Agentic AI is the next wave of artificial intelligence where agents can autonomously complete multi-step tasks to achieve a defined goal. Unlike traditional AI, which relies on a predefined set of rules, agentic AI can reason, plan, and adapt its actions without constant human intervention.
This shifts the role of the human from being a step-by-step instruction giver as an operator to a supervisory capable of setting and monitoring objectives. The role of breaking down the problem, executing the correct action, and even self-correction within the course remains that of the AI agent. While this is the promise of greater efficiency, human supervision remains essential in the upkeeping of ethical guidelines, verification of output, and in setting strategic decisions.
Even though agentic is of tremendous interest across almost all organizations, the issue of how and where to begin is the subject of debate in almost any meeting I have with IT and business leaders. To get a systems integrator (SI) perspective, I chatted with Jamie Timm, Global SVP of Service Delivery and Operations for TELUS Digital. SIs can be particularly valuable with emerging technology as they get a first-hand look at many early deployments.
Timm stated that the early, “low hanging fruit” for agentic AI is within call centers. She told me, “We are seeing a lot of activity with agentic in the back office. Agentic can be particularly useful to complete concurrent tasks or follow-up after a customer support call.”
This is consistent with my research. One of the underappreciated aspects of contact centers is they have very well-defined metrics as every conversation and transaction is measured. Contact centers also tend to have a high churn rate, with some businesses reporting their employee turnover in this area is over 50% annually. Using agentic agents to temporarily fill the gaps created by agent churn can be a boon to contact centers as it maintains a more consistent, high-quality service.
The back-office aspects are particularly interesting as many simple tasks such as creating customer call summaries and logging interactions can be done more fulsomely with agentic agents. For example, agentic agents can take detailed notes and summarize them much more accurately, quickly and efficiently than humans. This can lead to better data for the AI to work with, thereby creating a self-feeding loop that trains the AI to allow for continuous improvement.
Timm brought up another use case which was using agentic AI to revolutionize fraud detection and authentication. AI agents monitor a vast range of data points—from transaction metadata and user behavior to device fingerprints—to build a profile of “normal” activity. When a transaction or login attempt deviates from this established profile, the AI can autonomously initiate a response. This might involve blocking a suspicious transaction, escalating a case to a human analyst with a detailed summary of the risk, or triggering a secondary, dynamic authentication step.
Unlike the rule-based system of the past, agentic AI utilizes an iterative learning cycle that gets better with time and reduces false positives as the system gets more seasoned. The proactive, adaptive approach is critical in the presence of more sophisticated, AI-assisted fraud that can match human patterns and utilize multiple channels simultaneously. Some would characterize this as “fight fire with fire” because the bad guys are using agentic AI as part of making the fraud more undetectable. Businesses need agentic AI to identify those suspicious activities as today, it’s akin to finding a needle in a haystack.
Timm and I also tackled the highly sensitive topic of skills and staffing. There are currently significant discussions around the potential impacts of agentic agents on employment. While I believe agentic will eliminate some traditional agent jobs, I also believe that over time, it will create more opportunities. Every major technology innovation we have had has done three things – eliminated jobs, raised productivity and created new jobs, and agentic AI will be the same.
The reshaping is built around the core redistribution of work complexity. The agentic AI will streamline operations by effectively managing well-defined, narrow-scope intents and standard customer queries, enabling human agents to concentrate on more complex calls that demand subtle judgment, emotional intelligence, and advanced problem-resolution capabilities. The repositioning lifts the responsibility of human agents from the role of task performers to the role of critical thinkers who can handle new and undefined situations that the AI is unable to.
Timm confirmed my thoughts when she stated, “Looking ahead, agents will need an evolved skill profile. We’re moving beyond routine tasks to focus on call complexity and agents making sophisticated decisions and handling more challenging interactions. As this will eventually be driven by AI, the critical skills become knowing how to work with AI effectively and excelling at handling those difficult calls. CX leaders must prioritize training their teams for this evolution toward higher-complexity, higher-value interactions.”
Another topic we discussed was how to measure contact center effectiveness in the agentic era. Contact centers have used metrics such as average handle time (AHT) as a core key performance indicator. In the agentic world, customers should be able to solve more problems on a single call and agents will have the opportunity to upsell or take other actions to add to the experience. This will likely lead to an increase in AHT, which is counter to the way most contact centers operate today.
I asked Timm what she’s recommending to her clients, and she mentioned customer satisfaction (CSAT) and other metrics that indicate value delivered. Examples are retention metrics, ARPU and lifetime value. She noted, “I expect AHT to eventually go away and for contact centers to focus more on value generation. Call centers have historically been cost centers, but agentic AI is changing that and the way we measure this area needs to evolve with it.” This fundamental shift is also driving changes in how organizations contract with BPO providers, with Timm seeing an evolution toward more managed services contracts that include technology transformation components to achieve these value-driven outcomes, all facilitated by AI capabilities.
We rounded out the conversation on interoperability and the ability to orchestrate conversations between agents. If the agents are unable to communicate with one another, companies will have the problem of the agentic agents working in bubbles, and this may result in conflicting information being conveyed or propagated. An agent may take some action, and then another may reverse that, with the result being customer irritation and adverse effects on the overall customer experience. The issue spotlights the necessity of more effective communications protocols among the human and agentic agents as well as the importance of ongoing innovation in enhancing the multi-agent systems. On the horizon, agent-to-agent (A2A) and model context protocol (MCP) represent promising standards, but there is little chance that they will become widely adopted in the near future.
Agentic AI is quickly gaining traction and is paradigm-shifting in how we live and work. While Timm supplemented that schedules will differ depending on organizational risk appetite, the eventual penetration across all sectors of the technology is unstoppable. For decision makers, the question isn’t whether to adopt agentic AI, but where to begin. Contact centers offer an ideal starting point with immediate, tangible results, but successful implementation requires equal attention to workforce reskilling and development.