Generative AI based on Large Language Models (LLM) is a new entrant in the AI gamut of algorithms. It comes with a lot of promises to augment/enhance many use cases that are traditionally built using traditional AI or even non-AI algorithms. One such use case is the use of chat bots for consumer engagement. The state of the art LLMs have shown capabilities of doing moderate reasoning and generating mostly meaningful text. But they suffer from various limitations such as hallucinations, limited context, and even outright misinformation. So, if one has to leverage its strengths (doing moderate reasoning and generating mostly meaningful text) in any solution, they should be cognisant of its limitations as well. Here I present some Ux design considerations to this effect in the context of deploying GenAI powered chatbots for consumer engagement. I will cover other use cases of leveraging GenAI in solutions, in later posts.
Disclaimer
The first thing to include in a customer facing chat bot, which gives out responses generated by LLMs, is a disclaimer that clearly calls out the following things:
- The responses of the chat bot are generated by a machine, and not a human.
- It may contain misinformation
- The chat bot comes with no warranty – the company is not responsible for the damage caused to the consumer by the errors or misinformation provided by the chat bot
References for The Response
If the chat bot referred to any documents for generating the output, it would be a good idea to provide a link to the referenced documents, which can be accessed by the consumer for more detailed and correct information
Alternate Options for Availing Support
The chat bot gives immediate response for the queries of the consumer, which is low on cost as it is automated. But since the capabilities of the chat bot are limited, the customer may be given alternate options for availing support, like an option to file a customer support ticket, or to directly talk to a human support agent etc. The trade offs here for the consumer is delay in response and fee for the particular support modality as compared to the immediate and low cost but imperfect response from the chat bot. The advantage for the company in deploying a GenAI powered chat bot are the following:
- Reduction in the number of support tickets raised, as many repetitive queries gets fulfilled by the AI responses, and only the remaining queries gets in to the ticketing and on-call pipelines
- The AI responses generated for the consumer queries can act as a starting point for human agents assigned to respond to / solve the support tickets / support calls, reducing their efforts and turnaround time
Stay tuned for commentary on other use cases of GenAI powered solutions…
Nisanth’s Portfolio and Blogs 