Skip to main content

​
Summary

Agents are one of the core control surfaces in TruAgents. They are not just generic AI personas. In the current product, agents appear as reusable roles that influence how campaigns generate, review, or shape outbound work.

​
Who this is for

  • Anyone learning how AI behavior is organized in TruAgents
  • Campaign managers configuring outreach workflows
  • Operators reviewing how generated content was produced or checked

​
What the current app suggests

The Agents area currently separates at least two major agent types:
  • writers
  • reviewers
The current app also suggests agent records can carry details such as:
  • title or name
  • model choice
  • template or specialty
  • voice or style information
  • starter status
The app also suggests that writers and reviewers can be reused across multiple campaigns, and that teams can inspect active campaign counts for each one. That is a strong clue that agents are operational building blocks, not disposable one-offs.

​
The right mental model

  • a writer helps generate content
  • a reviewer helps evaluate or approve it
  • campaigns can depend on one or both of those roles
That means agents are closer to reusable workflow building blocks than one-off chat assistants.

​
Writers vs. reviewers

​
Writers

Writers are the generation layer. They shape how content gets produced for campaign work. In the current app model, writers can carry more descriptive surface area, including things like name, title, model, voice, and specialty or template associations.

​
Reviewers

Reviewers are the evaluation layer. They help determine whether generated work should move forward. They appear more tightly focused than writers, but they are still reusable configuration objects rather than isolated per-message actions.

​
What makes an agent reusable

An agent becomes useful when the team can explain:
  • what role it plays
  • which campaigns rely on it
  • what model or style choices define it
  • whether it is a starter/default option or a deliberate custom configuration
Simple version: if the team cannot explain what the agent is for, it is probably configuration noise.

​
How agents connect to the rest of the product

Agents should be read together with:
  • campaigns, because campaign setup references writers and reviewers
  • drafts and approvals, because agent behavior influences what reviewers later see
  • communications, because downstream outcomes are where agent choices become visible

​
Good agent hygiene

  • give agents names and titles that explain their job clearly
  • reuse strong agents intentionally instead of cloning near-duplicates
  • distinguish writing responsibility from review responsibility
  • treat model, voice, and specialty choices as workflow decisions, not decoration
  • review campaign impact before deciding an agent is working well

​
Common failure patterns

​
Too many near-duplicate agents

This makes campaign setup harder and hides which configuration is actually trusted.

​
Confusing campaigns with agents

A campaign is the outreach definition. An agent is one of the reusable pieces the campaign depends on.

​
Treating agents like chat toys

The current product model is much more operational than that. Agents are part of repeatable campaign and review workflows.

​
Common misunderstandings to avoid

  • Agents are not the same thing as campaigns.
  • Agents are not just prompts floating around without structure.
  • An agent definition can matter across multiple campaigns, drafts, and runs.
  • A writer and a reviewer are not interchangeable just because both use AI.

​
Plain-English example

If you change a writer, you may change how future campaign content gets generated. If you change a reviewer, you may change how future content is judged or approved. That is why agent changes can have effects well beyond one page or one run.