Eight by Eight's product leadership has crystallised a tension that CX teams are now confronting in production: the build-versus-buy decision for AI agents is no longer a theoretical exercise, but a practical reckoning with how custom workflows actually behave at scale. Dhwani Soni argues that the shift towards custom-built agents stems from two fundamental gaps in off-the-shelf solutions. First, most organisations operate with domain-specific terminology, escalation logic, integrations, and regulatory constraints that generic platforms cannot accommodate without extensive customisation. Second, and more critically, bolting AI onto existing contact centre infrastructure creates compounding technical debt—transcription intermediaries, API latency, context loss during transfers—that manifests as customer experience failures when customers must repeat information or agents cannot access conversation history. This architectural problem is not merely a technical inconvenience; it is the failure mode that separates a convincing demo from a production system that actually works.
The implications for CX leaders are substantial. Teams already committed to native AI architectures within platforms like Salesforce Agentforce or purpose-built solutions may have a structural advantage over those retrofitting AI onto legacy systems, but the real competitive pressure now sits elsewhere: in the ability to move from pilot to production without treating AI as a magic wand. Soni's practical framework—start with a workflow you understand, define KPIs upfront, map exceptions and escalations, design human handoff from day one—suggests that success depends less on choosing the right vendor and more on choosing the right starting point. The expansion phase she describes, where early adopters are moving from inbound reactive use cases into proactive outbound workflows like appointment scheduling and follow-ups, indicates that the market is maturing beyond chatbot novelty. Yet this raises a sharper question: as teams move into more complex, outbound-facing agentic workflows, will the build-versus-buy calculus shift again, and should smaller vendors without deep platform integration capabilities be concerned about being squeezed out?
As more CX teams deploy AI agents, the build-versus-buy question is no longer theoretical. Leaders are feeling real pressure to move beyond pilots and prove impact, but many are also learning that an agent that looks great in a demo can fall apart fast in production. In this CX Today interview, Dhwa