Walking the vendor floor at Enterprise Connect 2024 was like watching a parade of magic tricks—impressive demonstrations that left me wondering what was happening behind the curtain. Booth after booth proudly displayed agentic artificial intelligence solutions that promised autonomous customer service agents capable of independent reasoning, strategic planning, and adaptive problem-solving. The demos were slick, the marketing materials glossy, and the promises grandiose. But here's the rub: Most of what I witnessed wasn't agentic AI at all. It was clever chatbots embracing the newest buzzword.
In this time of unprecedented technological advancement, I have noticed something I never thought I'd see: marketing getting ahead of modern AI technology.
Before we dive deeper into my grievances, let's establish what true agentic AI means, because the term was horribly disfigured at Enterprise Connect. Forrester Research characterizes agentic software as having the following characteristics:
- Reflection – The ability of a bot to assess its performance of a task and find ways to improve.
- Planning – The ability of a bot to understand the requirements and tasks that need to be performed to meet a higher-level goal.
- Memory – The ability to maintain state to know where in a process the bot is so it can take care of business in the right order.
- Tool use – The ability to leverage external resources, such as other bots, or to make API calls to back-end systems.
- Multiagent collaboration – The ability of a core bot to orchestrate the activities of multiple bots, allowing a system to break tasks down into component pieces and have different processes own and manage the execution of various tasks.
- Autonomy – This is the result of the above, an autonomous agent that can perform complex tasks.
The Promise and the Problem
The potential benefits of true agentic AI for customer self-service are substantial and genuinely exciting. Imagine systems that could proactively identify customer frustration patterns and respond to customers' needs. For example, an AI agent might be able to provide a discount in response to a minor customer frustration.
These systems could theoretically provide 24/7 sophisticated problem resolution, handle complex multi-issue scenarios that currently require human intervention, and continuously optimize themselves for better customer outcomes. The efficiency gains and improvements in customer satisfaction would be transformational.
But, as our "ignore-the-man-behind-the-curtain" walk through Enterprise Connect taught us, we're not there yet. Several fundamental barriers prevent agentic AI from being ready for customer self-service applications:
- The Reliability Gap: Current AI systems, despite impressive capabilities, still struggle with consistency and accuracy. To be automated, any process needs to be well-defined and documented, but far too often this is not the case. When an autonomous agent makes errors, the potential for customer experience disasters multiplies exponentially compared to traditional system failures.
- The Control Paradox: Customer service requires strict adherence to brand voice, compliance requirements, and escalation protocols. True autonomy and corporate control create inherent tension that hasn't been resolved.
- The Integration Challenge: Agentic AI requires seamless integration with multiple back-end systems, real-time access to customer data, and the ability to execute actions across platforms, a problem that has haunted customer service since I was building screen scraping interactive voice response applications in the early 1990s. Tools use protocols like MCP, and agent integration options like A2A and ACP. These capabilities provide hope here, but these nascent standards for communications are a long way from settled.
When did generative AI become uncool? Here's where my curmudgeon status really kicks in. While vendors tout their agentic AI dreams, generative AI is doing its thing all over the contact center, summarizing calls, analyzing interactions, giving agent quality scores, helping agents know what to do next. What it is not doing much of is talking to customers. Despite the technology being accessible and demonstrably effective, adoption remains frustratingly slow. Most self-service implementations still rely on rule-based chatbots that frustrate customers more than they help.
For some reason, vendors faced with customers who perceive generative AI-driven chatbots as too unpredictable and unproven to put in front of their customers are responding by recommending a completely nascent technology that is even more unpredictable.
Marketing really is ahead of modern AI technology. But, taking off my curmudgeon glasses, the possibilities really are enormous.
Let's talk about what is possible when we combine the current strengths of generative AI with the future promise of agentic capabilities—a hybrid approach that could revolutionize customer self-service. Imagine a system where generative AI handles customer conversations with natural, contextual interactions that feel genuinely helpful rather than robotic. This same system could generate personalized explanations, adapt its communication style to match customer preferences, and provide detailed, accurate responses to complex queries.
Now layer on emerging agentic capabilities for strategic elements. The system could independently prioritize multiple customer issues, develop resolution strategies that consider available resources and policy constraints, and even negotiate within defined parameters to reach mutually beneficial outcomes.
The generative component ensures customers feel heard and understood, while the agentic elements provide the strategic thinking and decision-making that create truly satisfactory resolutions. Even my most curmudgeonly side gets excited thinking about the potential.
Which brings me to my fundamental frustration with the current market dynamics. Vendors are jumping straight to promoting agentic AI capabilities—technology that's still largely theoretical for customer service applications—while largely ignoring the massive untapped potential of generative AI for self-service.
The result? Disappointed customers, frustrated IT teams, and missed opportunities to leverage technology that could provide immediate value. This isn't just a technology problem, it's a strategic miscalculation that's slowing genuine progress in customer experience innovation.
So, here is my recommendation: Focus on genAI today. Your focus and energy might well be more about working with your legal team and convincing them that your new bot can be safely deployed, You might need your vendor to help. Can it explain why you should trust its guardrails and ensure the software can do what you want. It's worth it. GenAI enables better customer experiences built in a fraction of the time required with traditional approaches. Dabble with agentic AI. If you are nervous about genAI, you're going to have a lot to fear with agentic. That doesn't mean you should not have it in your lab, so you don't need to rely on articles from curmudgeons to get your insights on the software.
Max Ball is a principal analyst at Forrester Research.