Ayan Barua describes the difficulty of connecting agents to systems such as Salesforce, HubSpot, NetSuite and Oracle. Customer-specific schemas and business meanings complicate translation; a common-denominator API cannot necessarily preserve the context needed for useful actions.
Ayan Barua distinguishes MCP's tool-access protocolModel Context Protocol is a standard way for AI applications to connect to external tools and sources of context. from the broader integration problem. Data freshness, permissions and reliable writebackAn agent permission boundary defines and enforces which resources and actions an automated agent is allowed to use. still need engineering. Probabilistic actions in payroll or financial ledgers require governanceGovernance defines decision-making authority, rules and accountability within a system or organization. rather than an assumption that a connected model can act safely.
Ayan Barua also discusses opportunities in vertical industries with established legacy systemsAI adoption friction is the effort, uncertainty, disruption, or risk that makes people delay or avoid moving an AI capability into routine use.. The interview presents integration infrastructure as part of making agents dependable, not as proof that autonomous enterprise operation is already solved.
Watch on YouTube




