Building and Distributing Agentic AI Solutions A Comprehensive Guide to Productizing AI (Julian Soh, John Tidwell, Priyanshi Singh) (z-library.sk, 1lib.sk, z-lib.sk)

Page 1

(This page has no text content)

Page 2

Building and Distributing Agentic AI Solutions A Comprehensive Guide to Productizing AI Julian Soh John Tidwell Priyanshi Singh

Page 3

Building and Distributing Agentic AI Solutions: A Comprehensive Guide to Productizing AI ISBN-13 (pbk): 979-8-8688-2585-9 ISBN-13 (electronic): 979-8-8688-2586-6 https://doi.org/10.1007/979-8-8688-2586-6 Copyright © 2026 by Julian Soh, John Tidwell and Priyanshi Singh This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Trademarked names, logos, and images may appear in this book. Rather than use a trademark symbol with every occurrence of a trademarked name, logo, or image we use the names, logos, and images only in an editorial fashion and to the benefit of the trademark owner, with no intention of infringement of the trademark. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Managing Director, Apress Media LLC: Welmoed Spahr Acquisitions Editor: Smriti Srivastava Editorial Assistant: Marina Engler Cover designed by eStudioCalamar Cover image designed by kjpargeter on Magnific Distributed to the book trade worldwide by Springer Science+Business Media New York, 1 New York Plaza, New York, NY 10004. Phone 1-800-SPRINGER, fax (201) 348-4505, e-mail orders-ny@springer-sbm.com, or visit www.springeronline.com. Apress Media, LLC is a Delaware LLC and the sole member (owner) is Springer Science + Business Media Finance Inc (SSBM Finance Inc). SSBM Finance Inc is a Delaware corporation. For information on translations, please e-mail booktranslations@springernature.com; for reprint, paperback, or audio rights, please e-mail bookpermissions@springernature.com. Apress titles may be purchased in bulk for academic, corporate, or promotional use. eBook versions and licenses are also available for most titles. For more information, reference our Print and eBook Bulk Sales web page at http://www.apress.com/bulk-sales. Any source code or other supplementary material referenced by the author in this book is available to readers on GitHub. For more detailed information, please visit https://www.apress. com/gp/services/source-code. If disposing of this product, please recycle the paper Julian Soh Seattle, WA, USA John Tidwell Brandon, MS, USA Priyanshi Singh Danville, CA, USA

Page 4

This book would not have been possible without the support from family members, the technical community, esteemed colleagues, and customers who have welcomed us on their journey. Special thanks to my co-authors, John Tidwell and Priyanshi Singh. This book is dedicated to all the builders and creators.

Page 5

(This page has no text content)

Page 6

(This page has no text content)

Page 7

(This page has no text content)

Page 8

(This page has no text content)

Page 9

(This page has no text content)

Page 10

(This page has no text content)

Page 11

(This page has no text content)

Page 12

xiii About the Authors Julian Soh is a Forward Deployed software engineer with Microsoft and works with organizations to develop products that leverage emerging technologies on Microsoft and/or open-source platforms, specifically around applied/generative AI, and traditional ML like CNN and RNN for image and anomaly detection. With the rapid development in the field of embodied (physical) AI, Julian is leveraging his formal training as a mechanical engineer to help independent software vendors (ISVs) extend software design that could include wearables, embedded/real-time computing, and autonomous machines. His interest is in the intersection of software and the control of electromechanical systems, focusing on the areas of artificial intelligence and advanced analytics for independent software vendors (ISVs) who develop software solutions based on the Microsoft technology stack. Julian holds a BS in Mechanical Engineering and is pursuing his MS in Electrical and Computer Engineering at the University of Washington.

Page 13

xiv John Tidwell is Chief Technology Officer at YoungWilliams, where he leads a team responsible for the ongoing design, development, and deployment of Priya, an innovative AI-driven government service solution that leverages Microsoft Azure’s AI platform. With expertise in Azure cloud technologies, generative AI, DevOps, and Agile software development, John has guided numerous projects from concept to production. His technical leadership includes architecting scalable systems, driving strategic cloud adoption, and mentoring development teams to achieve excellence. John holds multiple Azure certifications, including Azure AI, and is actively pursuing a Master of Science in Computer Science specializing in Artificial Intelligence and Machine Learning at Georgia Tech. Priyanshi Singh is a senior artificial intelligence and machine learning technical specialist at Microsoft, specializing in designing end-to-end cloud solutions that leverage generative AI models and AI implementation best practices. She holds a master’s degree in Data Science from New York University and has a robust background as a data scientist, focusing on machine learning techniques for predictive analytics, computer vision, and natural language processing. Priyanshi is dedicated to helping the public sector and independent software vendors (ISVs) transform citizen services through artificial intelligence. She has been recognized as Microsoft’s FY24 State and Local Government Pinnacle Winner for her exceptional contributions to AI adoption and the growth of Azure business. Additionally, Priyanshi is a sports enthusiast, excelling in badminton and enjoying golf and billiards. abouT The auThors

