Hello, everyone. Thank you for joining this session on prompt engineering. As AI systems become more integrated into our workflows, understanding how to communicate effectively with them becomes increasingly important. Today, we will dive into the challenges, key considerations, and best practices for crafting prompts that yield effective AI responses. Prompt engineering is the practice of designing and optimizing inputs to AI systems to get desired outputs. Think of it as learning to speak the language that bridges human intent and AI capability. Just as we adapt our communication style when speaking to different people, we need to adapt how we communicate with AI to get the best results. Why should we care about prompt engineering? First, better prompt leads to more accurate and relevant responses, crafted prompts save time and computational resources. When using commercial AI services, efficient prompts can significantly reduce costs. And finally, a good prompt engineering helps minimize risks like bias and hallucinations. We can safely say that it is the art and science of creating effective prompts that guide AI to provide relevant and accurate answers. In any AI application, especially in conversational AI and generative tools, a well crafted prompt can be the difference between a useful response and one that misses the mark completely. One of the primary challenges in prompt engineering is ambiguity. An unclear prompt can lead AI to respond in the ways that may not meet our needs. Another challenge is bias. At certain prompts may inadvertently guide the AI toward biased responses, depending on the data it was trained on. Context dependency means that slight change in phrasing can alter responses. Character limits can restrict amount of context we provide. And finally, without proper structure, we might get inconsistent responses across similar queries. Before writing any prompt, consider these key factors. What's your ultimate goal? We need to consider the audience and purpose. A prompt intended for technical users will differ from the one aimed at general consumers. Then what does the AI, what context does the AI need? What are the constraints you are working within? It's essential to recognize the limitations of the AI, understanding what it can and cannot do. What format should the response take? And importantly, what safety considerations should you keep in mind? Here are some best practices for prompt engineering. First, use clear instructions. The more direct you are, the more likely the AI will respond accurately. Next, refine your prompts iteratively. Test and experiment with different versions to find out what works best. Testing and monitoring for bias and sensitivity is critical to ensure fair and balanced responses. And provide context whenever possible to narrow down the scope of the response. Let's look at the best practices for structure. While you're structuring your prompt, start with relevant context. Clearly state the task and then specify the desired format. Whenever helpful, provide examples and then outline any constraints or limitations. Following this structure helps ensure clear communication with the AI. Now let's go over an example of what a poor prompt is and what a prompt that follows the best practices I mentioned earlier looks like. Here, the user wants to know how can I improve customer satisfaction in my online store. While it's clear that the focus is on customer satisfaction, it doesn't specify areas of, what areas are of concern, which leaves the response too broad. The prompt doesn't specify how many suggestions are needed or what kind of suggestions are needed, such as whether they are looking for quick tips or they are looking for in depth strategies. It also doesn't indicate the type or size of the online store. Which leaves the AI to guess what might be relevant. Now, what on the right hand side is the improved prompt. Let's get into the details of how that prompt becomes effective. By following the best practices. This prompt specifies that the recommendations should focus on customer service, website usability, and post purchase follow up. It helps the AI target these key areas. It also clarifies that it's for a small online retail business, which allows the AI to tailor suggestions for a smaller team with limited resources. Requests five concise and actionable strategies. By setting expectations for a focused and structured response, that is manageable for the intended audience. Now let's look at another example and compare the responses from the AI. This is a prompt for creating emergency evacuation instructions. The initial prompt is straightforward. But it lacks the specificity needed for a complex setting like a concert venue, which is a detail which has been provided in the effective prompt. Write emergency evacuation instructions is a vague request that doesn't provide any context, format requirements, or audience considerations. As a result, the response is too generic, not tailored for the unique needs of a large venue, and failed to account for specific emergency types, languages, and accessibility needs. In contrast, the improved prompt is a great example of how incorporating the best practices can lead to a much more effective response. This prompt specifies the type of venue. A 1000 seat concert hall with multiple floors and exits, and outlines particular scenarios such as fires, medical emergencies, and security threats. It also includes special instructions for disabled patrons, staff roles, and event for even format requirements, like bullet points and universal emergency symbols to ensure clarity and usability. With this detailed prompt, the AI generated result is much more comprehensive and practical. For example, each emergency type, fire, medical, and security has clear and concise instructions with universal symbols and multilingual indicators. Specific assembly points and special instructions for mobility assistance and a breakdown of staff roles are all provided, ensuring that the instructions are easily understood and actionable. This comparison shows the power of a well crafted prompt. When we are specific about our requirements, structure, tone, and context, we guide the AI to produce a response that is detailed, organized, and ready for use. The improved prompt's clarity, structure, and relevance show how best practices in prompt engineering lead to much more accurate and usable outputs. Prompt engineering is an iterative process. So start with a basic prompt, evaluate the response, identify issues, refine your prompt and repeat. Each iteration helps you understand what works and what doesn't, which leads to better results over time. Use of advanced techniques can enhance your results. In chain of thought prompting, we guide the AI through multi step reasoning to enhance its accuracy. It helps AI break down complex problems. Role assignment can help set the tone and perspective, such as asking the AI to respond as if it were a historian. It provides helpful context. Few shot learning uses examples to guide responses. By allowing us to give specific examples within the prompt and setting a standard for responses. System instructions help overall behavior parameters. Using conditional prompts with if and then conditions can help clarify nuanced instructions. Adjusting the temperature of the model controls and the randomness of responses, allowing us to define, to fine tune between creative or factual outputs. Now let's look at some common pitfalls to be mindful of. Even after applying best practices, we need to keep these common pitfalls in our mind. First, overcomplication. Sometimes, in an effort to be thorough, we add too much detail, which can overwhelm the AI, causing it to miss the main point and diluting the focus. So aim to keep the prompts clear and direct. Assumption of knowledge is another trap. Never assume the AI has background information or specific knowledge without explicitly providing context. Lack of specificity is a frequent issue too. Without clear direction, the AI may provide broad or irrelevant answers. Unintentional bias can also slip in. Based on the wording we choose. So it's essential to carefully phrase prompts to avoid the AI going in a biased direction. Ignored constraints. Whether character limits, formatting requirements, or ethical boundaries, it can affect the quality and suitability of the response. for listening. So by keeping these pitfalls in mind, you will be more prepared to create prompts that consistently lead accurate and relevant results. Now that we have followed all the instructions for prompt creation, it is important to validate its quality. We start with an accuracy check. Does the AI response directly address what you are asking for in the prompt? Task completion check is for verifying that the AI successfully accomplishes the intended task. Relevance checks that the output is focused and appropriate. Bias evaluation involves reviewing the responses to catch any unintended biases, which may require adjusting the prompt. Consistency testing ensures that reports are accurate. Repeating the prompt yields consistent quality answers. And finally, user testing is invaluable. Gathering feedback from actual users can reveal if the prompt is achieving its intended clarity and effectiveness. As AI systems evolve, prompt engineering will too. We will see new capabilities, better tools, and evolving best practices. The key is to stay adaptable and keep learning as the field develops. Thank you so much for paying attention today. I'm Aditi Godbole. Please feel free to connect with me on LinkedIn for further discussions or any questions. Thank you.