22,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Fachliteratur
- Sprache: Englisch
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.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 296
Veröffentlichungsjahr: 2025
Ähnliche
Table of Contents
Cover
Table of Contents
Title Page
Copyright
Foreword
The Five Attributes of an AI Agent
Author’s Note
Agentforce Kickoff, San Francisco
Notes
Chapter 1
: Call My Agent!
Notes
Chapter 2
: What Is Agentforce, Anyway?
Notes
Chapter 3
: How Does Agentforce Actually Work?
Giving a Customer-Service Agent More Skills
Customer's POV
Building a Custom Agent
Notes
Chapter 4
: What Are Some Useful Things You Can Do with Agents?
Notes
Chapter 5
: Do You Need a Platform to Do Agents (and What's a Platform, Anyway)?
Notes
Chapter 6
: Are Agents Really Different from Chatbots and Co-pilots?
Note
Chapter 7
: What Are the Different Parts of Agentforce?
Note
Chapter 8
: What Is the Einstein Trust Layer, and Why Do You Need It?
Notes
Chapter 9
: Why Do You Need Data Cloud for Agentforce?
Note
Chapter 10
: What Is RAG, and Why Should I Care?
Notes
Chapter 11
: What Is the Atlas Reasoning Engine?
Notes
Chapter 12
: How Do You Control an Agent and Give It Orders?
Notes
Chapter 13
: How Do You Test an Agent in a Sandbox?
Note
Chapter 14
: What Are Some of the Prebuilt Agentforce Agents?
AI for Service
Agentforce for Sales
Agentforce for Marketing
Agentforce for Commerce
Agentforce for Employee Service
Notes
Chapter 15
: How Do You Build a Custom Agent from Scratch?
Notes
Chapter 16
: What Is the Best Way to Come Up with Ideas for Agents?
Brainstorming AI Agent Ideas
Notes
Chapter 17
: How “Human” Should Your AI Agent Be?
Notes
Chapter 18
: How Do You Make Sure AI Is Governed?
Pillars of Effective Governance
Notes
Chapter 19
: How Do You Build a Business Case for Agentforce?
Notes
Chapter 20
: So What Are We Humans Going to Do Now?
Notes
Chapter 21
: How Can I Get Started with Agentforce and Learn More?
Note
Chapter 22
: So What Was This Book All About, Anyway?
About the Author
Index
End User License Agreement
List of Illustrations
Chapter 3
Figure 3.1 Agentforce topic screen
Figure 3.2 Adding actions to Agentforce
Chapter 5
Figure 5.1 Agentforce and the Salesforce platform
Figure 5.2 Vivek's whiteboard of the Salesforce Platform
Chapter 6
Figure 6.1 Bots versus autonomous agents
Chapter 7
Figure 7.1 Agentforce agentic loop
Chapter 8
Figure 8.1 Einstein Trust Layer
Chapter 9
Figure 9.1 How Data Cloud works
Chapter 11
Figure 11.1 Agentic AI Planner and Data Cloud
Figure 11.2 Atlas Reasoning Engine
Figure 11.3 Atlas Cognitive Map example
Chapter 12
Figure 12.1 Five components of an AI agent
Figure 12.2 Creating prompts with Prompt Builder
Figure 12.3 Prompt Builder home screen
Figure 12.4 Agent Builder: description
Figure 12.5 Agent Builder: topics
Figure 12.6 Adding Agentforce: actions
Chapter 13
Figure 13.1 Testing agents in Agent Builder
Chapter 15
Figure 15.1 AI agent lifecycle
Guide
Cover
Table of Contents
Title Page
Copyright
Foreword
The Five Attributes of an AI Agent
Author’s Note
Agentforce Kickoff, San Francisco
Begin Reading
About the Author
Index
End User License Agreement
Pages
iii
iv
v
ix
x
xi
xiii
xv
xvii
xix
xx
xxi
xxii
xxiii
xxiv
1
2
3
4
5
6
7
8
9
10
11
13
14
15
16
17
18
19
20
21
22
23
24
25
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
45
46
47
48
49
50
51
52
53
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
157
158
159
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
207
209
210
211
212
213
214
215
216
217
Martin Kihn
Foreword by Marc BenioffCEO and Co-Founder, Salesforce
Agentforce
Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry
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, (978) 750-8400, fax (978) 750-4470, 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, (201) 748-6011, fax (201) 748-6008, 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 (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002.
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
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 post-mortems – from everyone on stage, in fact, in a concerted show of spontaneous conviction – that for various reasons, Salesforce itself is uniquely positioned to take the market.
“Almost by accident,” Benioff says at one point, “we built exactly what we needed to do Agentforce.”
He is referring to the Salesforce platform itself, with its secure global infrastructure; its flexible architecture, stressing what’s called metadata and a low-code user interface – above all, the fact that customers are already using it to do things agents liked to do, such as customer service, sales outreach, e-commerce, and digital marketing.
