Act now to set your future in motion

Transform AI potential into business reality, step by step

Every business leader today faces the same impossible equation: Create exponential growth with static budgets and overextended teams. Beat competitors who move faster every year. And of course, exceed customer expectations that reset higher and higher every day.

These mounting pressures demand a fundamentally new approach to work itself. Enter agentic AI — intelligent AI agents that don’t just follow scripts but reason, adapt, and act on their own. These agents can handle the repetitive, time-consuming tasks that drain your teams while helping employees focus on higher-value work that promotes real innovation and job satisfaction.

These advanced agents are central to the journey to becoming an agentic enterprise — where people and AI work together, achieving more than either could alone.

While this promise of digital labor is exciting, the way forward often feels overwhelming. Where do you start? How do you avoid costly mistakes along the way? What does success actually look like?

We’ve been exactly where you are.

At Salesforce, our own agentic AI journey has been filled with experimentation, breakthroughs, and hard-earned lessons learned. But we’ve also guided thousands of customers, from innovative startups to Fortune 500 enterprises, through their own agentic enterprise evolutions. We’ve seen what works, what doesn’t, and most importantly, the real benefits that come when human workers collaborate with AI agents on everything from sales support to marketing outreach to HR workflows. This guide distills those combined experiences into a practical roadmap that removes the guesswork from how to navigate the era of AI agents.

‎

How to use this guide

Read this playbook sequentially to learn our complete step-by-step roadmap, or jump to the chapter that addresses your immediate needs. You’ll find “lessons from the agentic enterprise” that include wins, pitfalls to avoid, and no-holds-barred examples of Salesforce customers on their agentic AI journeys. At the end of each chapter, you’ll also find downloadable activities for you to use with your team to put learning into practice.

As shown below, the guide moves from crafting a vision and preparing people through tactical execution — use case selection, process transformation, and data foundations — to realizing the “dual dividend” of customer delight and employee success. The final chapter brings ideas and inspiration for how to prepare for a future of advanced agent environments, robotics, and enterprise general intelligence (EGI).

The bottom line? Start now.

Companies that hesitate or approach agentic AI haphazardly will miss opportunities and be left behind by competitors who move decisively. In a March 2025 survey by McKinsey, 78% of respondents said their organizations already use AI in at least one business function, up from 72% in early 2024 and 55% a year earlier. AI adoption is accelerating rapidly year over year.

You’re much closer than you think to realizing the benefits that AI agents will bring to your business, and we’re here to help you get there.

Let’s begin.

Align on a tangible agentic AI strategy

Becoming an agentic enterprise can feel like standing at the base of Mount Everest with a backpack and a prayer. The summit looks impossibly far, and you’re not even sure you have the right gear. But here’s the thing about successful climbers: They don’t focus on the peak first. They focus on understanding their capabilities, mapping their route, and taking that crucial first step. When planned correctly, the success of that first step leads naturally to the next one, and what once felt like an impossible summit becomes a series of achievable milestones.

So, before diving into the technical details of this playbook, it’s important to understand what’s possible when AI agents become part of your organization’s fabric. We’re talking about intelligent AI agents that can take on tasks previously assigned to one employee or even a team of employees. These agents can analyze, decide, and execute, all on their own and within the guardrails you’ve established. Take a moment and ask yourself:

What if your marketing team could launch campaigns in days instead of months?

What if your customer service department resolved issues before customers even noticed?

What if your product development cycle accelerated by 70%?

And finally, what if you could reimagine your organization’s future not through incremental improvements, but through an overall transformation of how work gets done? Through our work with dozens of organizations, we’ve uncovered key insights that can speed up your journey.

Here’s what we’ve learned about envisioning your AI-powered future

Lessons from the agentic enterprise

• Agentic AI is a strategic growth multiplier, not a cost-cutting tool. Focus on building your AI strategy around scaling capabilities, not reducing headcount.
• Establish clear integration discipline early by deciding that every new tool must connect to your single source of truth, otherwise you’ll end up creating data siloes.
• Build your AI strategy on a unified platform foundation from day one to speed up your AI transformation‌ — ‌and business results.

Zota developed a bold AI-first growth strategy

Global payments marketplace Zota faced a strategic dilemma. They were planning for 30% annual growth and aiming to double from 500,000 to 1 million merchants, but CEO Avner Ziv was determined to cap the workforce at 300 employees. Although 180,000 annual support cases cut into their reps’ time for high-value work like closing deals, hiring more people wasn’t an option.

To solve this, Zota developed a bold AI-first growth strategy that scaled their capabilities from the start. Their agentic AI vision was to build a comprehensive digital workforce, starting with one agent and scaling to 30+ agents across every department. In just five weeks, they designed, built, tested, and deployed a merchant FAQ agent using Agentforce. This win was the beginning of their overarching strategy to create a scalable foundation using their unified Salesforce Platform and Data Cloud to power an environment of AI workers.

The result was a fully functional AI agent that could handle 180,000 inquiries annually, provide 24/7 support for 500,000 merchants, and seamlessly capture leads when questions fell outside the knowledge base. This targeted approach gave them immediate business impact while laying the foundation for their broader vision of deploying 10 agents in the first year. Zota’s vision to use AI as a strategic growth multiplier, not a cost-cutting tool, helped them build a digital workforce that will scale their team of 140 to perform like a team of 800.

The agentic enterprises most successful in AI implementations begin with a compelling vision, not with technology selection. This chapter will guide you through:

1. Understanding your agentic readiness
2. Aligning around a strategic foundation
3. Creating a foundational vision
4. Defining your agentic AI vision statement

Agentic AI-readiness

What does it mean to be “AI-ready?” Think of it like preparing for a cross-country road trip — it’s a journey! You need to know your starting point, check your vehicle, and map your route before hitting the highway.

The first step? Understand your current AI readiness. Assessing where your people, processes, and technology stand today will guide that journey — from selecting the right first use case to building a strategic roadmap for organization-wide adoption.

Part of understanding your readiness means knowing exactly where you stand on the agentic maturity spectrum — and more importantly, what your next achievable step looks like.

Salesforce has created a multi-level agentic maturity model to help you understand where you are today and where you can go. This progressive approach allows organizations to build capabilities incrementally while demonstrating business value at each stage.

Agentic maturity: Where are you now?

Level 1 agents deliver significant benefit and lay the foundation for more sophisticated deployments. This is where companies define their AI strategy, establish guardrails, and begin connecting data sources. Think: An AI agent that answers common employee questions like “What’s our vacation policy?” or “How do I reset my password?”

The evolution of AI leadership: No one-size-fits-all approach

Unlike previous technology implementations that could be handled in isolation, agentic AI requires a new kind of leadership collaboration. What successful implementations share isn’t a single decision-maker, but rather joint accountability between technical and business leaders who recognize that AI touches both technological infrastructure and business strategy.

The CIO’s role has evolved significantly in the era of AI agents. Today’s CIOs must flex their business muscles, working closely with functional leaders to identify use cases that deliver tangible value and measurable outcomes. Their role goes beyond managing technical infrastructure — we often see CIOs orchestrating cross-functional teams and translating business needs into technical roadmaps.

Success typically requires collaboration across these key roles.

