AI in Customer Experience: The First Thing CTOs Get Wrong

Every CX AI project runs into the same question, and most teams answer it badly: where do you actually start? Merlin Bise, CTO of Inbenta, and TSIA analyst George Humphrey rebuild a CX AI program from a blank page, sharing the mistakes they wouldn't repeat, the technologies they'd reach for first (from large-model retrieval to agentic workflows, voice, and agent-to-agent handoffs), and why deployments that should take days often stretch into quarters. You'll walk away knowing where to begin, what to fund first, and where to point your team.

Most CX AI programs stall for a reason that has nothing to do with the model: teams start in the wrong place, building an AI stack before the content underneath it is clean, structured, and governed. Deployments that should take days stretch into quarters, buried under professional-service hours and over-engineered setups that are built to answer questions real customers will never ask. Leaders who can't identify the real value drivers end up funding a bunch of pilots, none of which produce enough return to justify the next round.