What is Human-in-the-Loop and why is it an important part of your conversational AI and hyperautomation strategy?
Creating great conversational experiences isn’t easy. As a result, most of them are less than impressive. To avoid lackluster results, incorporate humans-in-the-loop (HitL). With this approach, when a conversational application (or IDW – Intelligent Digital Worker) gets stuck, it can bring a human in to help. That human can ensure the end-user experience is fulfilled while also training the system so future occurrences don’t require intervention.
This type of training can either be done in real-time, with algorithms to ensure the integrity of the training data, or through a moderation process where training is reviewed before it’s applied. HitL requires an interface that is seamlessly integrated for real-time interactions and tools to match.
For better or worse, the hype surrounding conversational AI has set end-user expectations high. HitL can help you meet those expectations in a timely manner that matches user needs.
Humans need to be able to monitor IDWs as they work, entering into the experience when the bot doesn’t know what to do or needs guidance. The IDW can learn from the ongoing process of human-led refinement of automated tasks. The continued evolution and expansion of an ecosystem relies on the human-in-the-loop process. Humans are a critical part of the ecosystem, working seamlessly alongside their IDWs, asking one another for help, querying, and establishing recommended responses and actions.
On the flip side of that example, an IDW looking at analytics data and seeing that a significant number of users are requesting the option to prepay can reach out to a human peer for training on how to complete the transaction. IDWs can also assist human agents during conversations with users, whispering in the agent interface with facts about the customer, their last interaction, or solutions that have worked to help in the past. The more time IDWs spend learning from their human counterparts, the faster the ecosystem can evolve. They become more capable and require less human intervention, freeing humans to move on to automating more tasks.
Opportunities also emerge for something we call co-botting, where people and IDWs design or modify skills together. This could be as simple as a human realizing they should train an IDW on how to collect payment before a service is rendered. In order to contribute to AI-training and step in where human-touch is needed, people need the ability to moderate IDW-managed conversations in real-time. From these in-experience interventions, IDWs can learn from the live human-to-human interactions—as the member of your organization guides a user to the next step, the IDW gains new contextual data. The knowledge and skills retained by IDWs through in-line training can be leveraged across your organization.
Who should be a human in the loop?
HitL is a powerful, fluid role with the ability to bind and strengthen your entire ecosystem. If that sounds like a superhero bio, we’re on target. This role is fluid because anyone within your organization could be the human in the loop. HitL is powerful because when someone assumes the role, they are leveraging knowledge, perspective, and experience surrounding a task that they have a deep understanding of to fill the gaps and accelerate training for IDWs. This role lets people play directly to their strengths and therefore requires very little training.