• CIO/Chief technical officer (CTO): Technical infrastructure, data strategy, and cross-functional orchestration
• Line of business leaders: Use case identification, outcome ownership, and adoption
• CEO: Strategic alignment and decision owner on the size and purpose of the shift to agentic AI
• Chief customer officer (CCO): Customer experience strategy and service automation workflows
• Chief human resources officer (CHRO): Workforce transformation and change management
• Chief financial officer (CFO): Investment strategy and return on investment (ROI) measurement

The critical question isn’t “Who’s in charge?” but rather, “How do we ensure joint ownership between technical capabilities and business outcomes?” Organizations that struggle with AI adoption often lack this collaborative model, treating implementation as either purely technical or purely business when it’s fundamentally both.

The best implementations begin with the right leadership model for your organization’s starting point and ambitions.

Build your AI strategy on these four pillars

While the chapters ahead guide you through the step-by-step journey to becoming an agentic enterprise, four foundational pillars represent an organizational readiness “snapshot.” Having the following in place will lay the groundwork for your AI agent implementation.

1. Strategic agreement
2. Data readiness
3. Technology infrastructure
4. Governance and measurement

Defining your strategic vision

Next, it’s time to begin aligning around the “why.” To succeed with agentic AI, start with a clear vision statement that communicates why your organization is pursuing AI agents. This statement sets the tone for success and should be shared across the organization.

Try this activity: Complete your AI vision statement

Throughout this playbook, you’ll find printable activity worksheets to help your teams put guidance into action today and speed up your agentic AI journey. Download the first worksheet below, and complete the fields with your team to develop your AI strategy and vision statement.

Remove blockers and set your employees up for human-AI agent collaboration

Remember your first day working with a new colleague who seemed impossibly efficient? Initially, you might have felt intimidated or worried about measuring up. Then you realized the best workplace relationships happen when you stop competing and start collaborating — their strengths make your strengths shine even brighter. Integrating AI agents into your company is like introducing the ultimate new team member: a powerful tool that’s highly efficient and always ready, but relies entirely on human wisdom, creativity, and judgment to perform at its best.

Managing this change well is one of the most critical parts of AI success. While setting up the technology is straightforward, only real change can happen when your people, from top to bottom, embrace AI agents as collaborators.

Recent research by McKinsey Digital found an AI readiness perception gap between leadership and employees. C-suite leaders are more than twice as likely to say employee readiness is a barrier to AI adoption as they are to blame their own role. However, the research showed that employees are actually more ready for AI than their leaders imagine, and the biggest barrier to AI success is leadership, not employee readiness.

This misalignment reveals a critical opportunity for leaders to shift from viewing employees as obstacles to recognizing them as willing partners in change. In fact, according to Salesforce’s latest Slack Workforce Index, workers who use AI daily prove to be 64% more productive and 81% more satisfied with their job than colleagues not using AI. When leaders view agentic AI adoption as an opportunity, this unlocks its real potential and an entirely new narrative around what AI means for the workforce.

The transformation mindset

Successful agentic enterprises view agents as an opportunity to elevate human work rather than replace it.

Desk workers report spending 41% of their time on tasks that are “low value, repetitive or lack meaningful contribution to their core job functions.” This research from Slack’s work trends report shows a clear opportunity for AI to help workers refocus their energy away from busy work toward more valuable activities. When leaders position AI as “digital labor” that handles routine tasks, they free their workforce to focus on work that requires human capabilities. This is about creating a compelling vision that addresses fears head-on while inspiring enthusiasm for an enhanced future.

This mindset must be actively championed by leadership and reinforced through consistent messaging, success demonstrations, and recognition of employees who work effectively with agents. Leadership buy-in from cross-functional areas of the business is incredibly important for a holistic approach — including technology leaders leading implementation, business leaders tracking strategic objectives, and HR leaders advising on career pathways and learning programs.

Here are a few practical approaches for leadership championing this mindset.

• Develop pathways for high performers to become “citizen developers,” encouraging them to build and implement agents to tackle repetitive tasks.
• Host regular “AI success showcases” where teams demonstrate what’s working.
• Incorporate AI collaboration skills into performance reviews and promotion criteria.
• Create internal communities where employees share best practices.
• Establish recognition programs that celebrate employees who work with agents in new ways.
• Have leaders openly discuss their own AI learning experiences, including successes and challenges.

Lessons from the agentic enterprise

• Be transparent from day one by offering free AI training, holding company-wide learning sessions, and creating dedicated communication channels for regular updates; otherwise, employees will fill knowledge gaps with fear and speculation.
• Create safe spaces to experiment by providing hands-on practice sessions in low-stakes scenarios and frame failures as learning opportunities; otherwise, teams will avoid using AI agents when they could be most helpful.
• Build clear decision frameworks by defining specific criteria for when to use AI agents versus human judgment; otherwise, employees will either over-rely on agents or avoid them entirely.
• Establish feedback loops that show impact by asking employees to report agent limitations and improvements based on their input; otherwise, they’ll lose trust in both the technology and leadership’s commitment to making it work.

The 4 R’s: How Salesforce built the blueprint as Customer Zero

As the first user of our own AI tools, we’ve learned that successful change management requires both strategic vision and tactical execution. We’ve taken a comprehensive approach to preparing our own employees for agent collaboration using what we call “the four R’s”:

• Redesigning jobs around human-agent collaboration
• Reskilling through programs like quarterly learning days
• Redeploying talent to high-growth areas through our AI-powered career platform
• Rebalancing the workforce by continually optimizing the mix of human and digital labor

This cultural change prepares every employee to manage agents or teams of agents, recognizing this is the last generation of leaders to manage only humans.

Redesigning jobs

Combining the best of AI and human capabilities will require new roles, AI agents, and working models. This involves a delicate balance of infusing AI into existing workflows for efficiency (for example, tasks like summarization or data querying), while also designing entirely new roles for humans like “redesign lead” who is responsible for driving large-scale organizational design.

Early on in our journey, Salesforce established a dedicated Workforce Innovation team focused on designing and implementing strategies to prepare the global workforce for the AI era, including optimizing talent mobility and driving internal AI adoption.

Organizations must rethink their structures, roles, and processes to maximize the complementary strengths of humans and AI agents.

Reskilling employees

Most of today’s workforce will need to brush up on topics like AI basics or data strategies to thrive in an agentic environment. At Salesforce we’ve been intentional about teaching, practicing, and adopting these AI skills every day. We started by identifying the top 10 enterprise skills that workers need to be successful in a future with agents; which fell across three broad categories:

• Human skills: Adaptability, accountability, collaboration, emotional intelligence
• Agent skills: Agent and AI literacy, and human/agent collaboration
• Business skills: Problem solving, data interpretation, creative thinking, storytelling

Here’s a deep dive on how some of these skills translate into day-to-day work in the era of digital labor:

AI literacy: Beyond basic understanding, employees need practical knowledge of:

• How to effectively communicate with AI agents: Learning to write clear, specific prompts, setting clear constraints, iterating with examples, and understanding how different phrasing affects AI responses. For example, asking “What are the top customer complaints this month?” versus “Show me customer feedback data” yields vastly different results.
• Understanding AI capabilities and limitations: AI agents excel at pattern recognition and data processing but struggle with nuanced judgment calls. Common limitations include hallucinations (confidently presenting false information), sycophantic behavior (agreeing with users even when they’re wrong), and difficulty with multi-step logical reasoning or complex calculations.
• Recognizing when AI outputs need human verification: High-stakes decisions, creative content requiring cultural sensitivity, or recommendations involving regulatory compliance should always undergo human review.

