The bottleneck constraining AI agent deployment in enterprise CX environments is not model capability but infrastructure architecture. At VB Transform 2026, representatives from LinkedIn, Walmart, and Zendesk converged on this diagnosis: legacy systems designed for human-speed workflows cannot accommodate the millisecond decision cycles that autonomous agents require. This distinction matters because it reframes the investment conversation entirely. Teams currently evaluating agent-native platforms versus retrofitting existing stacks face a critical question: is your infrastructure debt worth carrying forward, or does agent adoption demand architectural rethinking from the ground up? The implication is stark—organisations cannot simply bolt AI agents onto twenty-year-old ticketing systems and expect performance parity with purpose-built alternatives.
The infrastructure gap manifests across three dimensions that directly affect CX operations. First, data retrieval latency: legacy databases and knowledge management systems were optimised for human query patterns, not the rapid context-gathering cycles agents need. Second, permission and security models operate at human timescales, creating friction when agents must make decisions in real time. Third, integration layers between systems—ticketing, knowledge bases, customer data platforms—introduce cumulative delays that compound across agent workflows. For teams already running Zendesk or similar platforms, this suggests the real competitive pressure comes not from model improvements but from infrastructure vendors who can eliminate these latency points. Smaller CX platforms without the engineering resources to rebuild their data layers face genuine risk of obsolescence.
The practical consequence is that CX leaders must now evaluate infrastructure readiness as a prerequisite to agent adoption, not an afterthought. Teams should audit their current stack for millisecond-scale bottlenecks: database query times, API response patterns, and permission-checking workflows. The organisations that close this gap fastest will capture disproportionate value from agent investments, whilst those that delay infrastructure modernisation will find their agents constrained by the very systems they're meant to augment.
Legacy infrastructure, not the models themselves, is what's actually slowing AI agents down. That was the shared conclusion of three infrastructure leaders — from LinkedIn, Walmart, and Zendesk — at VB Transform 2026.The panel brought together Animesh Singh, senior director of AI platform and i