top of page

IVR to AI: Gleaning Insights from a Half-Century Journey in Customer Service Automation

While it is difficult to pinpoint an exact date for the birth of Interactive Voice Response (IVR) systems, I'll recognize IVRs as turning 50 this year. The technology began to emerge in early commercial applications in the early 1970s.

If you research the emergence of IVRs, you'll see several references to an order entry system developed by Steve Schmidt in 1973. Additional use cases began in banking in the 1970s, and American Airlines unveiled an IVR system in 1977, permitting customers to examine flight details and reserve seats via touch-tone inputs.

The use of the technology expanded in the 1980s and 1990s as companies began integrating these systems with computers. The IVR grew from collecting data and feeding back responses, finding a new role with computer telephony integration (CTI), and opening up options for intelligent routing of voice traffic.

Today, we all recognize IVR systems enabling us to self-serve by interacting with a computerized system before being connected to a live agent. Without human assistance, we navigate through menus and perform basic tasks, such as checking account balances, service outages, or inquiring about an order status. One of my first applications of enhancing an IVR was for pre-paid landline phone service at NOW Communications, automating the payment function. Straight-forward and transactional inquiries were a strong use case for how IVR technology created value for our call center employees, customers, and shareholders.

IVRs: The Good, Bad, and the Ugly

Research is extensive on how the introduction of IVR systems in call centers has brought benefits to customer service:

  1. It has improved efficiency by reducing the need for live agents to handle routine and repetitive tasks. Customers can quickly access the information without waiting for a human agent, leading to shorter wait times and faster service delivery (Tezcan & Behzad, 2012).

  2. IVR systems have increased accessibility by providing 24/7 service availability. Customers can interact with the system anytime, allowing for greater convenience and flexibility (Colladon et al., 2013).

  3. Self-service channels like IVR systems have empowered customers by giving them more control over their interactions with the call center.

Yet, if you ask anyone what they think about "IVRs," you will likely hear a negative response. IVR systems have become synonymous with customer frustration and dissatisfaction, supported by research:

  1. IVR systems can be complex and challenging to navigate, leading to confusion and frustration for some customers (Dean, 2008).

  2. The lack of human interaction in self-service channels can result in a less personalized and empathetic customer experience (Ellway, 2016).

  3. Some customers may prefer speaking to a live agent for more complex or sensitive inquiries, and the absence of this option in IVR systems can lead to dissatisfaction (Tezcan & Behzad, 2012).

While customers may find some incremental self-service convenience with AI-powered chatbots, the degree to which this further removes agent-assisted interactions from customer service organizations has its limits. The challenges faced with IVR systems underscore the necessity for a well-considered approach as we venture into the next frontier of customer service automation with AI. These can be expressed in a model customized to consider your business landscape. It could be a decision tree with two tiers:

Tier 1: Preliminary Evaluation These gates serve as an initial filter to quickly determine if a transaction type is even worth considering for AI automation:

1. Transaction Complexity:

  • Low: Proceed to Tier 2.

  • Moderate to High: Requires further evaluation in Tier 2.

2. Emotion Check:

  • Low Emotional Intensity: Proceed to Tier 2.

  • High Emotional Intensity: Likely direct to a human agent (with an option for re-evaluation in Tier 2).

3. Security Check:

  • Low Security Risk: Proceed to Tier 2.

  • High Security Risk: Likely direct to human agent (with option for re-evaluation in Tier 2).

Tier 2: Detailed Evaluation If a transaction passes the preliminary gates, it may undergo a more nuanced evaluation to decide on automation. This tier may include many more variables, depending on the business attributes:

  1. Customer Preference: Assess through previous interaction data or direct inquiry.

  2. Time Sensitivity: Evaluate based on the urgency and potential impact of delays.

  3. Cost of Service: Analyze the cost-effectiveness of automation versus human interaction.

  4. Regulatory Compliance: Ensure adherence to industry-specific regulations.

  5. Technology Readiness: Evaluate the capability and readiness of AI and IT infrastructure.

  6. Error Handling Capability: Assess the system's ability to handle errors and escalate when necessary.

  7. Feedback Collection: Mechanisms for post-interaction feedback to improve future decisions.

  8. Language and Cultural Sensitivity: Assess based on the demographic and linguistic diversity of the customer base.

  9. Skill and Training of Human Agents: Evaluate the readiness of human agents to handle transactions if escalated.

  10. Real-time Monitoring: Mechanisms for monitoring and intervention during AI-handled transactions.

  11. Data Quality and Availability: Ensure necessary data is accessible, accurate, and up-to-date.

  12. System Integration Check: Assess the seamless integration and data flow across systems, especially for complex transactions.

The evolution of IVR systems over the past 50 years has paved the way for the next leap in customer service technology—AI-powered service bots. The benefits of IVRs were clear, yet so were the limitations. As AI rapidly enters the scene, businesses must consider models incorporating experience, expense, and risk to navigate the transition. This structured approach helps determine the extent to which AI can be integrated while keeping customer satisfaction, security, and operational efficiency in check.

It's about finding the right balance between automation and the human touch, ensuring that the essence of customer-centric service remains intact as we move forward.


  1. Colladon, A., Naldi, M., & Schiraldi, M. (2013). Quality management in the design of tlc call centres. International Journal of Engineering Business Management, 5, 48. Link

  2. Dean, D. (2008). What's wrong with IVR self‐service. Managing Service Quality, 18(6), 594-609. Link

  3. Ellway, B. (2016). What’s wrong with IVR system service? a spatial theorisation of customer confusion and frustration. Journal of Service Theory and Practice, 26(4), 386-405. Link

  4. Tezcan, T. and Behzad, B. (2012). Robust design and control of call centers with flexible interactive voice response systems. Manufacturing & Service Operations Management, 14(3), 386-401. Link



bottom of page