Systems thinking: As agents handle individual tasks, humans must develop:

• Ability to see connections across complex workflows: For example, understand how automated inventory decisions affect supply chain partners.
• Understanding of how agent decisions impact downstream processes: For instance, if an AI agent approves a customer refund, employees need to track how this affects accounting, inventory, and customer lifetime value calculations.
• Capacity to identify when process redesign is needed: Recognize when workflows built for human execution need restructuring for AI agents.

Ethical decision-making: Humans must serve as the ethical backstop for agent activities.

1. Judgment on when to override agents: For example, when an AI agent suggests denying a loan application, humans must evaluate whether the decision reflects algorithmic bias or legitimate risk assessment.
2. Ability to detect bias: Recognize when AI recommendations consistently favor certain demographics or when training data limitations create unfair outcomes.
3. Understanding of privacy and data governance issues: Know when agent access to sensitive data crosses compliance boundaries and how to maintain audit trails for regulatory purposes.

Salesforce’s approach to reskilling? Start with culture

At Salesforce, getting employees comfortable with agentic AI starts with a culture of openness and experimentation. “It must come from the top-down and the bottom-up,” said Irina Gutman, RVP of global AI practice, professional services. “You also want to identify your change agents, those who advocate for the technology from within. It’s not just my boss telling me that it’s good for me. It’s my friend and colleague who I respect who’s sitting next to me saying, ‘You know what? This is really cool. You should try.’”

At Salesforce, we’re focused on making reskilling easy for employees. Resources like Quarterly Agentforce Learning Days encourage employees to spend company time developing their agentic AI skills. Last year, we introduced Career Connect, an internal AI-powered talent marketplace that gives employees tailored and personalized recommendations, including new skills to learn, online training, and even new job opportunities within the company. Read more about our approach to preparing our own employees for agentic AI.

Ready to begin reskilling? Tools like Trailhead, Salesforce’s free online learning platform, helped our team learn fast and adapt faster. Check out these trails for a foundational (and fun!) place to start.

Redeploying employees to high-growth areas

The most incredible opportunity that agentic AI offers businesses today might be this: As agents take on routine tasks, you can move talent to emerging roles that leverage uniquely human capabilities alongside AI. At Salesforce, we’re elevating our people into positions where they can influence greater impact and innovation.

Our redeployment strategy centers on three key initiatives.

1. Career Connect platform: Our AI-powered internal marketplace helps employees identify and develop new skills. The response has been strong, with 44% of employees signed up and nearly 40% of internal job applications flowing through the platform.
2. Future skills framework: We’ve identified the top 10 types of skills workers need for a future with agents — human, agent, and business skills — creating clear learning pathways for our entire workforce.
3. Workforce innovation team: We created a dedicated team to execute this vision, ensuring employees can meet evolving demands in an AI-enhanced workplace. This effort has resulted in 28,000 employees becoming certified as Agentblazers. These are pros who have the skills and vision to transform how work gets done and keep pace with innovation in the agentic AI era. Learn more about our Agentblazer certification program.

Rebalance your workforce with human-digital labor

Achieving the optimal blend of digital and human labor requires ongoing refinement. As agent capabilities evolve, businesses should continuously monitor their performance using workforce planning tools. A data-driven, dynamic approach to task allocation will ensure that agents manage more routine duties while human talent focuses on higher-value, engaging work.

The agentic enterprise: Humans and agents, together

Digital labor is as much about preparing people as it is about technology. By redesigning workflows thoughtfully and creating genuine learning opportunities, you build the foundation for sustainable AI adoption that benefits everyone. Now it’s time to put these principles into practice by mapping out exactly how you’ll optimize adoption and success across your organization. This can’t happen in isolation — it requires genuine stakeholder engagement from the start.

Try this activity: Build an alignment plan

We designed an activity to help you think through stakeholder perspectives before engaging them directly. Use it to build empathy and prepare thoughtful questions before holding conversations with each stakeholder group.

Apply AI where it’s needed most

Remember your first time behind the wheel? You didn’t start on the freeway during rush hour. You likely began in an empty parking lot, mastering the basics before taking on more complex challenges. Choosing your first AI use case requires the same wisdom. Begin where you can build confidence, learn the fundamentals, and create momentum for bigger journeys ahead.

This is the difference between agent success and disappointment: your first use case. We’ve seen hundreds of organizations navigate this choice, and those who select thoughtfully create momentum that transforms their entire business, while others struggle to get their second agent off the ground.

You don’t need to reinvent the wheel. Identify use cases that deliver immediate value and simultaneously build the foundation for broader transformation. This chapter will guide you through that process, removing the guesswork and setting you up for sustained success.

Your strategic starting point

Not all use cases are created equal. The art is to find that sweet spot where business value meets implementation feasibility, creating “credibility-building wins” that prove agentic AI’s value to your organization.

When determining your first use case, start with four essential questions that separate successful implementations from failed experiments.

1. What specific business goal will this address? Focus on use cases tied to measurable problems, not abstract possibilities.
2. What’s your organization’s risk appetite? If you’re in a highly regulated industry with low risk tolerance, consider internal agents first to build confidence.
3. What data does the agent need to succeed? The simpler and cleaner your data requirements, the faster your path to value.
4. How will teams actually engage with this agent? Clear deployment plans prevent agents from becoming expensive demos.

These questions might seem basic, but they’re where most organizations stumble. By getting clear answers up front, you’ll avoid the costly pivots that derail promising initiatives.

Here’s what we’ve learned about beginning with the right use case

Lessons from the agentic enterprise

• Start with a small and focused use case and resist the urge to implement AI everywhere at once. Starting with one clear use case builds value over time, creating tangible and exciting proof points for broader adoption.
• Structure your AI agent topics broadly rather than creating lots of narrow, similar options to help AI make faster, smarter decisions without second-guessing itself.
• Start small and learn fast. Deploy agents to 5% to 10% of your traffic initially. That gives you real customer data while allowing you to monitor and adjust. Then consider applying to more traffic, but only after proving the agent can handle the volume responsibly.

Indeed’s quick win is paying off in the long term

Jobs marketplace Indeed had a frustrating problem when employers tried to post jobs. They often got stuck dealing with basic issues like blurry uploaded documents or missing information, forcing employers to wait for human support to fix simple problems like a job description that was too short.

To solve this, Indeed started small and focused, resisting the urge to implement AI everywhere at once. Instead, they chose one clear goal: help support teams handle employer verification and job-posting issues faster. They trained Agentforce to instantly spot problems with job postings, like overly long job titles or missing details. Agents didn’t just explain what was wrong — they suggested fixes and even updated the posting in real time so employers could keep moving.

Now, instead of being stuck answering basic questions, Indeed’s human team can focus on building real relationships with customers while supporting their ambitious goal to reduce hiring time by 50% and help more people find jobs faster. Read the full Indeed story.

Travel platform Engine faced a costly problem

Engine’s 150-person support team handled 530,000 requests a year, with simple cancellations eating up valuable time that could be spent helping customers with more complex travel changes.

