Digital-native startups are abandoning traditional relational databases in favour of flexible, schema-agnostic architectures designed to support agentic AI systems. The core issue is architectural drag—the friction between what modern AI agents require operationally and what legacy data infrastructure can deliver. Rigid schemas, fixed data structures, and row-column databases create bottlenecks when agents need to process variable inputs, embed vectors, retrieve context dynamically, and iterate rapidly without schema migrations. This shift reflects a fundamental recognition that the data layer must evolve before agent capabilities can scale reliably. For CX teams already invested in Zendesk, Salesforce, or similar platforms, this raises an immediate question: are these incumbents architected to support true agentic workflows, or will they become constrained by their own database foundations as agents become central to customer interactions?
The implications for CX operations are substantial. Teams deploying agents for customer service—whether for routing, resolution, or escalation—depend on systems that can handle unpredictable query patterns, maintain context across multiple data sources, and adapt without requiring engineering intervention. Traditional CRM and support platforms were built for structured transactions: tickets, contacts, interactions. Agentic systems demand something different: they need to reason across unstructured data, maintain probabilistic confidence scores, and operate across multiple knowledge domains simultaneously. This architectural mismatch means that CX leaders implementing agents may find themselves constrained by their platform's data layer long before they hit limits on the AI model itself. The real competitive advantage will accrue to vendors—whether new entrants or incumbents willing to rebuild—that decouple their data infrastructure from their application layer.
The broader pattern emerging across Yorkshire Building Society's agent deployment and similar implementations suggests that CX teams will increasingly need to evaluate their tech stack not just on current functionality but on architectural flexibility. This is particularly acute for mid-market and enterprise teams: staying with a monolithic platform for comfort and integration may mean accepting architectural drag that limits agent sophistication, whilst migrating to purpose-built agentic infrastructure introduces operational complexity and vendor risk. The question is not whether to adopt agents—that's settled—but whether your current platform can evolve fast enough to support them without becoming a liability.
Presented by MongoDBThe gap between what AI models and agents can produce and what legacy infrastructure can reliably support is known as architectural drag, and it is the defining bottleneck of the agentic era. The data layer underneath an agentic system must handle variable schemas, vector embeddi