Competitors would notice, of course. Shortly after Dreamforce, which was “all-in” on Agentforce, Microsoft’s CEO Satya Nadella took to the stage himself at Microsoft’s UK sales kickoff to demonstrate an email-writing agent. Within weeks, other competitors like HubSpot and ServiceNow announced their agent-forward strategies.
Yet Salesforce undeniably had momentum. A week after the secret conclave, Salesforce’s stock hit an all-time high. Formerly skeptical reporters turned sympathetic.4 It was a change.
Would it last? Where would the company be in a year? In an agentic future?
“We better go fast,” Benioff tells his transformed, weary team at the end of day 2. “We have a window.”
Let’s start now.
Notes
i
Actually, Salesforce also advertised during the Super Bowl in 2022, but the sentence was more dramatic this way.
ii
In fact, Benioff’s cousin, David Benioff, is a showrunner and TV writer known for co-creating HBO’s
Game of Thrones
.
1
. On reflection, “I Chose Salesforce” gives too much credit to competitors. There really isn’t a choice, Salesforce implies.
2
. See
https://investor.Salesforce.com/press-releases/press-release-details/2024/Introducing-Agentforce-2.0-The-Digital-Labor-Platform-for-Building-a-Limitless-Workforce/default.aspx
.
3
. See Jim Collins, “From Good to Great: Why Some Companies Make the Leap … And Others Don’t” (2001)
https://en.wikipedia.org/wiki/Good_to_Great
.
4
. See
https://www.businessinsider.com/Marc-Benioff-ruptured-achilles-tendon-fakarava-Salesforce-Agentforce-ai-agents-2024-12
. A year earlier, Business Insider was relentless in picking on the company. It was also interesting to see sudden skepticism from the same source about Salesforce’s major competitors:
https://www.businessinsider.com/microsoft-ai-artificial-intelligence-bet-doubts-Marc-Benioff-satya-nadella-2024-11
.
Chapter 1Call My Agent!
Welcome to Agentforce: Harnessing the Agency of AI to Scale, Grow, and Lead Any Industry.
As you have probably guessed, this is a book about something called Agentforce. Agentforce was developed and is offered by Salesforce, one of the largest software companies in the world and the number-one customer relationship management (CRM) provider for many years.
If you've heard about Agentforce already and you're not a Salesforce customer, it's probably because it is the only AI agent platform that's advertised during football games. If you saw the ad with Matthew McConaughey drenched in the rain outside a high-end bistro while a gloating Woody Harrelson calls him “buddy” from a nearby sheltered table – and McConaughey explains that an AI agent could have prevented this dining disaster – well, that's Agentforce. Or if you were hustling through LaGuardia or the SFO airport and saw a billboard proclaiming “Agentforce can help thousands of shoppers with spot-on style recos. Every second.” – that too.
The topic of this book is much bigger than a single product from a single company. It is about nothing less than a total transformation in the way work gets done in the modern enterprise, large or small. It's about how you can actually use the recent, mind-bending advances in the power of artificial intelligence (AI) to act alongside human workers, doing real work like helping customers, nurturing leads, and – yes indeed – finessing dinner reservations.
That's right: AI agents can be used to enhance your human workforce, doing tasks that used to require direct human intervention, extending the scale and effectiveness of your teams. At the same time, agent-enhanced workforces can be much more efficient, delivering the same amount of work for a fraction of the cost. Humans can focus on more useful tasks – it's assumed – leaving sometimes dull, repetitive, and routine functions to their new AI teammates.
It is reliably estimated that some 41% of work in most jobs is repetitive, routine, and uninspiring. Perhaps you can relate? Wouldn't it be better to give such work to software, which doesn't need inspiration or purpose to feel seen?
And AI agents are not just a glorified or marginally improved version of the chatbots and “co-pilots” we've enjoyed in recent years. They are able to do things that bots and co-pilots could not, like remember conversations, find information in messy file folders, ask clarifying questions, and even make decisions – all within carefully curated boundaries that ensure the AI agents don't violate security, norms of conduct, or the reputation of your brand.
If all this sounds too good to be true or even futuristic, it isn't. Already companies – from the employment platform Addeco to OpenTable, from Young Drivers of Canada to Salesforce itself – are using Agentforce to build, test, and deploy AI agents into the field.
And they're already seeing the benefits. “Saving just 2 minutes on a 10-minute call lets our agents focus on strengthening customer relationships,” said OpenTable's SVP of Customer Success George Pokorny.
Now if you're worried this book is just an intricate pitch for Salesforce, rest easy. It is not. It's entirely about a new way of working and the practical day-to-day use of AI agents. That it's about a platform from Salesforce simply reflects the fact that Salesforce is a pioneer in agentic AI, as it was in cloud and other then-new technologies.
So even if you're not in the market for anything Salesforce and are not what the company calls a Trailblazer, one of its 19 million customers and users worldwide, you'll come away with something useful.
