Most enterprise AI investments follow the same arc. Promising pilot. Enthusiastic board presentation. Careful expansion plan. Then: the gap. Between what worked in the sandbox and what the production environment actually demands. Between what the model can do and what the organization is ready to receive. That gap is where AI budgets go to stall. Travelport decided to close it differently. Not by running another pilot. By rebuilding the infrastructure underneath — deploying Claude inside the engineering and delivery process, building MCP architecture that connects traveler intent directly to confirmed bookings, with customer-facing capabilities reaching market this year. The difference wasn't the technology. The technology was available to everyone. The difference was having a partner with the industry depth, engineering muscle and operational accountability to take it from capable to productive. That's the gap most enterprises are still trying to cross. Travelport just crossed it. Read on: https://cgnz.at/6043vZ8Jl #AIBuilder
Cognizant This is the right gap to target, but there is an even harder one underneath it. Connecting AI reasoning and planning to systems that can actually transact is a major enterprise step. But once traveler intent can move directly into confirmed bookings, exchanges, refunds, servicing, and disruption workflows, the problem is no longer only: can the AI understand the request? or can the platform fulfill it? The real question becomes: who governs the action? Because in travel, an AI mistake is not just a bad answer. It can become a booking error, a wrong refund, a missed connection, a policy violation, a customer trust failure, or a real-world operational consequence. MCP can connect agents to systems and data. Claude can reason across codebases and workflows. Cognizant can integrate at enterprise scale. Travelport can connect the transaction layer. But that still leaves the hardest layer: governed execution.
This is a very accurate description of the core enterprise AI problem: most companies overestimate the intelligence layer and underestimate the organizational readiness layer. Between “the model can do it” and “the system can safely operationalize it” sits a massive layer of processes, ownership, governance, and operational friction.
"Company values never match reality." Obviously! 😑 Otherwise, we'd all be walking around smiling, holding hands with unicorns and angels 🤨 .
What makes this especially relevant is the shift from isolated AI capability toward operational integration. Many organisations can already demonstrate what AI is capable of inside controlled environments. The more difficult step begins once these systems become embedded into real workflows, responsibilities and decision structures across the enterprise. That is often where entirely new forms of coordination, dependency and organisational complexity start to emerge.