Perplexity's hybrid compute architecture for its Computer platform represents a meaningful shift in how agentic AI systems handle sensitive data. By splitting processing between cloud-based frontier models and locally-running open-weight models on Apple silicon, the system routes confidential information to on-device execution, eliminating the need to transmit sensitive customer data to external servers. This addresses a fundamental tension in modern CX operations: teams need the reasoning power of advanced AI models to handle complex customer interactions, yet compliance frameworks, data residency requirements, and customer trust increasingly demand that sensitive information never leave the organisation's infrastructure.
For CX teams already managing customer data through platforms like Zendesk or Salesforce, this development creates both opportunity and complexity. The hybrid approach means you could theoretically deploy agentic systems that handle PII, payment information, or regulated data locally whilst still leveraging frontier model capabilities for nuanced customer reasoning—without the architectural compromises that have forced teams to either accept cloud processing or settle for weaker local models. However, this raises a critical question: how many CX platforms will actually integrate this capability, and will smaller vendors follow suit, or does this advantage accrue primarily to organisations with the technical depth to manage hybrid deployments? The broader implication is that data sovereignty is becoming a competitive differentiator in the AI tooling space, and teams should begin evaluating whether their current vendor roadmaps address this constraint.
The timing matters. As customer resistance to AI remains a documented concern and organisations grapple with AI governance, the ability to credibly state that sensitive customer data never touches external infrastructure becomes a material advantage in both internal adoption and customer-facing transparency. For CX leaders, this signals that the next generation of agent evaluation criteria should include data residency architecture, not just accuracy metrics.
Perplexity today launched hybrid compute for its agentic platform, Computer, a system that lets a single AI agent split its work between frontier models running in the cloud and smaller open-weight models running locally on Apple silicon Macs — routing sensitive data to the local machine so it never