Agentforce: Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry

By Martin Kihn and Marc Benioff

Introducing AI agents, the groundbreaking third wave of AI's integration in the workforce

Agentforce: Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry tells you how companies can create and control their own AI agents and build a virtual workforce. It goes behind-the-scenes on how Salesforce built a platform to drive AI agents, solving problems like hallucinations and bias through a framework that gives agents strict roles, data sources, actions, guardrails and channels to reach customers. This book draws from extensive research and exclusive access to Salesforce's leaders and their ambitious plan to dominate the race to develop and own the AI agent space.

In this book, readers will find information on:

• AI agents as a "third wave" of AI development that goes far beyond simple chatbots and "co-pilots" through harmonized data, Retrieval Augmented Generation (RAG), and Salesforce's innovative Atlas Reasoning Engine
• Steps to develop prompt guidance, topic creation (areas of work), explicit instructions, and a menu of actions allowed
• Salesforce customers, such as Saks and OpenTable, that are already using AI agents with success
• The effects of AI and automation on the job market

Agentforce: Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry is an indispensable, forward-thinking resource on the subject for all leaders in business seeking to supercharge their organizations' initiatives through the latest developments in a rapidly advancing field.

• Language: English
• Publisher: Wiley
• Release date: Jun 9, 2025
• ISBN: 9781394349234

Book preview

Martin Kihn

Foreword by Marc Benioff

CEO and Co-Founder, Salesforce

Agentforce

Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry

Logo: Wiley

Copyright © 2025 by John Wiley & Sons, Inc. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey.

Published simultaneously in Canada.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, [phone], fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, [phone], fax [phone], or online at http://www.wiley.com/go/permission.

Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at [phone], outside the United States at [phone] or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data is Available:

ISBN 9781394349227 (cloth)

ISBN 9781394349234 (ePub)

ISBN 9781394349241 (ePDF)

Cover Design: Wiley

A screenshot displays a tweet by Andrej Karpathy (at karpathy) stating, The hottest new programming language is English. The tweet was posted on January 24, 2023, at 3:14 PM and has garnered 5.2 million views.

Foreword

by Marc Benioff, CEO and Co-founder, Salesforce

I'm a member of the last generation of CEOs to manage an all-human workforce.

I can honestly say that when I co-founded Salesforce in a one-bedroom apartment on Telegraph Hill in San Francisco in 1999, I had no idea I would ever write that sentence. Artificial intelligence has come very far, very fast – and we've only just begun to realize its impact on our work and our lives.

Machines are now able to do things that would have seemed miraculous even 10 years ago. AI is able to carry on natural conversations, answer complex questions, summarize call transcripts, create images and videos based on simple prompts, and even write poetry.

As impressive as all this is, it's just a start. We have entered into an era when AI agents cannot just chat but also think, make decisions on their own, and take action on those decisions. AI agents are already taking their place as part of a hybrid agent-human workforce – what I called a global digital labor platform when we launched Salesforce Agentforce in 2024.

Agentforce is the first trusted platform that enables global enterprises to put AI agents to work, alongside a human workforce, to enhance efficiency and prime productivity. It is a major transformation, one that is already providing value to thousands of our customers around the world.

This agentic AI moment is like nothing I've seen before in my career. It's as important as the introduction of the Internet, which made Salesforce.com (as it was then called) possible. And it comes at an important instant in the evolution of business. In 1999, Salesforce introduced a new model that enhanced productivity, a new business model, and a new economic model. It was only the beginning.

In most developed markets, there is a labor shortage; workers are overwhelmed. As I shared at the World Economic Forum in Davos this year, I think agentic AI has the potential to address the shortage, making human workers more productive. Greater productivity drives up GDP, lifting the economy for all of us. McKinsey predicts that agentic AI could add $13 trillion to the global economy by 2030. In other words, agents introduce even more profound and transformative productivity, business, and economic models.

Of course, I've thought a lot about the impact of AI agents on jobs. Agents will change almost all forms of work. But history shows that new technologies create more jobs than they replace. From 1950 to 2020 – a time of incredible innovation – more than 100 million new jobs were created in America. A lot of these were unforeseen: after all, who knew prompt engineer would be a career option in 2025?

AI agents driven by Agentforce give companies capacity beyond human and physical limitations. Rather than just driving down costs, we're seeing our customers use Agentforce to do tasks they previously weren't able to do. At the same time, like all new technologies, AI agents come with real challenges. Mismanaged or unmindfully deployed, they can even cause harm. That's why I believe a platform like Agentforce isn't just nice to have but absolutely essential for the successful implementation of AI agents in the workplace.