Specifically, this book explains in commonsense, McConaughey-like terms:
What an AI agent really is, does, and will do in the future
Different ways agents are being used today
How to get up and running with an AI agent-enhanced workforce
How humans and agents can work together, and like it
Where it's all going so you can prepare your company (and yourself)
Of course, along the way you will also pick up a great deal about how the Agentforce platform works, how it fits within Salesforce's expansive ecosystem, how it's deployed and packaged, and so on. There's plenty in here that applies to the most teal-blooded, golden-hoodied Trailblazer1 … and to the AI novice who just wants to know how to spell retrieval-augmented generation.2
As Tangina said in Poltergeist, “All are welcome.”
Less technical readers may be reassured to learn that your author was a drama major who wanted to be a stand-up comedian and learned data science on the job. More technical ninjas will get a basic grounding in the topic and are referred to resources at the end.
Okay, so how is the book structured? We've noticed that at this stage in its lifecycle, as a relatively new and very fast-moving category, agentic AI (that is, AI agents and their milieu) gives rise to a lot of questions. Customers ask us questions (What is Agentforce? How does it work? Can I build it myself?) and we ask ourselves questions (What do customers need right now? How can we make this easier to use?) and so on.
For this reason, we've decided to structure each chapter as an answer to a question. That way, its contents should be clear. And if you already know the answer, you can move along.
Some of the questions we answer are:
How does Agentforce actually work?
What are some useful things you can do with agents?
How do you control an agent and give it orders?
How “human” should your AI agent be?
So what are we humans going to do now?
And many more.
We've tried hard to keep this book relatively short and jargon-free, so you could actually read it all on the flight from JFK to SFO and still catch an episode of Curb Your Enthusiasm. (We recommend the one where Larry David ties up a hotel desk with detailed feedback on sheets, cookies, etc., a situation where Agentforce could definitely have helped, no doubt inspiring its own hilarious episode on nonhuman workers.)
So let's proceed to our very first question, which is, of course …
Notes
1
. If you know, you know.
2
. A technique for making so-called large-language models like ChatGPT more customized for your particular situation; it's explained in
Chapter 10
.
Chapter 2What Is Agentforce, Anyway?
Agentforce is a platform that provides a safe way to bring artificial intelligence agents into the workplace so they can help people do their jobs. The Agentforce platform includes both out-of-the-box AI agents that are ready to go to work and a suite of tools that let you build your own agents to do almost anything you can imagine.
The goal of Agentforce is to provide a safe – there's that word again – way for companies to add a new kind of tightly defined yet autonomous team of virtual workers into their org charts. These virtual workers, which we call AI agents or just agents, are designed to make the human workers' lives easier by taking on some of their tasks and also doing things they didn't have time to do before.
The ultimate purpose of Agentforce is to make artificial intelligence more useful in the day-to-day flow of work to make it more like a good employee and less like a “science experiment.” Ultimately, Agentforce lets you put AI to work to improve your customers' experience.
Within a few days of Agentforce's launch in the fall of 2024, hundreds of agents had been built and many enterprise-grade agents were deployed, doing things like helping to handle product returns for Saks, managing reservations and loyalty for OpenTable, and dealing with back-to-school textbook order spikes for Wiley.1
Around the time of the launch, a technical reporter noted “Salesforce's push for more autonomous AI tools that handle repetitive tasks without requiring constant user guidance” – a dig at some of the many “AI co-pilots” then in vogue.2
What's truly new about these AI agents is that they can take a set of plain-language instructions – kind of like a job description for a person – and use them as a basis for some very human-like behavior. Interacting directly with customers, via text or voice or even video, they can understand everyday speech, ask clarifying questions, come up with a sequence of steps in an action plan, and put it to work. In other words, these agents can make decisions based on customer requests and act on them without needing direct human intervention.
That's to say they can be autonomous. And lest that word scare some of you, let me assure you that the agents have very clear boundaries about what they can and can't do and say and are supervised by no-nonsense virtual task managers that would make Mr. Burns on the Simpsons look fey.
An important part of the Agentforce platform is precisely this ability to instruct, control, and direct the AI agents. After all, they're supposed to work for you and not the other way around. Agentforce puts the fiction in all that dystopian science fiction like 2001 and She and … well, pretty much any story with AI characters.
Now that they're real, some of you may be wondering why AI agents are causing such a stir. Even Bill Gates recognized that “AI agents could change how we live our lives,” adding that “Clippy has as much in common with agents as a rotary phone has with a mobile device”3 – no doubt overestimating the power of Clippy, Microsoft's notorious on-screen help-bot.
The reason for the excitement is not solely due to advances in technology. After all, AI is not new; even ChatGPT 3.0 was preceded by GPT 2.0 and GPT 1.0. Salesforce opened its office of AI almost 10 years ago and launched its own generative model in 2023.4 And AI features powered 1.4T predictions during a recent cyber week for Salesforce customers alone. No, technology is a crucial but not sufficient explanation.
