AI in customer service hits a wall: adoption soars while orchestration lags

Artificial intelligence has become woven into nearly every customer service operation, yet the vast majority of implementations remain limited to isolated tools: a single chatbot or an assist interface next to a human representative. The genuine challenge—orchestrating data, systems and personnel to handle a customer request from start to finish—has largely gone unaddressed, and this gap is increasingly marking the boundary between organizations seeing tangible business results and those merely tracking deflected interactions.

Bridging this divide hinges not on technological sophistication but on organizational readiness: the governance structures, system integrations and operational workflows required to function as a unified whole. This concept underpins customer experience automation, or CXA, which Pedro Andrade, vice president of AI and generative AI business specialist at Talkdesk Inc., characterizes as an operating model rather than a discrete technology offering.

Customer Experience Automation, CXA, is what we defined as an operating model. It's not a category of technology. It's the system that coordinates a hybrid workforce of AI and human employees. Then connecting the systems, connecting knowledge and also connecting the workflows.

Pedro Andrade, Talkdesk Inc.

Andrade shared these insights during an exclusive conversation with Bob Laliberte, principal analyst for networking and observability, and Zeus Kerravala at The AI ROI in Contact Center Summit, broadcast through theCUBE, SiliconANGLE Media's livestreaming platform. The discussion centered on why orchestration and governance—rather than mere adoption—have become the decisive factors in determining whether AI investments in contact centers generate meaningful returns.

Orchestration is the real test for AI in customer service

Research from Talkdesk reveals that 98% of organizations have implemented AI somewhere in their customer journey, yet only 15% have integrated agentic AI with cross-departmental orchestration. The obstacles extend beyond technical limitations, with compliance concerns raised by 50% of survey participants, security worries cited by 48% and fragmented legacy systems mentioned by 45%.

Adoption is easy. The orchestration is the hardest part. Agentic orchestration requires AI to maintain a context across all those systems.

Pedro Andrade, Talkdesk Inc.

This distinction translates directly into measurable performance differences. Organizations that Talkdesk identifies as CXA leaders achieve roughly four times the net promoter score improvement compared to less mature implementations—22% versus 5%—while the gap in cost-per-contact reductions proves considerably narrower, according to Andrade. Churn prediction capabilities and personalization effectiveness demonstrate comparable disparities.

Savings alone don't tell the whole story. They understate the value that you can get with AI.

Pedro Andrade, Talkdesk Inc.

Overseeing this combination of artificial and human workers has spawned an entirely new supervisory role. Talkdesk developed its CXA Operations Center on the premise that team leads will assess, track and adjust AI agents using methods similar to those applied to human staff, transitioning from script creation to behavioral observation. The metrics used to evaluate performance evolve accordingly.

Now you don't measure just average handle time. You are going to measure how much time it takes from opening of a problem until it gets closed. Not because average handle time just measures the time of an agent, but what about the rest of the process?

Pedro Andrade, Talkdesk Inc.

The ultimate vision extends customer service operations well beyond reactive inbound support into proactive engagement—service notifications, contract renewals, debt collection—since the incoming call typically represents the final stage of a much longer customer interaction that originated elsewhere, Andrade noted. Automating a single failure point leaves the broader process unaddressed.

https://www.youtube.com/embed/gBFOWE5I_Fg?feature=oembed

Inbound is kind of the last piece of a journey is when everything breaks, people call in. Look at the whole spectrum, look at the whole journey and put your journey all on paper and now start thinking about automating the whole journey instead of just having a one-point solution that takes you nowhere.

Pedro Andrade, Talkdesk Inc.