Page 14

xv About the Technical Reviewer Mittal Mehta is an accomplished technology leader with deep expertise in AI-driven automation, DevOps automation, configuration management, and release engineering for both on-premises and cloud-based enterprise applications. With more than two decades of hands-on experience, he has built scalable automation frameworks and end-to-end delivery pipelines that enhance reliability, speed, and governance across the software life cycle. He is passionate about exploring and adopting emerging technologies, particularly in the areas of AI-driven automation, intelligent workflows, and application life cycle optimization. Mittal has extensive experience working with Microsoft technologies, including Azure DevOps, Power Automate, and PowerShell, and has architected numerous cloud automation solutions using Azure and modern DevOps toolchains. Based in Bangalore, he currently serves as a Principal Consultant in AI automation where he leads automation strategy, cloud enablement, and AI-powered engineering initiatives for enterprise applications.

Page 15

xvii Introduction This book is for the creators, builders, and entrepreneurs looking to incorporate AI in new or existing software (or maybe hardware) products. We hoped to capture the key aspects of product development, from ideation to the Go-To-Market (GTM) motion leveraging distribution channels like the Microsoft Marketplace. One of the challenges when developing an AI product is the understanding of cost and what adoption could look like. We have attempted to provide a good foundation to determine cost and a proposed way to model adoption. Entrepreneurs, product and marketing executives, and financial officers are our target audience because there are chapters in this book that are focused on the economic aspects of AI productization. Many new and trending AI technologies are also introduced and these form the platform on which products are being built. Technologies such as Azure Foundry, frontier models, agentic frameworks, and the Model Context Protocol are covered, along with sample codes and an accompanying GitHub repository. This portion of the book will provide our technical audience with actual hands-on exposure and insight to the learnings from our engagement with technical teams across various industries. All the topics were carefully selected based on real-world experiences and while this is ever evolving, we hope this book will provide a strong foundation for any AI productization journey.

Page 16

1© Julian Soh, John Tidwell and Priyanshi Singh 2026 J. Soh et al., Building and Distributing Agentic AI Solutions, https://doi.org/10.1007/979-8-8688-2586-6_1 CHAPTER 1 AI Landscape and Productization Opportunities At the time of writing, ChatGPT1 is three years old. Prior to that, generative AI was an experimental technology for niche firms, AI engineers, and data scientists. Fast forward to late 2025, it has evolved into a core business capability worldwide. No other technology has experienced a higher adoption rate. A study2 conducted by Vanderbilt University in August 2024 found that Generative AI reached a 39.5% adoption rate in just two years, compared to • 20% for the Internet (after two years) • 20% for personal computers (after three years) AI has driven the stocks of major technology companies like Microsoft, Amazon, and Google to record highs on the promise of this AI revolution. In return, these companies have poured billions into infrastructure, particularly power and compute (GPUs), and R&D to meet demand. 1 1 OpenAI unveiled ChatGPT to the world on November 30, 2022. 2 Vanderbilt Study: https://news.harvard.edu/gazette/story/2024/10/ generative-ai-embraced-faster-than-internet-pcs/

Page 17

2 We, the authors, have worked in technology and written about Cloud Computing since the day it became mainstream, and to witness a global shortage of compute at the biggest cloud providers is something we never thought we would see. But what is important, and therefore the focus of this book, is the need for builders, entrepreneurs, and engineers to build AI-enabled products and services or, at the very least, transform existing solutions by infusing AI capabilities. Companies are scrambling to hire skilled AI workers, and historically, we are seeing more startups by young entrepreneurs building businesses based on AI.3 This book is meant to help innovators, builders, and aspiring businessmen grasp an understanding on the foundational aspects of building an AI-enabled product or service. The topics we will cover in detail should touch on the entire life cycle of product development. A Mainstream Revolution Global adoption is extremely broad. Although the United States and China lead in AI development and patents, wrestling each other for talent and raw materials, the beneficiary is a worldwide audience. AI adoption is widespread: major hubs include the United States, India, the UK, Canada, and Germany, with cities like London, San Francisco, Bangalore, and Singapore emerging as vibrant AI centers. Over 16,500 companies and 6,000 startups are active in generative AI worldwide, reflecting a vast ecosystem of solution providers and innovators. This explosion is backed by venture funding (nearly $4 billion in VC poured into GenAI in just one quarter of 2024) and rapid talent growth. And this is not slowing anytime soon. 3 https://www.forbes.com/sites/sarahemerson/2024/12/03/30-under-30- ai-2025-the-young-entrepreneurs-coding-the-future/ Chapter 1 aI LandsCape and produCtIzatIon opportunItIes

