certainty about whom they deal with. The agent must clearly signal his identity so that he can never be seen for a person.

Fig. 01Clear boundary between human and agent users

An agent must inhabit the Native Platform

By default, agents must be able to work through existing onion patterns and standard actions of the platform in which they work.

Fig. 02The agent can use the same actions that a human user would do

An agent must give immediately feedback

Silence leads to uncertainty. When it is called, an agent must immediately, but unobtrusive, give feedback to reassure the user who received a request.

Fig. 03The agent immediately indicates that it processes the request

An agent must be clear and transparent about his internal condition

Agents must clearly indicate whether they think, wait for input, implementation or work completed. People must be able to understand what is happening at a glance and, if necessary, inspect the underlying reasoning, tool calls, prompts and decision logic.

Fig. 04The agent’s reasoning is completely transparent and is open to inspection

An agent must request to switch off

When asked to switch off, an agent must take a step back, and only once again received a clear signal to do this.

An agent cannot be held responsible

There should be a clear delegation model between people and agents. An agent can perform tasks, but the ultimate responsibility must always stay with a person.

Fig. 05Clear delegation flow between people and agent

The guidelines of the agent interactions are written with the community in mind. If you build agents and think the same challenges, join our Slack community to contribute to the conversation.

This page is a living document and we are constantly expecting it to add as we learn more in practice.

#Agent #Interaction #guidelines #AIG #Linear #developers

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *