Should you use LLMs to improve customer experience?

LLMs are the talk of the town today. But not every organization needs to invest in building sophisticated chatbots for their customer experience. Here's a list of problems, solved and better solved with customized chatbots.


Posted on Apr 24, '24

Solvable with simple tree-logic based chatbots, even if not LLM-powered.

  • Volume. What is the volume of requests you receive? Do you have more than 5 customer service agents?
  • Repetition. Do you have a high number of inquiries or repetitive tasks?
  • Bandwidth. Are you a solopreneur spending 2-3 hours tackling customer requests alone?
  • Availability. Do you have a global client-base? Is 24/7 availability important for your business?

Better addressed with customized chatbots augmented with a knowledge base.

  • Locale. Is your audience diverse and multi-locale?
  • Scope. What kind of requests do you get - open-ended or contained? Can they be answered with logical flows or do they require human judgment?
  • Search & Retrieval. Do your requests require quick information retrieval from different places that might be a struggle for a human? Is this information in the form of large blocks of text?
  • Automation inefficiency. Do you already have automation set up? What fraction of tickets are currently unsolved by your automation system?
  • User onboarding. How many onboarding flows do you have? Do you have multiple onboarding tutorials for your app or often have service agents “walking someone through a process”?
  • Voice and Branding. Can you envision a chat-first interface as the future of your business? Do you see chat assistants as a brand strategy?

Red flags, for not considering an LLM for your customer experience.

  • Risk tolerance. Can you risk your chat assistant being wrong or incorrect? What is the potential cost of being incorrect?
  • Business personality. Do you have a personal touch based business? Do you have a small number of elite users who value personal connection?