Salesforce’s 2026 Agentic Enterprise Index reveals a major shift in how businesses are deploying artificial intelligence, with AI agents moving beyond simple conversations to take on increasingly meaningful tasks across industries.

Based on aggregated AI usage data from the Agentforce platform, the report highlights a growing divide in how companies approach agentic AI. While consumer-facing industries tend to prioritize high-volume, task-specific deployments that address immediate customer needs, operationally complex and highly regulated sectors are increasingly turning to versatile agents capable of handling multi-step processes and cross-functional business logic.

Despite these different approaches, both models are delivering value. Salesforce’s findings suggest that the future of agentic AI will depend on the coexistence of fast, large-scale automation and increasingly sophisticated systems capable of handling complex work.

Businesses Are Deploying More AI Agents, Faster

The average number of agents activated per organization has nearly tripled over the past year. Once provisioned, businesses begin creating agents in an average of just two days, with that timeline declining by 53% across the analysis period.

But the growing number of agents is only part of the story. Salesforce has introduced the term Agentic Work Unit (AWU) to measure the amount of actual work completed by an AI agent. An AWU represents a discrete task performed by an agent—the point where AI capabilities translate into tangible business activity.

As of April 2026, Agentforce agents’ AWU output was increasing at a compound monthly growth rate of 15%.

Consumer-facing industries are leading in overall AWU volume, particularly businesses that manage large numbers of customer interactions. However, retail agents generally remain focused on one or two actions throughout most of the year, reflecting their role in handling high volumes of routine customer needs.

Pandora, for example, uses Gemma, an AI concierge powered by Agentforce, to manage spikes in customer inquiries during periods such as the holidays and Valentine’s Day. Gemma handles routine questions about order status, shipping, and jewelry care while also providing personalized gift recommendations.

By connecting with Pandora’s order systems and product catalogs, Gemma can handle 60% of routine support requests during peak traffic while contributing to a 10% increase in Net Promoter Score. This allows human representatives to concentrate on more complex and high-touch customer interactions.

AI Agents Are Becoming More Versatile

While many routine AI tasks require agents to use only one or two basic skills, businesses are increasingly relying on them to handle more sophisticated workflows.

The average agent can now act on six skills, up from two at the beginning of 2025. During peak shopping periods, the capabilities of retail agents expand even further. The average retail agent was able to act on nine skills during peak season—a 350% increase—suggesting that businesses are turning to AI to address more complicated, multi-step customer needs when demand rises.

Agents are also increasingly performing actions across different cloud environments. Instead of simply answering a customer question, for example, a service agent can retrieve sales information, provide a personalized recommendation, and initiate other workflows.

This broader capability highlights the importance of headless architecture, which separates an agent’s logic from traditional front-end interfaces and allows agents to execute tasks and trigger workflows across different environments.

Complexity Is Driving AI Adoption in Regulated Industries

While technology and retail have traditionally been viewed as early adopters of AI, Salesforce’s findings show that industries such as manufacturing, financial services, and healthcare and life sciences are building some of the most sophisticated agent networks.

These sectors are deploying AI across a broader spectrum of tasks, from retrieving and summarizing information to drafting communications, analyzing data, and updating records directly.

Salesforce measures this complexity through its Sophistication Index, which maps agent actions across five progressive levels of cognitive complexity. Lower levels involve tasks such as reading records, coordinating activities, and synthesizing information, while higher levels include writing to databases, analyzing information, and parsing complex inputs.

The findings indicate that operationally complex and regulated industries are increasingly prioritizing versatility and sophisticated decision-making over sheer volume.

Siemens, for instance, uses Agentforce to coordinate lead qualification across seven business units and 18,000 sellers. Faced with approximately 2,800 unqualified inbound leads each week, the company deployed a multi-agent workflow in which different agents engage and nurture leads, gather missing information, apply qualification rules, and route opportunities with cross-division context.

The approach allows Siemens to manage a complex sales process end-to-end and around the clock.

Financial Services Shows That Scale and Complexity Can Coexist

Financial services demonstrates how high-volume deployment and sophisticated agent capabilities can work together.

The industry accounts for a similar 10% share of total monthly agent AWU output, with activity driven in part by seasonal spikes in consumer demand, such as tax season.

At PenFed, a federally chartered credit union serving military members and their families, AI deployment must operate within strict compliance, security, and verification requirements.

The organization has deployed Ace, an AI agent secured behind online banking logins. Ace can evaluate account balances, check loan application statuses, transfer funds, and provide answers grounded in a curated knowledge base.

Another agent, Echo, extends these capabilities to voice interactions and is designed to replace traditional interactive voice response systems and automated teller services.

The example illustrates how AI agents can perform meaningful, multi-step financial tasks while operating within environments where governance and security are critical.

AI Agents Are Delivering Measurable Business Value

Across industries, AI agents are increasingly shifting from generating text to executing actions. Rather than simply drafting a response about a customer’s order, for example, an agent can retrieve the relevant record, apply a business rule, issue a refund, or rebook an appointment.

Salesforce reports that the ratio of actions to outputs is growing at a 15% compound monthly growth rate.

The benefits extend beyond efficiency. Businesses that use AI agents are also seeing stronger sales growth than those that do not, with the trend particularly pronounced among retail and consumer-facing companies during holiday shopping periods.

“Whether you’re spinning up agents to operate at massive scale or orchestrating them through deep, multistep pipelines, the bottom line is they’re shipping real value,” said Joe Inzerillo, Salesforce President of Enterprise AI and Technology. “That ROI isn’t just showing up on the top line in sales numbers but in execution efficiency. We are moving from passive chatbots and predictive models to execution-driven agents that actually roll up their sleeves and drive real value.”

Growing Trust Among Employees and Customers

The increasing use of AI agents also points to growing trust among employees.

According to Salesforce, the average employee engaged with an agent 300% more often per week over the course of the year. Slack agents averaged 67 sessions per week, with usage tripling in April compared with February.

Salesforce’s own Slackbot provides one example. Used regularly by 83% of Salesforce employees, the AI agent saves the average employee up to five hours of work each week. Employees use Slackbot for tasks ranging from summarizing missed conversations and drafting content to assembling briefs using information gathered across channels and applications.

Customer adoption is also increasing. Over the past five quarters, AI agents handled 170 times more customer service chats than in previous years, consistently resolving seven out of 10 interactions without human assistance.

The report also points to growing consumer confidence in AI-assisted shopping. A recent study cited by Salesforce found that 77% of shoppers who interacted with branded shopper agents felt more confident about their purchase than those who did not.

For customer service organizations, Salesforce says agents are making their greatest impact on customer satisfaction, surpassing other factors such as service representative productivity, average handling time, customer retention, and first-response time.

The Next Phase of Enterprise AI

The findings of Salesforce’s 2026 Agentic Enterprise Index point to an increasingly action-oriented era of enterprise AI.

Some businesses are using agents to manage enormous volumes of straightforward tasks, while others are building sophisticated networks capable of coordinating complex, multi-step processes. Both approaches are expanding as organizations gain confidence in what AI agents can accomplish.

Rather than replacing one model with another, the emerging enterprise landscape suggests that scale and sophistication will continue to develop side by side. As AI agents become more capable, businesses are moving closer to a digital workforce that can not only understand information and respond to requests, but also execute work and deliver measurable results.