Communicating with AI Chatbots: Prompt Engineering

We are all humans and we interact with each other through a language. Although we may use tools like Video conferencing, Virtual Calls, Face to Face meetings, Virtual meetings, events, conferences, meetings, etc. the one common factor among all of this is the use of one or two or more languages while communicating.

When it comes to chatbots or may I say Advanced AI Chatbots such as the likes of ChatGPT, Gemini, etc. the one common factor in communicating with these chatbots is the use of Natural Language though Prompts.?A quick caveat: Neither I work with the owners of ChatGPT or Gemini nor am I incentivized to promote one or other.

Prompts are natural language statements, sentences, questions, etc. which we usually type into the specific message tabs in these chatbots and then press Enter to let the chatbot understand our input and then respond with an appropriate output or intended response. Let me get a bit technical here and help sight the importance of LLMs and NLP in advanced chatbots.

NLP or Natural Language Processing is a technology that helps chatbots to understand and respond to human speech and text. Using a variety of algorithms and models, NLP enables chatbots or advanced chatbots to analyze and break down sentences to help them identify keywords, entities and relationships between them. This understanding helps chatbots to understand and grasp the intent of the users.?

LLMs or Large Language Models specifically in case of chatbots are models trained on huge amounts of text data and have an impressive understanding of the human language per se. This enables chatbots to generate coherent and contextually relevant responses. LLMs actually enhance chatbot responses by enabling them to understand language better, to get an awareness of the context and helping them engage in human-like conversations. This further results in generation of personalized and engaging user-chatbot interactions. Some common examples of prominent LLMs are GPT-3, GPT-3.5, GPT-4, GPT-4o, Lamda, Llama, Gemini, Claude etc. which we may be using or would have heard of.?

For example, if someone is a Supply chain intern and her manager may ask her to help the manager understand how she can improve the efficiency of the distribution center the intern is working at, and the intern wants to align her thinking with practical scenarios, she may want to log on to ChatGPT and enter the following prompt.?

Prompt -?

“Assume you are a supply chain expert. I am a supply chain intern at one of the largest distribution centers of an ecommerce company. My manager wants to create a report based on the 10 most important productivity metrics to improve the efficiency of the distribution center. Help me with the list of 10 most important metrics I can analyze for improving the efficiency of my distribution center. Also give me detailed explanations of the metrics and how I can apply these to my distribution center use case. Thank You.”

When I entered this prompt in the free version of ChatGPT, it not only resulted in informing the 10 most important productivity metrics to analyze the efficiency but also resulted in how I can apply these metrics to the distribution center use case.

A caveat here: The responses that result from these chatbots must always be taken with a pinch of salt. Why do I say so? Well, these advanced chatbots also suffer from something called hallucination which may result in a wayward response or result sometimes. And, when applying the recommendations we always must align with our own training and thinking before using and/or applying any recommendation as is.

Now, this art of communicating with these advanced chatbots or engaging with these tools using prompts is called “Prompt Engineering”.

This can be learnt and this is a skill which will enable greater productivity for firms and employees across functions and industries!?