Page 18

3 In 2025, several key trends started to emerge: • Multimodal AI (models that understand/generate text, images, audio, etc.) and conversational AI are major drivers of innovation. • Organizations have moved beyond just experimenting – roughly 44% of companies were running pilot programs for generative AI and 10% had reached production scale by late 2024, and these numbers have only climbed in 2025. • Early adopters report significant productivity gains and operational improvements, motivating others to follow. In fact, industry estimates suggest the question is no longer whether to adopt GenAI, but how to implement it effectively and responsibly. With respect to the last trend cited above, we would argue that first- to- market innovators are way past discussing the ethical nature of AI, and have started to adopt and implement industry-backed standards. It may be an unpopular opinion, but we feel that existing discussions that continue to question the use of AI in lieu of adoption is a futile effort to resist change and is bound to fail. Remember, there were a lot of naysayers and doomsday predictions at every inflection point. For AI, we have entered the maturation phase. This maturation that brings the adoption of standards includes • Ensuring factual accuracy, security, and ethical use of AI • Investing in governance and AI training programs as AI becomes central to products and workflows • Helping regulators with AI laws and legislation – from EU AI Act efforts to sector-specific guidelines – to set guardrails for responsible AI deployment Chapter 1 aI LandsCape and produCtIzatIon opportunItIes

Page 19

4 Overall, late 2025’s landscape is one of widespread enthusiasm tempered by pragmatic execution: generative AI is widely recognized as a game-changer (potentially $6–8 trillion in economic impact according to McKinsey). Furthermore, during the Q3 Calendar Year 2025 season of earnings reports, the “Magnificent 7” group of companies reported continued, but easing, global shortage of GPUs, which is an indicator of demand, and provided guidance on increased capital investment. While nothing is guaranteed, there are abundant signals that AI is not a passing fad, and that is why we should include it as a core feature in any product or service. Who Are AI Customers? Before we talk about AI-enabled products and services, the important question to ask is “who are the customers?” Just like how the PC became an indispensable tool spanning every industry and household, AI is doing the same, but more specifically, because of Generative AI. Generative AI brought us the Large Language Model (LLM) and LLMs speak our language, and we mean this in the most literal sense. As a result, the answer is everyone, every industry and every entity – private, public, education, consumer. We have always heard of the different inflection points in the history of technology – steam engines, electricity, personal computing, the Internet. We are now in the midst of another major inflection point with AI. When PCs debuted, users needed to possess knowledge to interact with the technology. These are primarily text-based commands with cryptic syntax that were not standards-based. Apple, followed by Microsoft, penetrated the consumer and enterprise markets with the Graphical User Interface (GUI), which is what opened the floodgates for adoption. Yes, we can argue that Xerox was the inventor of the GUI, but that is not the point. The point is that the GUI was introduced to the Chapter 1 aI LandsCape and produCtIzatIon opportunItIes

Page 20

5 masses and because the GUI aligns better with the natural cognitive abilities of most of the population, that is where we see the exponential adoption of personal computing. Since then, we have been pushing for better user interfaces (e.g., WYSIWYG, Drag-and-Drop, Cut-and-Paste, Intellisense, etc). The LLM is the most significant upgrade in UI since the GUI because it elevates the UI to the highest form of human cognitive alignment – Natural Language. An LLM interface provides the ability to use natural language to get the model to do things. No special syntax, format, programming ability, or even any experience in data science needed. For that matter, if you ask an LLM like ChatGPT using one of the more than 40 languages supported, it will understand what you are asking for, even if it is not able to provide the information or carry out the task. LLMs’ ability to converse in natural language is the main driver for adoption. No other technology was as accessible as an LLM. When ChatGPT was first introduced, consumers flocked to the site to ask it to generate dinner ideas, travel itineraries, poetry, letters, emails, party ideas, and almost everything imaginable. Realtors are leveraging it to create narratives for listing properties, government organizations use it to quickly understand and apply complex rules and regulations. Commercial organizations are using it to create advertising materials, marketing plans, and identify customer sentiment. Legal professionals and lawmakers use it to quickly understand legal documents and also create initial drafts of new legislation. Since the common skill needed to make use of this technology, in all these use cases, is the ability to communicate in natural language, the personal and professional personas of individuals overlap. On the regulatory compliance front, the gap between the use of AI in government versus the commercial sector is also starting to close. This is mainly due to the maturity of cloud computing, better understanding of risks, and improvement in technologies and processes that make it more feasible to provide a common standard across all clouds. As such, many Chapter 1 aI LandsCape and produCtIzatIon opportunItIes