What Google's AI patent defensive program reveals

What Google's AI patent defensive program reveals to us.

Seven quiet filings on Technical Disclosure Commons this year show the playbook:

1. RAG policy Q&A with citations.
2. Sentiment and sensitivity triage with inline PII scrubbing.
3. LLM briefings for bot-to-human escalation.
4. Synthesis of past cases into guidance.
5. Glossary-grounded translation.
6. Proactive intervention and data compliance.
7. Fatigue control for long-running agents.

My take:

Given the authors' background, this reads like the Salesforce and Agentforce support-agent playbook rebuilt inside Google's own People Operations. Lead author Deepkumar Raithatha is a former enterprise-Salesforce architect now leading AI solutions there. The filings sketch a reusable agentic support agent. It fits Google's wider platform push.

The filings put the mechanics in the public domain to guard against patent trolls and competitor patents, not to build a moat.

Sources:

The seven defensive publications by Deepkumar Raithatha and colleagues (Google, People Operations):

1. Agentic Framework Integrating RAG for Policy Queries and Autonomous Trend Analysis
2. Automated Service Case Triage Using Real-Time Sentiment and Sensitivity Analysis
3. Generative AI Synthesis of Automated Interactions into a Structured Agent Briefing
4. Dynamic Case Precedent: Accelerating Agent Proficiency and Resolution Consistency in Large-Scale People Operations
5. The Glossary-Grounded Universal Queue: Follow-the-Sun Support in Global People Operations
6. Agentic Trend-to-Knowledge: Automating Case Deflection through Proactive Content Surfacing
7. The Dual-Mode Privacy Guard: Augmented Compliance for Sensitive People Operations Data
8. Automatically Managing AI Fatigue Through Performance Tracking and Memory Summaries