As powerful as they are, LLMs are not perfect. They are trained on data from the Internet, which enables them to master the nuances of grammar but not to make detailed decisions about your customers or your business based on proprietary information. And we're all aware of incidents when LLMs hallucinated and showed bias.

The fact is that companies need a lot of additional tooling beyond the foundational LLM models themselves to put agents to work. That's why we built Agentforce, which has trust built into the core model.

Agentforce is the way companies can unleash the power of advanced AI on their own businesses. Using our platform, companies can ground agents’ decisions in their own data, set their own goals and guardrails, and determine how much freedom and access to give individual members of their new hybrid workforce. We provide the tools to spin up multiple agents for different purposes, test and refine them, scale them very rapidly – all on the highly performant, trusted infrastructure of the world's #1 CRM with the most advanced data capabilities to date.

The reason I'm excited about Agentforce is that I believe Salesforce is uniquely positioned to deliver on the promise of AI agents, if we do it right. From the beginning, Salesforce was built to be flexible, open, and highly customizable, using metadata, APIs, and other frameworks to ensure extensibility. And thanks to Data Cloud's real-time customer data management innovations, we support the enterprise data layer so crucial to the success of AI.

The power of AI agents isn't limited to businesses, of course. Agentic AI will transform our personal lives, our healthcare, how governments function, so many things. It will become ubiquitous to the point where it will be normal for us to have agents running 24/7 in our digital spheres, acting on our behalf and interacting with one another.

I hope I've convinced you that we're at an exciting moment for our industry and for Salesforce. That's why I'm delighted that Martin Kihn has written this lucid and highly readable introduction to Agentforce – how it works, why it matters, and how you can put it to work.

Welcome to the wonderful world of Agentforce. Things will never be the same.

The Five Attributes of an AI Agent

Role: What job should they do?

Data: What knowledge can they access?

Actions: What capabilities do they have?

Guardrails: What should they not do?

Channels: Where do they work?

Author’s Note

No generative AI was used in the creation of this book. This is not an indictment of the technology, which is brilliant, but rather to avoid confusion. All praise – or blame – for the contents of this book should go to its all-too-human author and his sources.

This is a moment in time like we have never seen – it’s beyond any description, there is no metaphor.

—Marc Benioff, CEO and Co-Founder, Salesforce

Agentforce Kickoff, San Francisco

The day after the 2024 presidential election, Salesforce’s co-founder Marc Benioff assembles the 500 top leaders of his 80,000-person company in a windowless basement room at the Ritz Carlton San Francisco for an emergency summit.

Security is tight: employees are required to scan their badge at every elevator and hallway; bags are checked; homework is mandated.

The election isn’t mentioned. The date is a coincidence. The topic at hand is what Benioff believes is the biggest thing to happen in any of our lifetimes – something he’s decided to call Agentforce.

It will soon be a recognizable brand: Salesforce is preparing a massive TV campaign, and there’s already a billboard on Highway 101 to the SFO airport with the company’s trademark cartoon animals, rebuilt as robots in Ray-Bans, attesting I Chose Agentforce.

In a few months, there would even be a commercial during that strategic conversation between the Chiefs and the Eagles otherwise known as Super Bowl LIX. All of this AI awareness features Salesforce’s brand ambassador, Matthew McConaughey, at his Texas-whimsical best, reciting lines he seems to have written himself.

The campaign’s tag line morphs to What AI Was Meant to Be.¹

Now the last time you saw an ad during the Super Bowl for a CRM company was … never.i Salesforce’s investment in associating its brand with AI and in making this astonishing, unsettling technology seem less frightening is unprecedented.

But this is an unprecedented era, an unanticipated chance. Almost nothing is predictable.

For one thing, when Benioff appears at the basement meeting wearing jeans and a charcoal sport jacket, he’s … limping? He wears a plastic boot on his left foot.

This could be an awkward moment, but he turns to his people at the front of the room and says, I’m wearing a boot.

They relax. All is well at the top.

So I was in French Polynesia, about 400 miles from Tahiti, he starts. And I’m on the fifth day of scuba diving and I jump out of the boat and I hit something on the way down. I walked around on it for a few days, but then I got an MRI … and turns out, I ruptured my Achilles.

