jesse zhang

Customer service offers quantifiable ROI (deflection rates), natural escalation paths for risk mitigation, and existing infrastructure — making it the fastest enterprise adoption vector.

Agents will simultaneously augment highest-paid knowledge workers (coding) and replace lowest-cost repetitive labor (customer service), driven by ROI clarity at both extremes.

Conversational agents will unify sales, support, and commerce into a single persistent interface with full user context, becoming the primary brand touchpoint.

Voice is the natural human UI, but voice-to-voice models currently suffer ~8x higher hallucination rates than text, requiring hybrid architectures for enterprise deployment.

LLMs can ingest 100% of customer interactions to auto-generate evals, detect gaps, and improve agents continuously — creating a data advantage that compounds with tenure.

Model inference costs drop exponentially; application-layer companies should prioritize market share and product quality over unit economics today.

Solving the end-user business problem allows application companies to command pricing power, while model providers face low switching costs and intense competition.

Board-level pressure drives AI initiatives; vendors must articulate clear cost savings or revenue uplift in a single sentence to get prioritized.