The contact center industry has invested heavily in AI capability but remains fundamentally unprepared to deploy it effectively. Across the sector, a consistent pattern has emerged: 79% of service professionals are investing in agentic AI, yet 55% of agents report receiving no training whatsoever. This gap between deployment ambition and workforce readiness manifests most acutely on the calls that matter most—high-stakes retention conversations, upsells, and complex escalations where agents face immediate, customer-facing consequences for underperformance. Cresta's launch of its Training Simulator addresses one dimension of this problem by creating realistic, adaptive practice environments grounded in real conversation data, allowing agents to develop competency before handling live interactions. Yet the training infrastructure gap is only symptomatic of a deeper architectural problem that extends across the entire contact center stack.
The more consequential issue, as Omilia's Claudio Rodrigues articulates, is that most contact centers operate on fragmented technology stacks where customer data, call recordings, workforce optimization tools, and knowledge bases exist in isolation. This fragmentation prevents the intelligence loop from closing—the system cannot simultaneously listen to both agent and customer, structure that information in a single place, and learn continuously from it. The result is that contact centers optimise for cost reduction rather than revenue generation, missing recurring interaction types that could be automated with minimal effort if the infrastructure existed to support self-learning systems. For CX leaders already running Zendesk, Salesforce, or similar platforms, this raises an uncomfortable question: are your current integrations actually creating a unified intelligence layer, or merely automating existing workflows within silos?
The organisations pulling ahead are making a deliberate mindset shift. They define clear business outcomes before selecting technology, treat the contact center as a revenue interface rather than a cost centre, and position AI as augmentation for human agents rather than replacement. This requires both technical architecture—connected data flowing through a single system—and organisational will to treat agent upskilling as continuous performance management rather than one-time onboarding. The compounding gains come not from single large automations but from hundreds of small, recurring cases becoming viable when the system handles the heavy lifting automatically. For teams currently managing fragmented stacks, the strategic question is whether incremental improvements within existing silos will ever generate the ROI your board expects, or whether a fundamental rearchitecture toward unified intelligence is now table stakes.
Contact center AI adoption is accelerating. Investment is up, vendor capabilities are maturing, and the business case for automation and agent assist has never been clearer. And yet deployment failures remain stubbornly common. A common weakness is starting to emerge in contact center AI depl
You’d be surprised at just how many CEOs are picking up the phone and calling their own contact centers. Not to check on staffing levels or review a transcript; just to experience it as a customer would. There is a pleasure in discovering that, despite being surrounded by and well-informed about