Back office ops · Production

Gemini Deep Research: Building an agentic research assistant that browses 100+ websites to generate comprehensive reports in minutes

The problem

Users conducting multi-faceted research faced the burden of opening 50-60+ browser tabs over hours or days and often gave up before completing their research.

Workflow diagram · grounded in source
1
User submits research query
trigger
“takes your query, browses the web for about five minutes, and then outputs a research report for you to review and ask follow-up questions”
2
Research plan generation
ai_action
“the model produces its first stab at the, at the research query at, at how it would break this down. And then invite the user to come and kind of engage with how they would want to steer this”
3
Parallel web search and browse
ai_action
“the model figures out which of the sub steps that it can start exploring in parallel, and then it primarily uses like two tools. It has the ability to perform searches and it has abilities to go deeper within, you know, a particular webp…”
4
Iterative grounded search
ai_action
“this notion of being able to read outputs from the previous turn, uh, ground on that to decide what to do next, I think was, was key”
5
Analysis and self-critiqued draft
ai_action
“we kind of entered this, uh, analysis mode and here there can be inconsistencies across sources. You kind of come up with an outline for the report, start generating a draft. The model tries to revise that by self critiquing itself”
6
Research report delivered
output
“outputs a research report for you to review and ask follow-up questions”
7
Follow-up and iteration
feedback_loop
“we broadly are basically trying to teach the model to be able to do all three and the kind of side-by-side format allows sort of for the user to do that more easily”
Reported outcome

Deep Research generates comprehensive research reports by browsing 100+ websites in about five minutes, with users reporting that tasks previously taking hours can now be completed in minutes.

Reported metrics
Research report generation timeabout five minutes
Websites browsed per research query100+
competitive research time (AI vs manual)10 mins vs previously at least 3 hours
coding task time (AI vs manual)about 30 minutes vs a day or so
Reported stack
GeminiGemmaGoogle Docs
Source
https://www.latent.space/p/gdr
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Deep Research generates comprehensive research reports by browsing 100+ websites in about five minutes, with users reporting that tasks previously taking hours can now be completed in minutes.

What tools did this team use?

Gemini, Gemma, Google Docs.

What results were reported?

Research report generation time: about five minutes; Websites browsed per research query: 100+; competitive research time (AI vs manual): 10 mins vs previously at least 3 hours; coding task time (AI vs manual): about 30 minutes vs a day or so (source-reported, not independently verified).

How is this back office ops AI workflow structured?

User submits research query → Research plan generation → Parallel web search and browse → Iterative grounded search → Analysis and self-critiqued draft → Research report delivered → Follow-up and iteration.

Related back office ops cases