The average number of artificial intelligence (AI) agents activated by businesses nearly tripled over the past year as companies increasingly deploy AI to perform tasks and execute business processes, US-based cloud software company Salesforce said.
The average number of agents activated per organization increased nearly threefold over the past year, according to Salesforce’s 2026 Agentic Enterprise Index, which analyzed aggregated usage data from businesses using its Agentforce platform from February 2025 to April 2026.
Salesforce identified two broad approaches to AI agent deployment. Consumer-facing industries generally use agents for high-volume, task-specific work aimed at addressing immediate customer needs, while operationally complex and heavily regulated industries deploy agents capable of handling more varied tasks requiring cross-functional business logic.
The company said AI agents are also moving beyond generating text toward executing work. Salesforce developed the Agentic Work Unit (AWU) to measure one discrete task completed by an AI agent. As of April 2026, AWU output from Agentforce agents was increasing at a 15% compound monthly growth rate.
Consumer-facing industries recorded some of the highest AWU output, driven by large volumes of customer interactions. Retail agents, however, generally remained focused on routine tasks, averaging one to two actions per agent during most of the year.
According to Salesforce, AI agents are becoming capable of handling more complex tasks. The average agent could act on six skills, up from two at the beginning of 2025. During peak shopping periods, the average retail agent could act on nine skills, a 350% increase, as businesses deployed agents to address more complex, multi-step customer needs.
Agents are also increasingly taking actions across different cloud systems. A service agent, for example, can move beyond answering customer questions to retrieving sales records and providing personalized recommendations. Salesforce said this expansion highlights the need for architectures that allow agents to execute tasks and trigger workflows across systems.
Meanwhile, industries with more complex operations and regulatory requirements are prioritizing agent sophistication.
Manufacturing, financial services and healthcare and life sciences are deploying more advanced agent networks than traditional AI front-runners such as technology and retail, Salesforce said. These industries use agents across a wider range of activities, including retrieving and summarizing data, drafting communications and directly updating records.
Salesforce said the approach is particularly evident in manufacturing.
Siemens, which sells hardware and software products across seven business units and has 18,000 sellers, receives about 2,800 unqualified inbound leads each week. Its sales teams previously spent time pursuing leads without sufficient information on budget, authority or timelines.
Using Agentforce, Siemens deployed a coordinated multi-agent workflow in which one agent engages and nurtures leads while another gathers missing information, applies qualification rules and routes leads with cross-division context. The system allows the process to operate around the clock, Salesforce said.
Financial services have similarly combined complex agent capabilities with high-volume deployment. The sector represented about 10% of total monthly AWU output, with activity partly driven by seasonal increases in consumer demand such as tax season.
Salesforce said AI agents are increasingly producing efficiency gains and supporting sales growth. Across industries, agents are taking more actions, including triggering background workflows and applying business logic, compared with simply generating text. The action-to-output ratio has grown at a 15% compound monthly growth rate.
Businesses using AI agents also recorded stronger sales growth than those that did not, with the trend particularly pronounced in retail and consumer industries 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,” Joe Inzerillo, Salesforce president of Enterprise AI and Technology, said.
“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,” he added.
Employee engagement with AI agents has also increased. The average employee engaged with an agent 300% more frequently per week over the analysis period.
Salesforce said its own Slackbot, which it describes as the fastest-adopted tool in the company’s history, is regularly used by 83% of employees and saves the average employee up to five hours of work per week. Employees use the tool to summarize missed conversations, draft content and assemble briefs using information from different channels and applications.
Customer use of AI agents has also expanded. Over the past five quarters, agents handled 170 times more customer service chats than in previous years, while consistently resolving seven out of 10 conversations without human assistance, Salesforce said.
The company also cited research showing that 77% of shoppers who interacted with branded AI shopping agents on websites felt more confident about their purchase than those who did not. In customer service organizations, Salesforce said agents are having the biggest impact on customer satisfaction compared with service representative productivity, average handle time, customer retention and first-response time.
Salesforce expects AI agents to increasingly combine rapid, high-volume deployment with the ability to handle complex workflows. Rather than choosing between agents designed for scale and those built for sophisticated tasks, businesses are expected to use both approaches to respond to demand while maintaining the cross-functional logic required for complex processes.
The index covered businesses that activated agents in production every month from February 2025 through April 2026. Salesforce said the findings are based on aggregated usage data from the cohort and are not indicative of its own performance. — Kaizzer Angela Marie V. Manuba
