I am struck by the blizzard of software announcements this year describing new features for CX tools that are “agentic,” meaning autonomous tools take actions without (much) human intervention. Industry conversation about agentic AI has proceeded ahead of clarifying definitions and a sense of how it fits on a continuum of rapid AI development.
I’m not sure that everything currently described as such in the marketplace actually fits a formal definition of “agentic,” but that’s a side issue. Instead, what’s in front of us is the pivot moment, where AI tools are transitioning from being advanced information retrieval systems into being action-taking systems. In the contact center this might seem obvious, because that unit has long used now-primitive methods of automating processes with rules-based modes or human/machine hybrids. So the transition seems a natural and welcome one. That doesn’t mean it is not disruptive.
For an enterprise to make sense of the agentic tools on offer, those tools must be tied closely to defined use cases. Those cases, in turn, must be backed up by evidence that the process being changed is clearly better with agentic tools than what was already in place. There are different perspectives to be brought to bear on that question; the most useful of them focus on the triad of improvement, transformation and disruption.
Looked at through the lens of “how does it improve things,” there are three contexts in which to consider agentic AI’s value. One is direct contact with customers, where we are asking automated systems to take over the conversation, contain it within self-service, and—this is the key—activate certain processes like transactions, product returns, renewal offers or requests. What makes a tool “agentic” is its ability to perform, rather than just converse. In this context, vendors are citing desirable outcomes like the ability to invoke greater personalization in interactions and to do more targeted proactive outreach, all in service of interactions that are more complete but also less expensive than human-led ones.
Another context is how agentic tools impact employee behavior and performance. Here, I have seen software providers pivot some of the earlier generative AI use cases to describe them as agentic, though I am not sure that represents something more than a narrow evolution of the tools. For example, real-time agent assist has been described this way, along with real-time coaching. The implied improvement in the transition to agentic is that instead of just retrieving relevant information for the human rep, an agentic system is also looking at broader factors like real-time customer sentiment to create suggestions or coaching that is more relevant than before.
There is also good evidence that automated quality management, wherein all or most interactions are evaluated by AI, can have a positive impact on agent morale, behavior and
performance.
I believe that the third context for evaluating agentic tools is the most interesting and potentially transformative for contact centers. Unfortunately, it is also the one where it’s hardest to measure the benefits of the output. That is in how agentic processes knit the
contact center more closely into the activities of other parts of the enterprise, even beyond
the back office. For example, systems that automatically invoke workflows in marketing or sales that affect the customer’s experience without direct involvement of the contact center, but that are based on information gathered during contact center interactions. We’re talking about efforts to recover lost customers or to prevent churn; to create hyper-personalized offers and deliver the outreach at scale; or to coordinate complex delivery of services like field repairs that are often where business processes break down. Indeed, ISG Research asserts that by 2028, software for managing customer experiences will focus on automating cross-departmental processes that touch customers.
These kinds of actions have always been possible, but they are so complex (from a human point of view) that they are rarely documented, repeatable or measured effectively. They are the kinds of exceptions to the norm that make them stand out. The agents or teams that build and implement them are sometimes the ones identified as performance stars that earn accolades.
So with agentic tools, you do have the ability to take best practices that are strikingly good, but rarely repeatable, and make them into standard actions regardless of whether agents have the skill or temperament to act on them. This is the underlying value to platforms and tools that create workflows across teams. Will the contact center get credit for the revenue boosts or experience improvements that come from them? That’s uncertain. It’s the responsibility of contact center leadership to articulate how cross-departmental processes and workflows rely on the expertise of the center and would fail without it.
One other argument that I hear often from software providers about agentic AI tools is that they make it much easier for non-analysts to extract meaningful information from complex pools of data, especially customer data. And this, I think, is where agentic AI (and its successors) will have the most profound long-term impact. Analytics in contact centers usually focus directly on performance improvement because that’s the most important set of KPIs that drive center activities. It is hard for people making operational decisions to see past the first level of analysis (Who is performing better than whom, and why, and what changes can we make to improve that?), to more sophisticated questions like What are the most common call drivers, and why? When a line-of-business person who has limited BI experience owns complex problems, the ability to ask intricate questions using natural language can be revolutionary. Could they do it without agentic AI? Yes, certainly, but only with time and assistance. Agentic AI for analysis brings the people with the problems closer to the solutions without intermediaries or the need for data expertise.
If you are an enterprise considering the prospect of agentic tools, I recommend taking a hard look at those three contexts and making a clear-eyed assessment of whether the improvement will be incremental or transformative. If it is transformative, ask whether your organization has the stomach for the internal disruption it will cause. And look closely at the third context: using agentic AI platforms with AI-oriented agents to build autonomous interdepartmental workflows. It’s there that the tradeoff between transformation and disruption has the potential to change the entire dynamic of customer-company relationships for the better.
Regards,
Keith Dawson
Fill out the form to continue reading.