Andriy Onufriyenko/ Moment via Getty Images
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- 70% of service organizations using AI agents report positive outcomes within 60 days of deployment.
- Agentic AI adoption for service organizations has grown from 39% to 66% in the past 12 months.
- A new outcome-based pricing model will accelerate the adoption of agentic AI in the enterprise.
Adoption of AI agents in customer service has grown from 39% in 2025 to 66% in 2026, according to a Salesforce survey of 3,075 service professionals representing 13 countries across five continents. To maintain customer trust, service organizations continue to ensure that people are in the customer service loop. In fact, 77% of companies with AI agents allow customers to connect with human agents at any point.
AI agent adoption in customer service
The Salesforce survey found that 85% of service organizations use AI. The current use is generative AI at 78%, predictive AI at 71%, and agentic AI at 66%. The use of agentic AI by the end of 2026 is expected to be at 88%.
The customer-facing adoption of AI agents is at 89%, meaning the agents are used across the entire service lifecycle and across all channels, including web, voice, apps, text, and social networks. The top use cases for AI agents include: proactive outreach, personalized product recommendations, resolving cases, case routing, and after-call work.
Adoption drives new skills development for human agents
The survey revealed that service organizations are building more skills to support digital labor, aka AI agents. The roles expected to expand due to AI adoption include data management (66%), specialist (62%), AI architect (61%), prompt specialist (50%) and AI generalist (48%). Expanding the capabilities of AI agents will require autonomous design engineers and relationship design engineers to ensure the proper hand-offs between humans and AIs. Most companies are investing in AI training for their staff. The survey found that only 3% of service reps report no engagement with upskilling programs. The AI training curriculum includes workshops and conferences (53%), internal training programs (53%), and online courses (49%).
The skills priorities for services professionals as they adopt AI agents in the workplace include AI oversight and judgment, complex problem solving, and adaptability and learning agility, which includes strategic thinking.
AI agents highlight need for productivity improvements
Nearly 9 out of 10 respondents said they are using AI for internal employee-facing functions, including how teams are optimally managed. Half of service leaders are using AI agents to analyze trends and adjust their workflows. Service leaders are using AI to track employee performance (50%), predict demand (47%) and recommend staffing schedule adjustments (40%). The results of using AI for back office performance management is very promising. The vast majority of service leaders (92%) note that AI improves their ability to coach at scale.
Impact now measured in precision and desired outcomes
The survey found that 83% of service organizations with AI agents have deployments across five or more channels. AI agents are no longer a single touchpoint for companies. The top channels include email, online chat, messaging apps, SMS, and phone. AI agent deployments include 74% for online chat, 72% for email, messenger apps, phone, and text/SMS, and 69% for customer portals and collaboration tools. The key challenge for AI agents is the hand-off to humans across any channel of choice for customers. The AI agents must have contextual understanding of each engagement in order to properly inform their human colleagues.
The big surprise for some regarding this global survey of AI agent adoption is that return on investment is coming faster than what businesses initially forecasted. The survey found that 40% of the time AI is used in case resolution, the work is done completely autonomously. This outcome can potentially drive an average of 20% decrease in case resolution time. Service organizations are measuring AI adoption based on tangible measure of business outcomes, including metrics like case resolution time.
The survey found that 70% of service organizations with AI agents observe measurable value within 60 days of deployment, while 25% of service organizations see value from AI agents within 30 days. The focus is now more on business outcomes, as it should be. The lessons learned from nearly two years of business adoption of AI agents is clearly showing that technology must serve the business needs in order for adoption to accelerate. Time to resolution, leaner workflows, ability to anticipate outcomes and ultimately customer and employee satisfaction is key to business success and wider deployment of AI agents.
The survey found that the performance metrics that most improved with use of AI Agents include customer satisfaction, service rep productivity, average hand time and customer retention. First-response time was also improved.
The use of agentic AI for customer service at Salesforce is approaching more than 4.5 million conversations, double the number of cases managed by humans during the same time period, with a staggering 70% resolution success. After 1 million customer conversations with AI agents, Salesforce shared some valuable and surprising lessons, including the need for agents to have a dynamic brain and a caring heart.
Now that Salesforce has quadrupled its workload with AI agents, there are more lessons to be shared, including the focus on ease of deployment and a better alignment of objectives aimed at outcome maximization instead of token usage maximization. Salesforce’s emphasis on benchmarking AI agents based on business outcomes came to life by its introduction of a help agent, a pre-packaged service agent designed to deliver faster value to businesses.
The help agent is connected to a company’s knowledge base, workflows, and sanctioned actions, and service channels in only a few minutes. But the innovation goes beyond just ease of production deployment but rather how customers can truly measure return on their agentic AI investments. The pricing model for the help agent is pay-per-resolution pricing. This outcome-based resolution pricing model means companies only pay when the AI agent resolves an issue autonomously — without human intervention.
The confidence needed to package innovation and pricing based on real outcomes only comes from millions of customer interactions and tens of thousands of customers using the help agent.
To learn more about adoption of AI agents global survey, you can visit here.
