Agents overview

Agents are AI assistants that automate complex tasks in Slant. Unlike simple workflows that follow fixed rules, agents can reason about situations and take appropriate actions based on context.

​

What agents do

Agents combine AI reasoning with automated actions. You define what the agent should accomplish, and it figures out how to get there using the tools you give it. Common agent use cases:

• Research and summarize — Gather information about a client before a meeting
• Draft communications — Write personalized emails based on context
• Categorize and route — Sort incoming items based on content analysis
• Process data — Extract and organize information from emails or documents

​

How agents work

​

Agent types

​

System agents

Pre-built agents that ship with Slant for common use cases:

• Meeting prep — Summarize client history before meetings
• Email categorization — Sort and tag incoming emails
• Contact enrichment — Research and fill in contact details

System agents are ready to use with minimal configuration.

​

Marketplace agents

Agents created by the Slant team or community that you can install:

• Browse available agents in the marketplace
• Preview what each agent does before installing
• Install with one click and configure for your needs

​

Custom agents

Build your own agents for unique workflows:

• Define the agent’s goal and available actions
• Configure when it should run
• Test and refine based on results

​

Agents vs workflows

Use agents when: The task requires judgment, content varies widely, or you’d need many branching conditions in a workflow. Use workflows when: The logic is predictable, you need guaranteed consistency, or the task is simple enough for explicit rules.

​

Agent capabilities

Agents can perform many actions:

​

Communication

• Draft emails for review
• Send notifications to team members
• Create follow-up tasks

​

Data operations

• Research information from records and history
• Categorize and tag records
• Update fields based on analysis

​

Analysis

• Summarize conversations and meetings
• Extract key information from text
• Identify patterns and insights

​

Running agents

Agents run in different modes:

​

Automatic

• Triggered by events (new email, meeting completed)
• Scheduled at regular intervals
• Initiated by workflows

​

Manual

• Click to run on demand
• Available as actions on record pages
• Useful for testing and one-off tasks

​

Monitoring agent activity

Every agent run is logged:

• Run history — See when the agent ran and what triggered it
• Actions taken — Review what the agent did
• Results — Check outcomes and any errors

This transparency helps you understand agent behavior and troubleshoot issues.

​

Best practices

Start with supervision. Configure new agents to draft rather than send, so you can review before actions complete. Define clear goals. Agents work best with specific, well-defined objectives rather than vague instructions. Provide good context. The more relevant data an agent has, the better decisions it makes. Review runs regularly. Check agent activity to ensure it’s behaving as expected and refine as needed.

​

Next steps