Call us at 650-400-3029 (PST)

Managing AI Decision-Making Part 3: Human on the Loop

by | Jan 21, 2022 | AI, Business Architecture, Business Intelligence, Decision Automation, Decision Management

Continuing our series on AI management options (kicked off by the HBR article Managing AI Decision-Making Tools), the next option is Human on the Loop (HOTL).

HOTL systems are most common in situations there is a need for 100% straight through or automated processing – the decision must be automated as it needs to be delivered in a situation where the response time is too short to include a human or the channel is completely automated such as a mobile app. HOTL systems allow for regular and rapid intervention on the part of human decision-makers not by injecting them into the system as it makes decisions, but by creating a rich data set about how the decision was made that allows the humans to quickly see when the decision-making needs to change.

For instance, one of our clients was a global life and health insurer who needed to increase the value of sales made by its large agency force. These agents used a mobile device to work with prospective clients, gathering financial needs and situation data and helping them apply electronically for protection and investment products. Increasing revenue meant helping agents identify the right upsell and cross sell opportunities while they were talking with prospects. 

The marketing department used advanced analytics and AI to identify offers for each prospect based on their unique financial situation and the products they were buying. This customer-level decisioning system was seamlessly embedded into the mobile application and ran in real-time, completely autonomously. 

The marketing team managed the feedback loop. They got detailed data on how offers were being selected, could track champion strategies against challengers, and make changes to the behavior of the system. This combination of seamless integration, real-time decision-making and human on the loop continuous improvement led to 98% adoption by agents, millions of dollars in additional revenue and over 50% lift in upsell results.

Many customer treatment and marketing decisions fit this pattern where regular and specific updates are possible and desirable, even though the decisions are made more autonomously. Which, of course, brings us to the final category, stay tuned for next week’s post about fully autonomous or Human out of the loop decision-making.

Click here to check out previous posts discussing other options, such as: Human in the Loop (HITL) or Human in the Loop For Exceptions (HITLFE).

If you have any additional questions on this article or other related topics, drop us a line – we’d love to connect.

0 Comments

Join our Newsletter