As companies rethink how they provide customer experiences (CX), a new form of AI capability, agentic AI, is quickly changing how work is accomplished in contact centres.

In the recent episode of the Tech Transformed podcast, Dialpad Lead Product Manager Calvin Hohener sits down with host Jon Arnold, Principal at J Arnold & Associates. They discuss the transition from legacy chatbots to more autonomous agents capable of completing tasks and improving customer interactions.

The conversation highlights the importance of understanding the technology's impact on enterprise architecture, the need for clean data, and the strategic implications for C-level executives. Hohener emphasises the importance of starting with clear use cases and working closely with vendors to maximise the potential of AI in business operations.

From Legacy Chatbots to Agentic AI

Most people have used chatbots and found them lacking. Hohener explains why: earlier conversational AI was based on retrieval-augmented generation (RAG). These systems could take user input, search a knowledge base or the internet, and provide an answer. This was helpful for customer service queries, but limited.

“Previous AI models could retrieve and return information, but now we’re moving into a new phase with agentic AI.” Agentic AI can take action rather than just providing information.

For AI agents to succeed, organisations must first organise their data. “How your internal knowledge is structured is crucial. Even if the data is unorganised, you need to know its location and ensure it’s clean,” stated Hohener.

Agentic systems depend on internal knowledge, including knowledge base articles, CRM notes, and process documentation. If this foundation is disordered, the agent’s output will not be reliable.

This isn’t about achieving ideal data cleanliness from the start; it’s about knowing what information exists, where it is, and whether it can be trusted. If an AI agent bases its decisions on outdated, conflicting, or incomplete content, it will struggle to perform tasks aptly, regardless of how sophisticated the model is. Enterprises need at least basic clarity about which systems hold which knowledge, who is responsible for them, and whether there is consistency across sources.

Hohener noted that organisations often overlook how quickly conflicting information can undermine an otherwise well-designed agent. A single outdated procedure or mismatched policy in a knowledge repository can lead an AI to produce incorrect results or halt during workflow execution.

Keeping internal content clean, deduplicated, and consistent gives the agent a reliable, valid source. This reliability becomes crucial when AI starts taking meaningful actions, not just providing answers.

By focusing on data readiness early, enterprises not only reduce deployment obstacles but also set the stage for scaling agentic AI across more complex processes. In many ways, preparing data isn’t just a technical task; it’s an organisational one.

How Human Agents Work with AI Agents?

The Dialpad Lead Product Manager noted that human roles, too, will evolve with agentic AI entering the contact centre. For instance, human agents will take on more of an advisory role—reviewing conversation traces and helping adjust the models.”

Instead of just resolving customer issues, they will help refine and oversee AI workflows. This includes reviewing conversation logs, noting where an AI agent may have misinterpreted intent, and providing feedback to improve the models.

This mentor-mentee relationship between human and digital agents becomes crucial as organisations increase automation. Human agents bring domain knowledge, contextual judgment, and the ability to handle unique situations, all of which help the AI improve over time. In return, digital agents reduce the repetitive workload and allow humans to engage in higher-level thinking.

Takeaways

  • The time is now for adopting agentic AI solutions.
  • Agentic AI represents a significant shift in customer experience technology.
  • Legacy chatbots have limitations that agentic AI can overcome.
  • Real-world applications of agentic AI include mundane tasks like scheduling and verification.
  • Quantifying time savings is crucial for measuring ROI with AI.
  • AI agents can improve customer satisfaction metrics when implemented correctly.
  • Data organisation is essential for effective AI deployment.
  • Human agents will focus on more complex cases as AI handles routine tasks.
  • C-level executives should view AI as a strategic investment.
  • Collaboration with reputable vendors is key to successful AI implementation.

Chapters

  • 00:00 Introduction to Agentic AI
  • 05:10 Evolution Beyond Legacy Chatbots
  • 11:44 Real-World Applications of Agentic AI
  • 15:38 Impact on Enterprise Architecture
  • 21:41 Strategic Considerations for C-Level Executives