Workspace Studio's Ask NotebookLM grounds no-code flows

The Breakthrough

Workspace Studio introduces an 'Ask NotebookLM' step that lets no-code flows query specific NotebookLM notebooks, grounding automations in a team's uploaded documents like compliance checklists and data agreements.

What Actually Worked

• Workspace Studio flow triggers on incoming Gmail with subjects containing 'partnership', 'vendor', or 'proposal'.
• First step uses Gemini with prompt: "read the sales email below and extract the following facts I want to know the company name what are they selling" to pull structured facts from email body, keeping it grounded without web search.
• Second step selects a NotebookLM notebook (e.g., one with PDFs of 'compliance red flag checklist' and 'data processing agreement standard') and runs prompt: "you're helping merit health team review an incoming sales email below is a structured summary of the email extracted by an earlier step for each concern you flag tell us which document and section in the rule book it is a problem recommend what to do next and then also identify who specifically needs to be involved".
• Final step notifies Google Chat space with extracted facts, NotebookLM assessment (e.g., 'high-risk vendor that should be rejected'), flagged concerns with doc/section cites, recommended actions, and involved parties.

Live Demo Outcome

In the demo, a test email from 'Marcus' about a 'partnership proposal' (data processing, service improvement, DPA request) triggers the flow. Gemini extracts: company name, product claims, agreement requested. NotebookLM assesses as high-risk (e.g., data storage issues citing specific rulebook sections), recommends rejection, and posts full report to Chat instantly.

Context

Previously, Workspace Studio (Zapier-like for Google Workspace) handled triggers like new emails and AI actions with Gemini, while NotebookLM acted as a separate grounded memory for uploaded docs, slides, and videos. The integration unlocks automations with persistent, hallucination-free knowledge bases. The vendor compliance example targets repetitive manual checks (e.g., data storage, AI training on customer data) that consume hours per inquiry in midsize firms like healthcare companies; now, teams get instant Chat assessments without coding.