Deep Research Agent Ai Agent Workflow Inspiration

Conduct iterative deep research on a topic by generating queries, summarizing web results, reflecting on gaps, and finalizing insights.

How this Ai Agent Workflow Helps

This workflow powers a deep research assistant that takes a user-defined topic and performs iterative cycles of web research and summarization. At each cycle, it generates a search query, gathers online sources, summarizes them, and reflects to identify gaps. The cycle repeats for a configurable number of times before producing a final markdown summary with citations. It is useful for writing comprehensive briefs, conducting market research, or understanding complex topics in depth.

Who this Ai Agent Workflow Is For

Go-To-Market Pros

You care about improving the entire lifecycle of buyer and customer journeys. And you want inspiration for areas where Ai may help.

Ai Builders

You want inspiration for real-world ways to help people as you're building your Ai product.

Pros Focused on the Stage

You're focused on the stage in your daily job and want to find ways for Ai to help you do more with less.

Ai Agent WorkflowExample as Inspiration

Does this AI agent workflow rely too much on AI and not enough on human know-how? Or the reverse? Is it missing steps or tools?

Note that this Ai workflow is presented as inspiration for what's possible. Adjust the amount, type and quality of the data inputs. Adjust how much or how little your human team mates (or you), AI and fully autonomous agents contribute.

And test it! Learn what works and what doesn't.

Don't forget! In the end, it's not just about efficiency. It's about delivering great experiences for your customers and customers-to-be.

1. Start Research
Trigger. User inputs a topic or question to research.
Trigger Event
2. Generate Search Query
LLM Call. Use a local LLM to generate a focused web search query based on the research topic.
Prompt
Ollama Google Gemini OpenAi LMStudio

Prompt

Generate a concise, relevant search query to explore this research topic: {researchTopic}

3. Perform Web Research
Tool Call. Use a search engine API to fetch relevant summaries or snippets from web pages related to the query. These sources will include titles, URLs, and summarized content used for downstream LLM summarization.
Research Asset
Tavily Perplexity SearXNG DuckDuckGo
4. Summarize Sources
LLM Call. Summarize the main findings from the gathered web pages using a local LLM.
Document
Ollama Google Gemini OpenAi LMStudio

Prompt

Summarize the key points from the following sources related to {researchTopic}. Include source references.

Sources:
{webSources}

5. Reflect on Summary and Decide
Router Agent. Analyze the current summary using an LLM to identify knowledge gaps. If gaps are found and max iterations has not been reached, generate a new search query and repeat the loop. If no gaps, proceed to final summary.
Decision
Ollama Google Gemini OpenAi LMStudio

Prompt

Based on this summary of {researchTopic}, determine if there are any significant knowledge gaps. If gaps exist, generate a revised search query to fill them. If the summary is sufficient, indicate that no further search is needed.

Summary:
{currentSummary}

6. Finalize Summary
LLM Call. Consolidate all research cycles into a single, well-structured markdown summary with citations.
Research Asset
Ollama Google Gemini OpenAi LMStudio

Prompt

Combine all findings from multiple research cycles into a single, structured markdown summary. Include clear citations from all sources used.

7. End Research
End. Workflow ends after outputting the final research summary.

Ask Pathdraft Ai agent to plan your next (or first 👊) Ai agent

Craft an Ai automation or agent plan, just for your b2b. Simple. The first step in automating across your buyer and customer journeys: inspiration for what to build. Talk with an Ai agent via your phone to quickly create your custom Ai agent plan.

1. Call (813) 906-1084 2. Chat with Pathdraft Ai agent 3. Agent emails your agent plan
Awareness Agent
Consideration Agent
Purchase Decision Agent
Onboarding Agent
Retention Agent
Expansion Agent
Advocacy Agent
Awareness Agent
Consideration Agent
Purchase Decision Agent
Onboarding Agent
Retention Agent
Expansion Agent
Advocacy Agent