To tackle this, Engine partnered with Astound Digital and built their first AI agent in just 12 days, starting with customer cancellations, using a test-learn-and-scale approach. During implementation, they discovered that when AI has too many similar options, it gets confused. So, instead of creating separate topics for cancellations, upgrades, and adding drivers, they simplified everything under broader topics like “car management” with multiple actions underneath. Structuring AI agent topics broadly rather than creating lots of narrow, similar options helps the AI make faster, smarter decisions without second-guessing itself.

Now, Engine’s AI agent, Eva, handles cancellations autonomously in seconds, reducing average handle time by 15% and freeing up human agents to focus on complex requests while supporting Engine’s goal of 70% YoY growth without adding overhead. “Agentforce enables our people to work more effectively,” said Joshua Stern, director of go-to-market (GTM) systems at Engine. “It’s not a replacement for them. It’s a way to maximize their efficiency so that they are not stuck doing repetitive tasks. The real power is how we use that data to create a holistic picture and offer the best support to customers. It’s about meeting customers where they are.” Read the full Engine story.

The use case selection framework

We’ve developed a practical framework that maps agentic AI opportunities along two critical dimensions: implementation complexity and business impact. This visual tool has helped hundreds of organizations prioritize their agentic AI investments with confidence.

Taking a look at the above graphic, from bottom left, clockwise, the tool features:

Quick wins (low complexity, focused impact)
Best for organizations beginning their agentic journey. These use cases leverage existing data, involve limited stakeholders, and deliver clear, measurable outcomes. Example: an internal agent handling employee HR questions and onboarding workflows.

Scaling capabilities (low complexity, broad impact)
Perfect for organizations with established data foundations. These use cases leverage existing infrastructure but affect multiple departments or customer touchpoints. Example: customer service agents handling product inquiries across all business units, or sales development representatives qualifying leads across different geographic regions.

Strategic foundations (high complexity, focused impact)
Ideal for organizations with AI experience ready to invest in infrastructure. These require significant data integration but deliver clear, transformative outcomes. Example: an intelligent underwriting agent that integrates medical records, credit histories, and risk assessments to automate insurance approvals.

Transformational initiatives (high complexity, broad impact)
Reserved for organizations with mature AI capabilities and strong executive sponsorship. These initiatives require substantial resources but can fundamentally reshape business models. Example: an autonomous procurement agent that negotiates contracts, manages vendor relationships, and optimizes supply chains.

A word about unified data, the AI implementation shortcut

As you’ll learn in detail in Chapter 5, analyzing your data quality is just as important as thinking through the right use case. Leapfrog ahead in your implementation with a platform that already deeply unifies your data for agentic AI. Agentforce — Salesforce’s platform for building AI agents — makes it easier to build and deploy agents with low- and no-code templates. It also works with both internal and external data (and both structured and unstructured data, too). Learn more about how unified data is helping many Salesforce customers fast-track their agentic AI implementation.

A 3-step framework for implementation success

By now you know how important use case selection is to success. From our work with customers, we’ve developed a quick three-step process to help your implementation team choose a use case.

Measuring what matters

Success requires more than good intentions — it demands rigorous measurement from day one. Organizations that track both efficiency gains and business impact create compelling narratives that build support for expansion.

Focus on metrics that resonate with different stakeholders.

• For executives: ROI, revenue impact, cost savings, competitive advantage
• For operations teams: Time savings, error reduction, capacity expansion
• For customers: Response times, resolution rates, satisfaction scores
• For employees: Task automation, skill development opportunities, job satisfaction

Plan for both leading and lagging indicators

Some benefits may take time to materialize, especially those tied to longer business cycles. For example, Salesforce’s web agent is designed to send more qualified leads to sales and handle general inquiries, but evaluating lead quality requires tracking whether pipeline is created faster or deals close faster. These metrics can take months to validate.

In these cases, identify both your long-term goals and the leading indicators that signal you’re on the right track. Early signals might include engagement rates, escalation patterns, or preliminary feedback scores that suggest the desired outcomes are developing, even before final results are measurable.

Remember: Your first agent implementation will help to prove the value of agentic AI as well as solve the immediate problem. Choose metrics that tell a compelling story about agentic AI’s potential for your organization and balance quick wins with patience for longer-term impact.

Try this activity: Select the right use case

Now it’s time to apply these principles to your organization. This collaborative exercise works best with a diverse team representing different perspectives across your business.

Preparation: Gather six to eight stakeholders from various functions (operations, customer service, sales, marketing, IT, and executive leadership). Block a few hours for deep discussion and prioritization.

From automation to intelligent workflows that adapt

Remember when driving somewhere meant printing directions and hoping for the best? Miss one turn or hit unexpected road construction and you’re lost! Today’s GPS adapts in real time, giving you directions and routing around problems when the original path is blocked. That’s the difference between traditional automation and intelligent workflows: One follows rigid instructions and the other focuses on getting you to your final destination.

While traditional automation has delivered significant efficiency gains, it has a fundamental limitation: It can’t adapt to changing circumstances or unique scenarios without human intervention. Agentic AI is designed to adapt, with dynamic learning systems that can reason, plan, and help us evolve how we approach business processes.

Agentic AI focuses on desired outcomes rather than fixed procedures. Instead of programming every possible scenario, you define what success looks like and allow the agent to determine the optimal path.

Consider the traditional approach to a customer service interaction.

Procedure-driven: A chatbot follows a fixed sequence of scripted questions and decision trees, frustrating customers with unique needs.

Outcome-driven: An agent understands a customer’s request and automatically asks further questions, pulls documents, and takes action to resolve the situation.

The outcome-driven approach enabled by agentic AI delivers several significant advantages.

• Adaptability to exceptions: Instead of breaking when encountering unforeseen scenarios, intelligent workflows can recognize novel situations and adapt.
• Continuous improvement: By focusing on outcomes, systems can learn from both successes and failures to refine their approach over time.
• Reduced cognitive burden on employees: Workers are freed from rigid procedures to pitch in where human capabilities are needed, such as a complicated service call.

How AI agents reason

AI agents can reason dynamically in response to real-time data and context, similar to humans in many ways. However, there’s a key difference: AI agents operate deterministically — given the same inputs, they’ll produce the same outputs every time. While you can improve an agent’s performance by updating its training data, instructions, or the underlying model, the agent itself doesn’t learn from individual interactions the way humans do. Each interaction is processed independently based on the agent’s current configuration.

Core steps of AI reasoning

To understand how this works in practice, let’s explore the fundamental steps that enable agentic AI to reason.

1. Understanding: Agents gather and interpret data from relevant sources, using context to comprehend the problem or request at hand.
2. Reasoning: Agents process the information using large language models (LLMs) to analyze context, evaluate options, and make informed decisions about the best course of action.
3. Planning: Based on their analysis, agents develop structured plans to achieve the desired objective, considering available resources and constraints.
4. Coordination: Agents communicate their plans with users or systems to ensure agreement and promote collaborative decision-making.
5. Acting: Agents implement their plan and execute the necessary actions across relevant systems or interfaces.
6. Adaptation: Agents assess outcomes and can incorporate feedback to improve future performance when their instructions or underlying data are updated.

This six-step reasoning process — from understanding to adaptation — transforms from abstract capability to real-world business impact when applied thoughtfully to our customers’ challenges.

