A Strategic Bet to Win Back Market Share

The timing is strategic. The CCaaS market, valued at roughly $2.55 billion in the U.S., is expanding at an estimated 14% to 15% annual rate and could reach about $97 billion globally by the mid-2030s. With AI now the primary driver of contact-center modernization, AWS sees an opening to reset the competitive hierarchy. Amazon Connect’s financial momentum, combined with AWS’s roughly 30% share of cloud and contact-centre technologies in the U.S., aims to position the company as a formidable force.

“As third-party LLM-based tools actively work to disintermediate brands from their customers, businesses still focused on getting people off the phone quickly or eliminating interactions completely are essentially inviting others to replace their direct customer relationships,” explained DeMaio. “The organizations that thrive won’t be the ones automating away human interaction, they’ll be the ones using AI to make every human interaction more meaningful and valuable to the customer relationship.”

The upgraded platform introduces AI agents that can parse multi-intent queries, maintain memory across interactions, retrieve account information, file service requests, update orders and execute actions across enterprise systems, without human intervention. In contrast to standalone AI tools that often require months of integration work and extensive model training, Amazon Connect’s agentic AI arrives pre-built and fully customizable. It can draw on live enterprise data through new Model Context Protocol (MCP) integrations, enabling agents to fetch CRM records, check order status, access inventory platforms or query loyalty databases in real time during a conversation.

“This transforms the representative role to emphasize emotional intelligence and relationship-building rather than system navigation or data entry, which is why we’re seeing both efficiency metrics and customer satisfaction scores dramatically increase simultaneously,” said DeMaio.

AWS is also rolling out Nova Sonic, a new speech technology that gives interactions with agents a more natural, human-like voice quality, supports over 30 languages, and is built to understand interruptions. Moreover, Amazon Connect now supports speech partners such as ElevenLabs and Deepgram, catering the enterprise demand of open ecosystems rather than locked-down vendor stacks.

Agentic AI Assistance for Superhuman Productivity

While the new agentic layer introduces autonomy, AWS emphasized that human representatives remain central to the experience.

“The AI teammate model is about making humans superhuman, not replacing them,” DeMaio told me. “While the representative focuses on building rapport with the customer, the AI analyzes the conversation in real-time and actively completes administrative tasks in the background.”

One of the most persistent concerns executives raise about agentic AI is trust. Amazon Connect is attempting to address that head-on with a full observability suite that exposes how its AI agents operate—how they interpret customer requests, why they choose certain actions, which tools they invoke and what outcomes they generate. Enterprises can simulate thousands of interactions before a system goes live, evaluate agent behavior using the same scorecards applied to human representatives and track performance across every channel in real time.

“Observability transforms AI from a black box into a transparent system where you can see exactly what the AI understood, which tools it accessed, and how it reached decisions,” said DeMaio. “What makes our approach different is applying the same evaluation framework to both AI and human agents, combined with testing capabilities that validate AI behavior across thousands of scenarios before deployment.”

Amazon claims that companies deploying early versions of Connect’s agentic features report measurable impact. Centrica, one of Europe’s largest energy companies, deployed agentic assistance across its 10,000-agent environment, reducing average handle time by 38% and boosting customer satisfaction scores by 19 points. Likewise, Toyota Motor North America reportedly processed more than one million annual calls using Amazon Connect, reducing handling time by 20% and Traeger Grills consolidated five systems into one unified Amazon Connect deployment.

The Competitive Curve Bends Toward Operational AI

Most enterprises still face a widening gap between rising customer expectations and the limitations of legacy systems. Nearly one-third of customers abandon a brand after a single bad experience. AWS contends that contact centers need AI not as a feature layer, but as core infrastructure. With North America accounting for more than one-third of the global CCaaS market, AWS sees this moment as a strategic opening, especially as cloud migration, agentic AI adoption and omnichannel transformation accelerate across the region.

“The future of customer experience is continuous conversation—not isolated transactions, but ongoing relationships that build context over time and across every touchpoint,” said DeMaio. “A customer starts a conversation on your website’s chat, continues it via SMS while they’re on the go, calls in when they need more complex help, and receives proactive messages with relevant updates at every step, with full context traveling with them at every interaction.”

Alongside the new Amazon Connect capabilities, AWS also unveiled Interconnect – Multicloud, now in preview with Google Cloud, to simplify multicloud networking and offer a private, high-bandwidth link between the two providers via AWS Interconnect–Multicloud and Google’s Cross-Cloud Interconnect. Together, the providers have introduced an open specification for network interoperability with dedicated bandwidth.

Yet regaining market share will take more than speed and automation. It will require proving that AI can be trusted with the most sensitive part of the business: the customer relationship. Whether AWS has built a platform capable of earning that trust will become clearer as the CCaaS race enters its next chapter in 2026.