AI Research Step

Step Guide: AI Research

Knowledge

Search the web and synthesize findings with AI

Overview

The AI Research step runs an agentic web search loop — it issues searches, reads results, and synthesizes a brief with citations. It supports two actions: Search returns a markdown summary with sources, and Structured Search populates user-defined fields from the research results.

When to Use

Build account briefs for prospects by researching their company, recent news, and key people
Gather competitive intelligence from public sources
Research a topic and extract structured data points (e.g., company size, industry, funding)
Monitor industry trends by searching recent sources within a time window

How It Works

Write a research goal as a prompt with {{stepId.fieldName}} references to upstream data. The agent issues multiple web searches (controlled by search depth), reads the results, and synthesizes findings. Search depth controls thoroughness: Quick runs up to 2 searches, Standard up to 5, and Deep up to 10. Advanced options let you restrict or exclude specific domains and filter by recency.

Actions

Synthesized brief with citations

How it works

The agent searches the web based on your prompt, reads the results, and produces a markdown summary. Output fields are locked: summary (markdown write-up), sources (array of citations with title, URL, snippet, and published date), source_count, and query_count.

Use Standard search depth for most use cases — Quick is faster but less thorough, Deep is slower but covers more ground
Combine with AI Text Generate to reformat or extract specific information from the summary

Structured Search

Search the web and populate user-defined structured fields

How it works

Define custom output fields and the agent populates them from its research. Citation metadata (sources, source_count, query_count) is appended automatically. Use this when you need specific data points rather than a narrative summary.

Name output fields clearly so the agent knows what to look for (e.g., employee_count, headquarters, recent_funding)
Sources, source_count, and query_count are appended automatically — don't include them in your custom fields

Tips

Be specific in your research goal — "Find the CEO and employee count for {{step.company}}" works better than "Research this company"
Use site restrictions to focus research on trusted sources (e.g., linkedin.com, crunchbase.com)
Use the recency filter to limit results to recent sources when freshness matters

Related Steps

Step Guide: AI Text Generate

3 actions

Generate text or structured output with AI

Step Guide: Form

Collect user input with custom forms

Step Guide: For Each

Iterate over array elements