Lessons from the agentic enterprise

• Focus on what you want to achieve rather than scripting every step; otherwise, your workflows will break when they encounter situations you didn’t anticipate.
• Don’t over-engineer workflows upfront. Begin with basic requirements and allow your agent to learn from real usage patterns; otherwise, you’ll build overly complex systems that cause user drop-offs and miss opportunities for improvements.
• Create ways for agents to improve from both successes and failures, and for humans to provide feedback on agent performance. Just like your human workforce, you want your digital labor to get smarter, not stay static.
• Define precisely when complexity, stakes, need for empathy, or time sensitivity should transition control from agent to human; otherwise, you’ll either over-rely on AI for sensitive decisions or waste human time on routine tasks.
• Ensure humans receive all necessary context and information when taking over from agents; otherwise, valuable time gets lost reconstructing what already happened and decisions get made with incomplete information.

To learn how reasoning capabilities work in practice, consider how healthcare provider Precina is applying agentic AI to one of medicine’s most complex care challenges. Then, read how Absa is effectively implementing adaptive workflows.

Precina’s agent-first approach to transforming diabetes care

Precina faced a healthcare challenge that seemed impossible. They had an ambitious goal to transform diabetes care from patients seeing their doctor every few months to providing personalized, daily support that could actually change lives. Traditional healthcare workflows were rigid, allowing patients to receive care every 90 days, but the real work of managing diabetes happens every day between visits.

To solve this, Precina built intelligent workflows with Agentforce that could adapt to each patient’s unique needs, with AI handling routine therapeutic interventions while clinicians focused on complex cases requiring human judgment and empathy. Their pilot program created clear handoff points between AI and humans, allowing the system to reason through daily patient interactions and escalate appropriately.

Results from their pilot were promising. In just 12 weeks, 50 patients in rural Louisiana saw their A1C levels drop to 6.4% from 9.6% — three times better than traditional annual improvements — while saving an estimated $80,000 per year for every 5,000 patients.

“In my lifetime, I want to help a billion people improve their health and have a better life because of the work we’re doing,” said John Oberg, CEO of Precina. “With Agentforce, that dream of helping a billion people is a reality.”

Effective implementation of intelligent workflows requires thoughtfully designed human-machine partnerships that leverage the strengths of both. Watch the video below to learn more about their transformation, and go deeper into Precina’s complete agent-first approach.

Absa Group blazes new trails in agent-first banking with Agentforce

Absa faced a scaling challenge. Their menu-driven chatbot could only handle simple questions from their 10 million customers across 15 African countries. That left customers confused about which of the dozens of business banking products fit their needs, potentially leading to them applying for wrong loan types.

To solve this, Absa deployed Agentforce to provide personalized, real-time assistance that understands conversational requests, asks follow-up questions to uncover customer needs, and delivers precise guidance. Built on their unified Salesforce platform with Data Cloud, Agentforce taps into customer profiles, financial accounts, and interaction history to explain product details in plain language and guide customers step by step through applications.Now Agentforce expects to double the impact of Absa’s 5,000 contact center reps, resolve charge disputes 88% faster, and automatically handle 50% of fraud cases. By making AI a seamless part of every interaction, Absa is opening doors of financial empowerment across Africa. Read more about the Absa story here.

Anticipate technical challenges

Implementing agentic processes comes with technical challenges — integration, governance, and resource requirements are a few we’ve experienced ourselves. Consider addressing named challenges proactively with these five core architectural principles.

1. Scalability: The platform must handle growing computational demands as usage expands.
2. Flexibility: The architecture should support seamless integration with existing systems and adaptation to evolving AI capabilities.
3. Data accessibility: Agents need reliable access to accurate data sources, including databases and APIs.
4. Trust: Implement guardrails, evaluations, and observability to ensure reliability and continuous improvement.
5. Security and compliance: Strong measures to protect data with granular access controls and monitoring.

Shift your thinking from steps to outcomes

The most fundamental change required for implementing intelligent workflows is shifting your thinking from “What steps do we follow?” to “What outcome are we trying to achieve?”

To make this shift in your organization, complete the following worksheet activity to reimagine a critical business process around outcomes rather than procedures.

Try this activity: Design outcome-focused workflows

To help you reimagine your business processes around outcomes rather than procedures, we’ve created the Outcome-Focused Workflows activity. This guides you through the essential steps of creating business processes that think.

Complete these statements.

Turn your data into a competitive advantage

Think of your organization’s data as a vast, sprawling library accumulated over decades — some books meticulously catalogued and shelved, others scattered in dusty boxes, and still others written in languages few can read. Just as a world-class librarian doesn’t simply throw all books onto shelves and hope scholars find what they need, creating a comprehensive data foundation for AI agents requires the same thoughtful preparation and intentional organization.

Agentic success means providing access to clean, reliable data — both structured data like patient records or transaction details, and unstructured data like video files or social media posts. But simply having access isn’t enough. The quality of agents’ work depends heavily on how accurately they can find and use the right information.

This is where ontology — the structured map of how information relates — becomes crucial. Just as a master librarian understands not only where each book is shelved but also how every concept within connects to others across the entire collection, your AI agents need a rich understanding of how your customer data relates to your product information, sales processes, and broader business outcomes. Without this metadata framework, even the most advanced AI is like a speed-reader in a library with no organizational system, capable of finding individual facts but unable to generate the meaningful, contextual insights that increase real value.

Where to start: A unified knowledge foundation

Creating a comprehensive data foundation for AI agents requires structured and unstructured data from across your organization. This process involves more than simply connecting systems.

Structured data preparation:

While structured data (information organized in tables, databases, or spreadsheets) is typically better managed in organizations, challenges remain in enabling AI models to understand rows and columns effectively. The “text-to-SQL” task — translating natural language questions into database queries — requires specific preparation:

1. Semantic mapping: Create clear metadata descriptions for your database schemas that explain not just field names but their business meaning and relationships (e.g., “customer_id” links to “Customer records in CRM system”)
2. Query patterns: Document common query patterns your business uses to answer specific questions (e.g., “How do we typically calculate customer lifetime value?” or “What data points indicate churn risk?”)
3. Data validation: Implement consistent validation rules to ensure data accuracy and completeness (e.g., required fields, data format standards, acceptable value ranges)

Unstructured data preparation:

IDC predicts that by 2028, 90% of the estimated 400 billion terabytes of global data will be unstructured, creating massive headaches for IT departments. Unstructured data (PDFs, images, videos, emails, chat transcripts) often contains valuable information but requires additional preparation.

1. Content extraction: Use AI-powered tools to automatically extract text, entities, and relationships from various file formats (PDFs, Word docs, images, etc.), making previously locked information searchable and usable.
2. Semantic organization: Apply consistent tagging and categorization to make content discoverable (for example, tagging customer support transcripts by issue type, product, or resolution status).
3. Versioning control: Establish clear processes for handling document versions and updates to ensure agents always access the most current, authoritative information.