And so on … about his decision to avoid surgery and try a nonsurgical technique involving large needles, meditation, and no anesthesia. One doesn’t have to spend much time in Benioff’s orbit to learn that absolutely everything is a story with him. He’s part of a long line of maggidim, itinerant storytellers unfurling inspiring homilies with a message.ii

Later in the day, after much agent-related discussion, Benioff gets back to that message:

So I’m scheduled for a CAT scan … but is there any follow-up? It would have been nice to have somebody reach out and tell me what to do, update me on what happened, keep me informed about the process. But nothing. It would have been a perfect job for an AI agent.

And bing-bang: "Any company that’s adding an agentic layer is putting an expert by their side to help deal with customers," he says.

An agentic layer. And agents. And of course Agentforce. That’s the point.

Now Agentforce is a way to build, customize, test, deploy, and monitor AI agents. And AI agents are simply very malleable pieces of software that can interact with humans, automate business processes, and make plans and decisions. They’re a form of digital labor, or what Agentforce version 2.0, released the following month, would call A digital labor platform for building a limitless workforce.²

At the time of the emergency conclave, Agentforce is all of four weeks old. It emerges – as most dramatic moments in this founder-led, 25-year-old company do – directly from Benioff’s late-night brainstorms.

As usual for Salesforce’s top-to-top retreats, the Agentforce Kickoff is held over 12-hour days with few breaks. The hotel may have chandeliers, but they are unremarked on, and the spa is a fragrant, lifeless desert.

And there is a consciously cult-like vibe, in the positive sense of Jim Collins’ classic From Good to Great, which said: A cult-like culture can actually enhance a company’s ability to pursue Big Hairy Audacious Goals, precisely because it creates that sense of being part of an elite organization that can accomplish just about anything.³

Just about anything is exactly what needs accomplishing now. It’s one of those moments – a call to action, a decisive point, when ability meets opportunity and a historic advantage is won or lost – and Benioff wants to make sure his team is alive to their chance.

There is no ambiguity here. This time will not return.

All of us have to change our minds and realize, he says, that this is the single biggest opportunity of our lives.

And: This is the single most important piece of technology to come along in the history of business.

Obviously, he’s had an epiphany. He’s seen a future. The development of so-called large-language AI models like OpenAI’s GPT and Google’s Gemini and High-Flyer’s DeepSeek made software conversational and smart. Seemingly overnight, computers could look up information, summarize and organize, make plans and suggestions – do a lot of the things that people do but faster and with better grammar.

But there were still limitations. AI couldn’t really do much work, not the kind you and I do when we’re, well, at work. Salesforce was going to change that. It was going to put AI to work using virtual agents, AI agents that could work alongside humans, making everything easier.

They already exist. The group sits through a live demonstration of an agent that helps a family of four plan and change a trip on the fly to a theme park, asking questions, making reservations and changes, getting real-time updates on ride status – all using slangy American rat-a-tat without talking to a single human being.

And they’re new. Benioff tells a suspenseful story of the build-up to Dreamforce, the company’s annual mega-event that takes over downtown San Francisco, held four weeks earlier. Agents weren’t originally part of his keynote until some customer meetings and an encounter with a prescient tech-startup CTO rerouted his code.

We were doing the [customer] demo, he says, mentioning an impressive agentic AI case study of a European luxe customer that trained its customer-service agents to speak in a luxurious tone of voice, and the CTO said, ‘That might just be the best software ever made.’

So he tells the team to tear up the keynote and go all-in on Agentforce.

There’s another story, of the time he met with Steve Jobs, years ago, and Jobs was launching the iPad. And Jobs tells Benioff a secret that wasn’t really a secret: he only did one thing at a time. Just one.

Benioff decides that from now on Salesforce would only do Agentforce. This is what is called focus.

Two days; 10 in-person customers, talking about their Agentforce adventures on the ground; hands-on Agentforce training for everyone in the room, 500 highly educated, meticulously dressed achievers going back to school to learn to use a tool.

And it’s unsettling how good the software already is. It can already mimic human-like call center agents and sales reps. It can already build marketing plans, write campaign briefs and emails, and create websites with personalized images and thousands of product descriptions.

It can write elegant computer code and document it without complaint. And building an agent using Agentforce is almost as easy as writing this sentence. It does not feel like computer programming but rather conversation, which is the future of software it seems.

Across the two days, there is another message as well.

A strong case is made by the head of sales, the heads of product and engineering, a rising sales star, and various customers and sales reps walking through Agentforce deal recaps and