Gemini Deep Research Rawks! (step-by-step ish)
With a bunch of up-front work, it can save many, many hours
First up, I’ve been aggressively upgrading my AI knowledge, toolkit, approach, etc, for the last few months. I’m nowhere near, say Tris Hussey or Christopher S. Penn (probably never will be), but I am pretty good at getting Gemini (as my primary “let’s figure this out” bro) to help me produce some pretty great stuff.
This week’s task was to jump into Gemini Deep Research, because I’m following Tris’s absolutely awesome 30 Days of AI content. It took until about Day 13 for me to be like “okay, time to slow down and figure this out”, and Deep Research was a big part of that (I’m good with Gems, okay with Opal Apps, so the idea was interesting).
Based on Tris’s suggestion, I worked with Gemini to figure out what a report I could review was. My success criteria was “what can Deep Research do, and what can’t it do”, but I didn’t want to just tell it “make a marketing report on AI”. Based on my experience with Gems and Opal Apps, spending half an hour working with Gemini on building a really robust prompt was worth it’s weight in gold, or maybe even latinum.
Before I share my process, here’s the end result:
I’ll explain the thinking after, but this was generated in less than an hour between prompt engineering with Gemini, report direction, review, editing with Deep Research, and tweaking the output into a fully functional webpage.
The actual 10+ page report itself is full of gold, but the point of the prompt was to understand an Agentic GTM model for a mid-level B2B SaaS company (I chose Riva so there was real world data) that was using Hubspot for content marketing/automation, and Salesforce for CRM and sales enablement.
Here’s the prompt that, eventually, Gemini and I figured out (Gemini has a tendency to “rabbit hole” with me, which can be great for finding new ideas, but is why I use gems for specific tasks to keep it focused!):
# GOAL
Conduct a comprehensive strategic deep dive into the “Agentic GTM” (Go-To-Market) shift, with a focused application for **Riva (rivaengine.com)**. The goal is to produce a report that defines the transition from traditional deterministic automation to autonomous, goal-seeking agentic workflows within a HubSpot/Salesforce environment.
# CONTEXT
I am producing a report for Riva. Riva is a startup currently operating with a classic B2B stack: HubSpot for content/marketing automation and Salesforce for CRM. We are seeing the 2026 shift where “Systems of Record” (CRMs) are being replaced by “Systems of Action” (Agents). I need to demonstrate a phased transition plan that modernizes their GTM engine without disrupting their current revenue.
# CORE CONTENT REQUIREMENTS
1. **The 2026 Landscape:** Why traditional “HubSpot Nurtures” and “Salesforce Cadences” are losing efficiency in the 2026 buyer climate.
2. **The Riva Audit (Current State):** Research Riva’s current market positioning and public technical footprint. Identify likely points of “Human Latency” in their HubSpot-to-Salesforce handoff.
3. **Multi-Agent Architecture:** How to deploy a “Research Agent” (to enrich Salesforce data) and an “Engagement Agent” (to handle HubSpot inbound) that collaborate autonomously.
4. **Transition Roadmap for Riva:** Provide a 3-phased plan:
- *Phase 1 (Crawl):* Data Enrichment & Signal Mapping. Using agents to turn CRM “static data” into “intent signals.”
- *Phase 2 (Walk):* The “Agentic Nudge.” Implementing agents that suggest the “Next Best Action” for AEs directly in Salesforce.
- *Phase 3 (Run):* Full Autonomous Outreach. Deploying agents to manage the “Boring 80%” of prospecting and mid-funnel follow-up.
5. **Tooling & Integration:** Research 2026-specific tools (like Composio or Agentforce) that bridge the gap between HubSpot and Salesforce most effectively.
# STYLE & STRUCTURE
- Use the **Pyramid Principle**: State the primary strategic recommendation for Riva first.
- Include a **Phased Migration Table** (Phase 1, 2, 3) with specific KPIs for each.
- Format for a C-suite audience (Laura Sweet, VP Marketing).
# INSTRUCTIONS
Before starting the research, provide a **Research Plan** for my review. Ensure the plan includes searching for Riva’s specific 2025/2026 product updates and competitor landscape.
This was far better than the inital 5 prompts that I won’t show here!
The full report, with some tweaking is pretty solid (I can send you a link if you want to read it):
I was pretty floored by Hubspot as a legacy system, but in the world of Agentic tools… it sort of is. The concept of objective-driven, vs prescriptive/descriptive dashboards/automation flows/etc is incredibly compelling and I’ll be exploring more in the weeks to come.
Even the infographic it created (and the website, above) were pretty damned incredible. Yeah, it took some time, and 20 years of experience, and hundreds of hours of working with Gemini for it to get my drift, but this would have been weeks of work before. Is it client-ready? No, but would it take much to get it there and could an entire year-long plan be built from this? Absolutely. And that’s pretty freaking insane.
Here’s that infographic, if you’re interested on how an org that is on “modern” (legacy still makes me laugh) tech would evolve into this world - as well as some ideation on hwo much real world work is required to build and use this infrastructure (ignoring internal training, processes, and staffing changes):
For now, I’ll keep plugging, and if you want to read Tris’s article on Deep Research to get yourself started, here ya go!