Connecting data sources:

Creating a unified knowledge foundation requires connecting these structured and unstructured data sources effectively. To do this, a number of methods are typically used, including:

• API-first approach: Developing standardized APIs for accessing structured and unstructured data, ensuring consistent data retrieval methods across all systems
• Embedding strategies: Implementing consistent vector embedding approaches across data types to convert all data — text documents, images, or database records — into a common “language” that AI can understand and compare. Embeddings convert text, documents, and data into numerical patterns that help AI agents understand meaning and find related content
• Integration patterns: Establishing clear, repeatable patterns for connecting data across systems, including real-time synchronization protocols and data transformation rules

Lessons from the agentic enterprise

• Invest in semantic organization and unified data foundations from the start; otherwise, your agents can find individual facts but can’t generate the meaningful insights that influence real value.
• Ensure data quality across every department as your agent is only as good as your content. If your data’s a mess, your agent will be confused. Be especially vigilant for content collisions, where multiple articles say different things about the same topic.
• Use decades of past decisions and outcomes to train your agents rather than starting from scratch; otherwise, you’ll miss the institutional knowledge that can dramatically improve decision quality.
• Build data foundations that can support broader use cases than your first implementation; otherwise, you’ll limit the unexpected value that clean, connected data can bring.

Big Brothers Big Sisters of America makes lifelong matches with Agentforce

Big Brothers Big Sisters of America (BBBSA) faced a data challenge. With over 30,000 young people on the Big Brothers Big Sisters BBBS wait list 30,000 children waiting up to four years for mentors, BBBS’s match specialists were overwhelmed by unstructured free-text data from applications and interviews, making it nearly impossible to identify optimal pairings quickly.

To solve this, the mentoring network built their “Matchforce” platform on Salesforce and used Agentforce with Data Cloud to extract insights from decades of matching data, analyzing similarities in location, interests, family history, and preferences to generate compatibility scores and predict match duration. They transformed scattered text into viable recommendations that specialists can filter by radius and compatibility scores.

Now BBBSA expects to cut matching time in half while improving quality and reducing the average cost per match, helping more youth find life-changing mentors faster.

“We’re trying to reach more young people and we’re trying to do it faster. But we also need to make sure the quality of matches increases,” says Travis Gibson, CTO of BBBSA. “Agentforce is helping streamline our process and elevate our match recommendations at a rate humans just can’t do. Every day matters that a child doesn’t have a mentor, and we expect with Agentforce to be able to cut that time in half.” Read more about the Big Brothers Big Sisters of America Agentforce Story.

Good360 faced a critical data matching challenge

With only two employees coordinating disaster-related donations, disaster recovery charity Good360 had to manually search through their database of tens of thousands of nonprofit partners, calculate shipping distances, and verify needs. This tedious process constrained their ability to accept donations when every second counts in disaster recovery.

To solve this, Good360 built a resource-matching agent powered by Data Cloud that harmonizes data from Nonprofit Cloud and NetSuite, instantly analyzing donor, partner nonprofit, community, and logistics data to generate curated match lists.

Their clean, well-organized data foundation became their competitive advantage — and the key to their Agentforce implementation. When searching for tents during the 2025 California wildfires, the agent not only found tents but also surfaced work boots, recognizing additional needs they hadn’t considered.

Now Good360 expects to route disaster donations 3 times faster while saving more than 1,000 hours annually and reducing their carbon footprint by 20%. Read more about how Good360 is finding success with Agentforce.

See how Good360 Uses AI Agents to Aid in Disaster Response and Recovery

Governance and trust

For AI agents to work at scale, they need secure connections with enterprise data and unified governance to manage their access, similar to how you control employee access. Here’s what to keep in mind in establishing governance systems for AI agents in your business.

Data access controls: Implement the following governance measures:

1. Role-based access: Define clear roles for agents just as you would for employees.
2. Attribute-based access: Control access based on data attributes and classification.
3. Purpose limitation: Restrict data usage to specific, documented purposes.
4. Audit trails: Maintain comprehensive logs of all agent data access.

Privacy safeguards: Ensure appropriate privacy protections:

1. Data minimization: Limit agent access to only the data necessary for their function.
2. Anonymization/pseudonymization: Apply appropriate techniques for sensitive data.
3. Retention policies: Implement clear data retention and deletion procedures.
4. Consent management: Ensure proper consent tracking for data use.

Security measures: Implement these security best practices:

1. Encryption: Protect data in transit and at rest with appropriate encryption.
2. Multi-factor authentication: Apply strong authentication for sensitive system access.
3. Backup and recovery: Ensure comprehensive backup procedures for all data.
4. Security awareness: Train teams on security practices for AI systems.

Data quality imperatives

Data quality has always been important, but agentic AI raises the stakes significantly. Poor data quality doesn’t just create inefficiencies — it can lead to AI agents making actively harmful recommendations or decisions. The more grounded and reliable, the better! Use the following checklist when considering your agentic AI data strategy.

Remember: In the agentic AI era, the quality of your data determines the quality of your outcomes. By investing in a comprehensive data strategy, you’re not just improving your agents — you’re building a foundation for sustainable competitive advantage.

Try this activity: Score essential data sources

Use the Essential Data Sources activity to assess your data readiness and identify the highest-impact improvements you can make today.

The dual dividend: Empowered people, delighted customers

An outstanding symphony doesn’t replace musicians with technology — it gives them better instruments. The violinist still brings passion and interpretation, but now they’re playing a Stradivarius — a legendary instrument known for its richness and clarity, making every note more powerful and expressive. In the same way, AI agents don’t replace your people. They become precision instruments that amplify human creativity, empathy, and strategic thinking to create experiences your customers will never forget.

When done right, agentic AI delivers the “dual dividend” — simultaneously unlocking the human potential of your employees while transforming your customers’ experiences. This is the essence of digital labor: Rather than choosing between efficiency and empathy, two value drivers of excellent customer experience, digital labor allows us to amplify both by creating a seamless alliance between human expertise and AI capabilities.

As we explored in Chapter 2, agentic enterprises don’t fear human-agent collaboration — they strategically design it. While that chapter laid the groundwork for organizational change management, this chapter reveals the business outcomes that emerge when companies thoughtfully integrate humans and agents as partners.

Companies that grasp this duality don’t just automate processes — they redesign entire business relationships, creating value for customers while elevating their human workforce to more strategic, fulfilling roles. While it hasn’t been easy, at Salesforce we’re proud of what we’ve learned about increasing customer loyalty and human potential.

The dual dividend in action: Salesforce Help portal’s transformation

The power of this dual approach extends beyond individual customer interactions. At Salesforce, we recently leveraged Agentforce to transform our help site into an agent-first experience, moving from reactive support to proactive success.

“Six months ago, our help portal was typical — documentation, search bars, frustration for our customers,” explained Bernard Slowey, senior vice president of digital success and team leader for the project. “Today, we’ve had over 1 million conversations with our agent, achieving an 85% resolution rate.” Thanks to Agentforce — and the change management, upskilling, and holistic adoption of a new way of working — we’re seeing the “dual dividend” play out in action.

• For employees: Human reps shifted from handling routine password resets to driving product adoption, conducting strategic account planning, and building deeper customer relationships.
• For customers: They had 24/7 availability across multiple languages, instant responses to common questions, and seamless escalation when needed.

“The real story isn’t just about technology, it’s about what happened to our people. Our support engineers aren’t losing their jobs — they’re gaining leverage,” Slowey noted. “They’re becoming customer success partners, proactive advisors, and strategic consultants. The agent handles the routine; humans handle the relationships.”

The ‘heart and head’ approach

How’d we accomplish this? The evolution required moving beyond what Slowey calls “old-school bots” — chatbots that follow preset scripts — to AI agents that can reason, adapt, and provide contextual responses that feel genuinely helpful.

