Solving the live agent vs AI conundrum: The human-in-the-loop approach

Artificial intelligence is transforming customer interactions by making them faster and more efficient. The challenge for organizations is ensuring that automation does not eliminate the human touch that customers value most.

Michael Housman, Chief Science Officer and co-founder of RapportBoost.AI, advocates a human-in-the-loop approach to solving this problem. Rather than replacing live agents with bots, his model creates a collaborative system in which AI supports humans, and humans, in turn, train the AI.

In this framework, artificial intelligence analyzes millions of customer conversations to identify behaviors that improve outcomes such as conversion rates, order size, and customer satisfaction. The system then offers real-time recommendations to live chat agents. Crucially, the human agent decides whether to follow or ignore the guidance.

When agents follow a recommendation and outcomes improve, the system learns from that interaction and increases the likelihood of offering similar advice in the future. Over time, the AI becomes better calibrated to real-world human behavior.

This approach has delivered measurable results. In one retail case study involving an e-commerce company generating $25 million in annual revenue, selective rollout of AI recommendations led to improvements of 10%–70% in targeted agent behaviors. These changes produced a 7.6% increase in total order revenue and a 7.8% increase in conversion rates, pointing to a multimillion-dollar upside at scale.

A core insight from Housman’s research is that the success of customer interactions is driven far more by how something is said than what is said. Tone, empathy, responsiveness, and conversational nuance account for most of the variance in outcomes such as satisfaction, repeat visits, and purchasing behavior.

For this reason, RapportBoost.AI focuses first on understanding and codifying the elements of authentic human interaction. Only after these patterns are well understood does the company apply them to bots, with the long-term goal of building emotionally intelligent AI capable of adapting to context and customer intent.

Looking ahead, Housman predicts that the next wave of AI in customer experience will emphasize personalization, emotional intelligence, and memory of past interactions. Bots will move beyond customer service into sales and recommendation roles, powering the growth of conversational commerce at massive scale.

For organizations early in their AI journey, the guidance is clear: start small, identify high-impact use cases, and avoid turning control over to machines too quickly. Begin with humans, use AI to enhance decision-making and efficiency, and gradually automate repetitive tasks once best practices are well understood.

The greatest long-term impact will come from technologies that democratize data science—tools that allow non-specialists to ask questions, test hypotheses, and act on insights. In customer experience especially, empowering frontline teams with this capability can lead to sustained and meaningful improvement.

The overarching principle is focus and discipline: avoid sweeping AI initiatives that promise to solve everything at once. Instead, target specific problems where AI can deliver clear value, build momentum through early wins, and expand thoughtfully from there.