---
title: "What K3X agents do"
description: "Agents are goal-based AI workers you describe in plain English. They run on triggers, webhooks or a selection and log every action."
url: https://help.k3x.ai/en/articles/74-what-k3x-agents-do
collection: "AI Agents"
updated: 2026-10-04
language: en
---

# What K3X agents do

Agents are goal-based AI workers you describe in plain English. They run on triggers, webhooks or a selection and log every action.

An agent is an AI worker that lives in the **Agents** tab. You give it a goal in plain English, tell it when to wake up, and it does the work: qualifying companies, following up, moving deals, creating tasks, enriching records, pushing data to other tools.

## Goal-based, not flowchart-based

You do not draw a flowchart. You write what the outcome should be, the way you would brief a teammate: "Follow up with every new company that has not replied after two days, and stop as soon as they answer or book a meeting." The agent breaks that into steps and adapts to what each company does.

Each agent has four parts, all on its **Agent** tab:

-   **Goal**: free text describing the outcome and the rules.
    
-   **Triggers**: the events that start a run, such as New Company, Deal Updated or Webhook.
    
-   **Cadence**: how often it keeps working, for example Continuous.
    
-   **Action** steps: the plain-English steps it executes, with the AI Analyzer checking for a response after each one.
    

## Three ways an agent starts

1.  **On a trigger.** Something happens in your workspace, such as a company being created or a deal changing. The agent runs on that record automatically.
    
2.  **On a webhook.** Another tool (Make, Zapier, n8n, Calendly, your website form) posts JSON to the agent's own URL. The agent reads the payload and runs.
    
3.  **On demand.** You pick companies or people in a grid, open the **Action Bar** and choose **Agents**, or click **Enroll Agent** on a single record page.
    

## Every action is logged

Agents never work in the dark. Each run appears on the agent's **Activity** tab with what it saw and what it did, and every email, SMS, note, deal change or field update lands on the record's timeline as a System or channel entry. The Agents list shows **Errors**, **Success Rate**, **Credits (7d)** and **Last Execution** per agent so you can spot a problem at a glance.

## Agents spend credits

Agent runs are AI actions, so they spend credits from your single pool, the same pool that pays for calling, texting and data reveals. There is no separate agent bucket. If the pool runs out, the CRM keeps working and agent runs pause until you top up or auto-recharge kicks in. See the current usage in **Settings → Billing** or in the **Usage this cycle** bar in your profile.

## What agents are good at

-   Speed to lead: reply to a new inbound company within a minute.
    
-   Follow-up chasing: multi-step email and SMS sequences with built-in waits.
    
-   Data entry: normalize inbound data, create the company, person and deal, assign an owner.
    
-   Pipeline hygiene: flag stalled deals, update stages, create tasks.
    
-   Call debriefs: summarize a call and write the next step on the deal.
    
-   Handoffs: post results to Slack, a sheet or another CRM through an outbound webhook.
    

Start from one of the 13 templates or describe your own. The next article walks through creating your first agent.

## Related

-   Create your first agent: describe it or pick a template
    
-   Goal, triggers and cadence
    
-   Monitor agents: Activity, Deliveries, errors, success rate and credits
    
-   Your 7-day free trial and the Adaptive plan