Slowey’s team discovered that successful agentic AI requires both technical knowledge (the “head”) and emotional intelligence, empathy, and soft skills (the “heart”). Initially, they focused solely on feeding their agent the right content and measuring accuracy. But they quickly learned that customer service is fundamentally human, even when delivered digitally.

“We realized we were coaching our agent like an old-school bot — if this, then that,” Slowey recalled. “The breakthrough came when we started coaching it like a human employee. We gave it values, empathy training, and a service mindset. Just like our human support engineers, it needed to understand not just what to say, but how to say it.”

Smart guardrails, smarter handoffs

Slowey’s team spent hours defining the guardrails about what the agent should not handle, such as complex billing issues and pricing discussions. But equally important — and a technical challenge in its own right — was designing the whens and hows of an agent transferring a customer to a human representative.

Immediately showing the employee the summarized issue, the background, and the customer’s information — and doing it seamlessly — was a crucial step in the process. It was important that the customer not have to repeat information or start over.

This approach exemplifies the dual dividend: Customers get continuity and faster resolution, while humans focus immediately on problem-solving rather than information gathering.

Key metrics that matter

Slowey’s team tracks both efficiency and experience metrics.

• Resolution rate: 85% resolution of conversations that are started (for example, the chat window conversations) are closed without handing over to a rep.
• Customer satisfaction: This is measured through post-conversation surveys.
• Handoff rate: This is currently at 4% (up from 2% as they made escalation easier).
• Content quality: Continuous monitoring is in place to eliminate “content collisions” where multiple sources conflict.

“We look at our scorecard daily,” Slowey explained. “Someone in your business needs to own this performance, tweaking prompts and improving responses just like you’d coach a human employee.”

Your dual dividend implementation guide

Here’s your tactical roadmap to increasing customer loyalty and human potential, according to Slowey.

Lessons from the agentic enterprise

• Technical accuracy alone isn’t enough. Agents need emotional intelligence and a service mindset to truly delight customers. Train AI agents like human employees (with values and empathy) to create more authentic customer interactions.
• When agents escalate complex cases to human experts, they must provide complete case summaries and preliminary analysis. Context influences both faster resolution (customer benefit) and immediate problem-solving focus (employee benefit).
• Measure both sides of the dividend by tracking customer metrics (wait times, resolution rates) and employee metrics (time spent on strategic work, job satisfaction) to prove the transformation is working for everyone.
• Your agent is only as good as your content. Messy data results in confused agents. Watch out for content collisions, where multiple articles contradict each other. Quality content is key.

1-800Accountant is scaling client satisfaction with Agentforce

1-800Accountant faced an impossible scaling challenge during tax season. Their existing chatbot could only handle 10% of customer inquiries, creating wait times that were three times longer than their target and causing their CPAs and financial experts to spend a large portion of their time answering routine deadline questions instead of providing strategic tax advice. Each tax season required nine recruiters working full-time to find and train seasonal staff who could speak intelligently about financial services and troubleshoot complex tax questions.

The seasonal hiring challenge had become unsustainable. “We’ve learned over the years that it’s very difficult to scale in a service-oriented business without being able to scale people,” explained Ryan Teeples, CTO at 1-800Accountant. “It’s never been more difficult to scale people. We simply can’t hire fast enough to keep up with the demand and the growth that we have.”

This is where digital labor redefined their approach. Unlike basic chatbots that follow scripted responses, Agentforce can reason through complex tax scenarios, access multiple data sources, and provide the nuanced support the situations demand, while seamlessly escalating complex financial advice questions to human CPAs with full context and preliminary analysis.

The dual dividend is clear.

• Small business owners now get instant answers to complicated tax questions at 11 p.m. when panic strikes, with agents that understand their specific business type and can provide tailored suggestions – no more wait times.
• CPAs leveled-up their role. They moved from reactive problem-solvers spending most of their time on “When is the tax deadline?” questions to proactive business advisors, building strategic relationships and helping entrepreneurs navigate major financial decisions that truly differentiate the practice.

Agentforce now resolves 50% of customer support inquiries, with over 1,000 client engagements handled in just the first 24 hours. Read more about how 1-800Accountant is resolving its customer issues with Agentforce.

Building the competitive moat

Companies that master the dual dividend create a sustainable competitive advantage. Happy customers become loyal advocates. Empowered employees become innovation drivers. This combination becomes increasingly difficult for competitors to replicate.

“We’re not just deploying agents, we’re reimagining what customer success looks like when humans and AI work together,” said Slowey at Salesforce. “Our customers get better service, our employees do more meaningful work, and our business scales in ways we never thought possible.”

The companies that thrive in the age of agentic AI won’t simply automate their way to efficiency — they’ll amplify human potential while delighting customers. This approach is the blueprint for sustainable growth in an AI-first world.

Ready to realize your dual dividend?

Follow Slowey’s advice: “Begin with one use case where you can clearly measure both customer and employee impact. You don’t need to transform everything at once — you just need to start with something that matters to both your people and your customers.”

Ready to see these principles in action? Watch how other organizations are increasing customer service efficiency with Agentforce and discover practical implementation strategies in the video below.

Try this activity: Orchestrate how you team with agents

Use this exercise to foster rich collaboration between your human employees and AI agents.

A glimpse into the future and how to prepare today

In 1995, most people thought the internet was just email and basic websites. Few imagined we’d be video-calling grandparents from our phones, scheduling grocery delivery with an app, or running entire businesses in the cloud. The organizations that started building web presences early, even just simple websites, were positioned to capitalize on each wave of innovation. Today’s AI agents are like those early websites — the foundation for transformation many of us can barely imagine.

At Salesforce, we take a unique view of AI’s future. An approach that’s grounded in scientific research and focused squarely on enterprise needs rather than speculative sci-fi scenarios hinting at “artificial general intelligence” or AGI. Our global AI Research team, by contrast, is invested in what we call Enterprise AI — AI agents and systems that meet the highest standard of quality, reliability, security, and of course, ethical use of data, to fuel not only Salesforce products like Agentforce but also businesses around the world.

In this forward-looking chapter, curated by our AI Research group, you’ll discover three macro trends that will shape the future of agentic AI and transform how agentic enterprises operate in the coming years.

1. Agent to Agent (A2A) ecosystems
2. Robotics and AI
3. Enterprise general intelligence (EGI)

Macro trend #1: A2A ecosystems — The next evolution of digital labor

The first major trend reshaping the future of work? The emergence of orchestrated agent ecosystems — networks of specialized AI agents that can collaborate across organizational boundaries to negotiate, collaborate, and co-create with agents from other companies.

AI agents are progressing from solo performers to orchestrated ensembles, just as music evolves from simple melodies to complex symphonies.

This evolution of AI agents will unfold in three stages.

1. Specialized contributors handling discrete tasks (aligns with Level 1 of our Agentic Maturity Model described in Chapter 1)
2. Seamless collaborators working in concert within your organization
3. Business orchestrators spanning organizational boundaries

In this final stage, we’ll see the emergence of A2A interactions that create entirely new patterns of business relationships. Your agents will negotiate with your vendors’ agents, serve your customers’ agents, and collaborate with your employees’ personal AI assistants. They’ll handle complex interactions like price negotiations, supply chain optimization, and customer experience orchestration with efficiency and intelligence that would be impossible for humans alone.

