Название: Scaling Generative AI: An Operational Readiness Framework for Enterprises
Автор: Amit Prabhu
Издательство: Business Expert Press
Год: 2025
Страниц: 318
Язык: английский
Формат: pdf, epub (true)
Размер: 10.1 MB

This book contains the operational readiness framework, providing step-by-step guidance to enterprises to prepare themselves for the scaled adoption of Generative AI. Different enterprises have reacted differently to the Generative AI hype. The real value of enerative AI lies in the scaled adoption. Only 10 percent of the enterprises have been able to scale. A staggering 90 percent of them are lagging. This has caused a "huge gap" between the scaling and lagging firms. Closing this huge gap is a daunting task.

To bridge this gap, enterprises must be operationally ready in the following four areas:
• Customer
• Technology
• Data
• People

This book contains the operational readiness framework, providing step-by-step guidance to enterprises to prepare themselves for the scaled adoption of Generative AI.

In a prompt based business architecture, LLMs are developed, pretrained, and fine-tuned with built-in applications provided by the vendor. A user just needs to write prompts to generate the required inference. In application based business architecture, LLMs are developed, pretrained, and fine-tuned by the vendor. But the applications need to be developed by the business before starting to prompt them. In fine-tuning based business architecture, LLMs are developed and pretrained on a large corpus of unlabeled data by the vendor. But the model needs to be fine-tuned and applications need to be developed by the enterprises before applying prompts on them. In pretraining based business architecture, only the LLM frameworks are developed by the providers. Pretraining, fine-tuning, and application development need to be done by the enterprises before applying prompts to them. In development based business architecture, the enterprises build an LLM from the scratch, then pretrain and fine-tune it, and develop an application before starting to apply prompts to it.

Fine-Tuning Versus RAG: In the RAG approach, rather than directly inputting a prompt to the LLM to generate a response, the process involves first parsing the user query through a RAG module. This module comprises a retriever and a knowledge base. The retriever uses the query to search the knowledge base for relevant information, which is then extracted to form an input prompt for the LLM. The knowledge base consists of a stack of documents that need to be prepared in four steps: loading the documents, chunking them into smaller pieces to fit the LLM’s context window, translating the chunks into text embeddings, and finally storing these embeddings in a vector database for efficient retrieval during the search process. This approach ensures that only relevant information is used to generate responses; thus improving system performance.

Although the framework is primarily for the executives, leaders, managers, consultants, strategists, and transformation drivers at the lagging firms, the scaling firms can use it to assess their current operational readiness levels and mitigate the prevailing gaps. It can also provide useful insights to the entrepreneurs in the Generative AI value chain to develop unique solutions. Additionally, it can help technology and management students to align themselves better to embrace new challenges of the corporate world they will soon enter.

The success of this book lies in how effectively the readers apply the framework at their workplace. This book is not just about information...it's all about transformation!






ОТСУТСТВУЕТ ССЫЛКА/ НЕ РАБОЧАЯ ССЫЛКА ЕСТЬ РЕШЕНИЕ, ПИШЕМ СЮДА!