Artificial intelligence in the contact center is moving beyond basic call routing, frequently asked questions, and simple self-service. The next phase centers on AI agents capable of completing multistep tasks, supporting outbound engagement, collaborating with human agents, and producing measurable business outcomes.
During the CX Summit, Zeus Kerravala, Principal Analyst at ZK Research, and I had the opportunity to speak with Ram Rajagopalan, Head of Product AI for Zoom CX, and Joe Rittenhouse, Co-CEO of Zoom partner Converged Technology Professionals. We discussed what this evolution means for enterprises. Their message was clear: The technology is becoming increasingly capable, but successful adoption depends on selecting the right use cases, connecting trusted data, establishing governance, and measuring results against the full cost of deployment.
See below for the full interview.
Moving from containment to resolution
Historically, contact center automation has often been evaluated by containment, the percentage of interactions handled without reaching a human agent. While containment remains useful, it doesn’t show whether the customer’s problem was actually solved.
Rajagopalan said enterprises are beginning to adopt a more meaningful standard. “We are going from not just looking at containment to end-to-end resolution of consumer inquiries. This is what we are calling within Zoom ‘Conversation to Completion.’” Why does this matter? A customer abandoning an automated interaction may technically count as containment, but it should not be considered a successful outcome if the customer leaves frustrated or without an answer.
Zoom is addressing this through explicit and implied resolution measurements. Explicit resolution can be captured through a post-interaction survey. Implied resolution uses a separate large language model to evaluate the conversation and determine whether the interaction appears to have achieved its intended outcome. These signals can then be combined with customer satisfaction and sentiment data.
This reflects a broader industry shift from measuring activity to measuring value. Contact centers will continue to track call volume, average handle time, deflection, and cost per interaction. However, enterprises should increasingly connect those operational metrics to resolution rates, revenue generation, customer retention, and overall experience.
More sophisticated use cases are becoming practical
AI agents are expanding into workflows that extend beyond inbound customer support. Rajagopalan cited outbound political surveys, healthcare appointment reminders, prescription-refill notifications, and interactions that require navigating complex interactive voice response systems.
These workflows may involve qualifying a customer, responding dynamically based on previous answers, updating a system of record, detecting voicemail, or completing a business transaction. That requires greater integration with customer relationship management, ticketing, scheduling, and industry-specific applications.
As Rajagopalan explained: “We are not just looking at answering the basic inquiries [or] deflecting the call from going to a human agent but actually completing the task that the consumers or users are calling in for support.”
Zoom CX provides nearly 40 out-of-the-box integrations, including connections to platforms such as Salesforce and Microsoft Dynamics, as well as support for custom scripting. The strategic value lies less in the number of connectors than in the ability to preserve context across systems and throughout the customer journey.
If customers must repeat their identity, history, and reason for calling every time an interaction moves between a virtual agent and a person, the automation may reduce costs without improving the experience. A connected platform should transfer the conversation history, collected information, customer context, and recommended next steps to the human agent.
Trusted knowledge is a business requirement
AI performance is directly tied to the quality of the information available to it. Outdated knowledge bases, conflicting policies, and undocumented employee expertise can limit even the most advanced virtual agent.
Rajagopalan described an enterprise where human agents routinely relied on knowledge gained through experience rather than information documented in the company’s knowledge base. By analyzing conversations after escalation, the organization could identify those gaps, update its knowledge, and improve the virtual agent. “That feedback loop is very critical. The virtual agent and human agent are always talking to each other and learning from each other.”
For business leaders, this highlights an easily overlooked element of AI readiness: Knowledge management must become a continuous operational discipline. Organizations need clear ownership for validating content, resolving conflicts, tracking changes, and determining what information AI systems are authorized to use.
Start with a focused opportunity
While the technology can support increasingly ambitious use cases, Rittenhouse recommended beginning with a clearly defined, lower-risk opportunity. After-hours service is one example because it can expose unmet demand without disrupting daytime operations.
He described a physical rehabilitation provider that was missing calls from patients discharged from emergency rooms at night. If those patients could not schedule an appointment, they frequently contacted a competitor the following morning. Adding automated after-hours engagement enabled the provider to capture appointments and generate millions of dollars in revenue.
Rittenhouse said a focused implementation of this type can begin producing data and results relatively quickly. “Roughly, start to finish, that’s 30 days where you’re going to start seeing containment, start seeing data, and start seeing results.”
The objective is not simply to deliver a quick technical win. An initial deployment gives the organization an opportunity to learn how teams must collaborate, which integrations are required, how customers respond, and what ongoing operating model will be needed. “We can do all those things, but we have to prioritize and we have to have a plan.”
Governance, cost control, and continuous improvement
Rittenhouse emphasized that contact center AI is not solely an IT or CX initiative. Implementations may involve customer service, operations, security, compliance, finance, data management, and executive leadership. Organizations therefore need an AI committee or comparable governance structure to prioritize projects, define responsibilities, approve metrics, and oversee risk.
He also identified dedicated project management as an emerging indicator of success. “The organizations that we’re seeing that are really starting to thrive have a PM-based model now that reports to the C-suite and is the choreographer of the business.”
Financial governance is equally important. Consumption-based, metered, conversational, and outcome-based pricing can produce very different economics as volumes increase. Enterprises should model expected usage, define what constitutes a billable outcome, and monitor the full cost per successful resolution. A project that improves containment but creates unpredictable operating expenses may fail to deliver its intended return.
Testing must also continue after deployment. Zoom is applying LLM-based simulation to test virtual agents across different scenarios, accents, and background conditions. Combined with scorecards and A/B testing, this can help organizations compare agent versions and detect performance or knowledge gaps before directing more production traffic to them.
Why it matters
Contact center AI is progressing from isolated automation to a connected operating model in which virtual and human agents share context, knowledge, and tools. The business opportunity includes more responsive service, higher employee productivity, expanded customer access, new revenue capture, and lower costs, but none of those outcomes are automatic.
The enterprises most likely to succeed will start with a focused business problem, establish baseline metrics, calculate the complete cost of operation, and expand only as evidence supports doing so. Technology provides the capability, but disciplined execution, reliable knowledge, cross-functional governance, and continuous measurement will ultimately determine whether AI experimentation becomes sustainable business value.