Macro trend #2: The new age of robotics, the convergence of digital and physical AI

The second trend transforming agentic enterprise is the convergence of digital AI and physical robotics — a development that extends the impact of agentic AI beyond information processing and into the physical world.

This convergence marks a pivotal moment for enterprise leaders. “Humans today spend countless hours on routine physical tasks — from sorting warehouse inventory, to folding hospital linens, to organizing stock rooms,” Savarese explained in a recent article. “Just as AI agents now handle email triage and report generation, freeing knowledge workers for strategic thinking, robotics promises similar liberation in the physical realm.”

While AI agents discussed in this guide operate through APIs and software interfaces, robots interact through motors and sensors. But the intelligence layer of modern robots — the ability to plan, adapt, and learn — remains consistent. We’re beginning to see these two types of AI work together to gain myriad skills across industries — from warehouse inventory management, to life-saving surgeries, to even mundane household tasks like laundry. This collaboration between digital and physical AI is exciting, because it dramatically scales up the capacity of what a single human can do, augmenting their physical as well as cognitive abilities.

Salesforce and Boston Dynamics: Pioneering the physical-digital future

At Salesforce’s own AI Research lab within our Palo Alto office, the future of digital-physical AI convergence isn’t just a vision — it’s a working reality. In partnership with Boston Dynamics, Salesforce has created what may be the first enterprise demonstration of Agentforce directing physical robots to complete real business tasks.

The integration showcases a “dispatch council” that manages both human workers and robots as a unified digital labor‌ force. When facility inspections are needed, managers can request tasks through Slack, and Agentforce intelligently dispatches the right resource — whether human or robotic — with the appropriate skills for each specific job.

In a recent demonstration, an ad-hoc pressure gauge inspection was initiated through a simple Slack conversation with a facilities agent. Within moments, “Spot 007” — one of the Boston Dynamics robots integrated into Salesforce’s system — was autonomously navigating to the inspection site. The robot positioned itself precisely and captured a clear image of the analog gauge, and Agentforce’s vision agent automatically processed the reading to determine if pressure levels exceeded safety thresholds.

But the system goes beyond simple automation. The robots are equipped with conversational AI that allows for natural interactions with employees. When approached by a new team member asking for directions to the printer, Spot 007 seamlessly provided helpful guidance, demonstrating how physical AI can enhance workplace experiences through both task execution and human interaction.

Perhaps most remarkably, the robots can perform complex manipulations; for example, carefully opening cabinet doors to inspect medical supplies, coordinating multiple cameras for optimal documentation, then automatically updating work order line items in Salesforce Field Service with completion status and photographic evidence.

“What you’re seeing here is Agentforce taking physical actions to solve real problems,” explained Bert Legrand, senior director, frontier AI product management, who led the project. “Rather than traditional AI agents that process information, we can think of these robots as ‘physical agents’ that can manipulate the environment, all orchestrated through the same intelligent reasoning systems.”

The key insight from Salesforce’s recent implementation? The boundary between digital and physical AI is dissolving faster than most organizations realize. The same reasoning engines, data integration capabilities, and workflow orchestration that power digital agents can seamlessly extend into physical operations, creating entirely new possibilities for how work gets done.

Macro trend #3: EGI, the future of AI for business

The third trend shaping the future of AI in business is the emergence of enterprise general intelligence (EGI) — a practical, business-focused definition of AI custom-built to bring an unprecedented level of reliability and skill capabilities to augment the enterprise workforce.

Unlike the vague concept of artificial general intelligence (AGI) often portrayed in science fiction, EGI takes a different approach. It focuses on developing AI systems that excel across two critical dimensions: capability (the power to navigate complex business environments) and consistency (the delivery of reliable, predictable results with seamless integration into existing systems).

“EGI must not only perform tasks reliably but also demonstrate quick reactions, strategic pivoting, and adaptability when encountering new information or challenges,” he said. “EGI systems will succeed by combining reliable execution with intelligent, adaptive strategy.”
They accomplish this by using multiple specialized agents — collaborating seamlessly with humans at the helm — whose higher levels of reasoning help them perform complex and operational tasks in rapidly changing environments.

Preparing for the future of the agentic enterprise

These three trends — orchestrated agent environments, digital-physical convergence, and EGI — won’t emerge overnight. They’ll unfold through what Salesforce AI Research calls “boring breakthroughs”: incremental yet profound improvements in everyday operations that collectively reshape industries.

Yet the organizations that will thrive in this future aren’t those with the most advanced technology — they’re those with the clearest vision and most thoughtful strategy. By focusing on the well-designed handoff between humans and agents, aligning AI initiatives with business objectives, and thoughtfully preparing both systems and people, leaders can position their organizations at the forefront of the agentic AI revolution.

Forward-thinking organizations can begin preparing for this future today through several practical steps, many of which are outlined in this guide.

1. Develop your agent strategy: Define which business functions can benefit most from specialized agents and how they might evolve toward collaboration.
   
2. Build your foundation: Implement the integrated infrastructure, data architecture, and governance frameworks that will support increasingly sophisticated agent environments.
   
3. Focus on human-AI collaboration: Invest in crafting seamless handoffs between human expertise and AI capabilities. This represents the key differentiator between organizations that merely deploy AI and those that truly transform.
   
4. Evolve your oversight approach: Move beyond simple “human-in-the-loop” models toward nuanced “human-at-the-helm” frameworks that balance autonomy with appropriate oversight.
   
5. Cultivate future-ready skills: Develop the human capabilities needed to work effectively with increasingly sophisticated AI systems.

“The organizations that will thrive in this new era,” concluded Savarese, “are those that understand AI isn’t just about technology — it’s about reimagining how humans and machines can work together to achieve what neither could accomplish alone. The future belongs to those who shape AI to serve real human and organizational needs — and the time to build that foundation is now.”

Learn more about Salesforce’s AI research by following the team on X at @SFReseearch.

Try this activity: EGI Readiness Assessment

As you continue your journey toward becoming an agentic AI company, evaluate your organization’s readiness for these future developments using the following assessment areas.

By addressing these questions today, you’ll be well positioned to capitalize on the transformative potential of agentic AI tomorrow — creating a future where humans and AI work together to achieve what neither could accomplish alone.

Companies are building competitive advantages today

While there are many measurable outcomes to becoming an agentic AI company, the biggest may not be measurable: unprecedented ingenuity and innovation across your organization. When your most talented people are freed from routine tasks to think strategically, when customers experience seamless service, and when teams pursue breakthrough ideas once impossible due to availability constraints, you’re unlocking human potential at scale.

From our work with customers all over the world, we know that the companies defining the next decade aren’t waiting for perfect conditions — they’re starting today, building competitive advantages through intelligent human-AI collaboration. Your journey begins with small steps captured in this guide, but leads to a future where your organization’s capacity for innovation knows no bounds.

The question isn’t whether this transformation will happen — it’s whether you’ll lead it or follow it.

Try this activity: Identify your innovation white space

The most powerful thing about agentic AI is not that it automates existing processes — it’s the promise of the limitless human potential it unlocks.

This activity helps you discover the “innovation white space” in your organization: high-value opportunities that become possible when AI agents handle routine work, freeing up your most talented people to focus on what only humans can do.

Use this structured approach to identify where your organization could achieve exponential growth by strategically redeploying human capital toward innovation initiatives.

No time to read? Get each chapter of this guide delivered to your inbox. Sign up now!