What this book covers

Chapter 1, Recognizing the Power of Design in ChatGPT, begins with a brief introduction to the relationship between design and LLMs, including the art and science of UX and the history of LLMs. The various design frameworks for deploying an LLM are discussed, including a chat UI, a hybrid UI that includes chat with graphical user interfaces, recommender UIs that are not interactive, and designs intended to work behind the scenes with backend solutions. There is a small hands-on lab for building a simple model using a no-code playground.

Chapter 2, Conducting Effective User Research, provides many tips and tricks for using some of the most critical user research tools to evaluate adding ChatGPT and LLMs to enterprise solutions. We cover methods such as surveys, needs analysis, interviews, and digging into data to create a conversational analysis.

Chapter 3, Identifying Optimal Use Cases for ChatGPT, teaches how to identify the breadth of solutions to which an LLM can add value and explains when an LLM is not suited for a use case. We briefly cover classic use case design and then spend time aligning an LLM’s capabilities with user goals. We ensure you know ChatGPT’s limitations and biases and how to handle inappropriate responses.

Chapter 4, Scoring Stories, helps you become an expert in prioritizing user stories. This chapter is also valuable after a product or service goes to customers. You learn to prioritize updates, patches, and bug fixes so the customer gets the most value from the team’s efforts. You will be able to balance customer priorities with the cost of development and make rational decisions to help plan and deliver the most value for the least cost. It explains in simple terms how to apply some special Agile tools to prioritize all this work. No road is without some bumps, but we share some complexities so you can navigate this successfully with the entire team.

Chapter 5, Defining the Desired Experience, is the final chapter before we get serious about the inner workings of ChatGPT. You will uncover specific considerations, design issues, and solutions for the full range of contexts of use. These include chat experiences, hybrid UIs (a graphical user interface merged with chat intelligence), recommendation UIs, and backend solutions (those without a customer-facing UI). We will address overarching considerations for these desired experiences, ensuring you know how to handle accessibility and internationalization while creating trust and handling security in any of these solutions.

Chapter 6, Gathering Data – Content Is King, dives into the complex nature of enterprise data, which is fundamental to creating a ChatGPT solution based on customers’ needs. Explore how data sources such as knowledge bases, databases, spreadsheets, and other systems provide a source of truth. This helps connect customers to actions and explains how product people like yourself can contribute at this stage. Hands-on activities and a case study on annotating and cleaning data help explain the key points. We will cover retrieval augmented generation to help bridge the gap between an enterprise’s vast data sources and the LLM.

Chapter 7, Prompt Engineering, coaches you on creating instructions that control, adapt, and personify the communications from the LLM to the customer. You will learn the difference between prompts anyone can give to an LLM and the more refined nature of instructional prompts for enterprise solutions.

Chapter 8, Fine-Tuning, explains what happens within the fine-tuning process, provides a tutorial on how to start fine-tuning, and continues our in-depth case study. You will be shown different methods to apply when training models. This includes a hands-on exercise to fine-tune a very sarcastic chatbot.

Chapter 9, Guidelines and Heuristics, steps past the technical nature of ChatGPT design to examine how to interpret ChatGPT style, tone, and voice. Essential guidelines and heuristics adapted and applied to evaluating ChatGPT solutions are reviewed so you can learn how to use design thinking to create clarity in the output from your LLM solutions. Dozens of examples are provided, along with a case study and example prompts that tie together the suite of heuristics covered in the chapter.

Chapter 10, Monitoring and Evaluation, focuses on knowing if the solution is doing well. It covers evaluating successes and failures, defining quality, and judging whether the UX improves. Our approach is one of care and feeding, following the life cycle of learning from the product’s users and feeding back any learnings to have it grow and mature. Statistical measures of model performance, user quality metrics, and heuristic evaluation methods are covered, with tips on improving quality.

Chapter 11, Process, focuses on adapting traditional Agile and modern development methods to more interactive and customer-driven needs to improve ChatGPT solutions rapidly. We cover practical strategies to integrate a care and feeding approach into traditional Agile or Agile-like development while explaining why you should advocate for a continuous improvement life cycle.

Chapter 12, Conclusion, is the final chapter and provides additional suggestions and coaching to wrap up the entire life cycle covered in this book to set you up for success.