Salesforce's AI annual recurring revenue surged more than 200% to $3.9 billion, with Agentforce ARR alone hitting $1.5 billion, the company reported during its Q2 fiscal 2027 investor webinar. For customer support leaders, the numbers reflect a market that is moving fast from AI experimentation to production deployments tied directly to business outcomes.
More than 10,000 customers now use at least one Salesforce AI product, and the number with AI in production has roughly doubled since February. Conor Marsden, Salesforce's President of Sales and Chief Consumption Officer, said customers are increasingly adding second, third and fourth AI solutions after initial deployments. Agentic work units - a measure of work conducted on the platform - rose 97% to $7 billion.
Customer support deployments show concrete results
Several deployments highlight what adoption looks like in practice. SharkNinja launched service and shopper agents and reported a 6% increase in conversion. Its custom "unboxing agent" for espresso machines achieved a 93% resolution rate, with only 7% of interactions escalated to a human. Wyndham deployed a contact-center agent that produced a 25% decrease in average handle time. Live Nation handled 37,000 guest interactions at its BottleRock festival with an agent deployed within 30 days.
Marsden said out-of-the-box agents can be deployed in 30 to 45 days, while AI coding tools have improved the speed of Salesforce environment configurations by 40%. One global retailer that received an internal estimate of 35 weeks for a custom contact deployment went live with Salesforce in six weeks. Salesforce has roughly 600 "builders" - employees embedded with sales teams to help customers deploy AI - and plans to more than double that investment by year-end.
Outcome-based pricing changes the cost equation
Bill Patterson, Salesforce's President and Chief Commercial Officer, said the company is simplifying its AI pricing with an emphasis on outcomes. For help agents like Casey, pricing is based on resolutions delivered. "If they do not resolve the issue, you do not pay for the offering," Patterson said. The company plans to expand outcome-based pricing across sales, service and other areas, tying costs to leads processed, orders managed or cases resolved rather than token consumption.
Flex Credits, pay-as-you-go options and broader enterprise agreements are designed to address unpredictable AI costs. Marsden said the structure lets organizations using multiple Salesforce products achieve more predictable spending. The company is also packaging work capacity into "headless" add-ons that provide access to AI-powered workflows outside the traditional full Salesforce application interface.
Slack and model choice expand the interface layer
Salesforce executives positioned Slack as a key interface for AI adoption. Slackbot will be monetized through Slack's per-user subscription model, with included capacity and additional usage available through Flex Credits. Patterson said integrating external AI models and productivity systems into Slack should increase user engagement, retention and platform usage. Valmik Desai, senior director of investor relations, said Slack posted its strongest net-new annual contract value performance since the acquisition, and upgrades have tripled since Slackbot became generally available.
On model selection, Patterson said customers can bring their own models to Agentforce and choose models for particular prompts or workflows. Salesforce plans to select what it considers the best initial models for use cases while allowing customers to override those choices. Marsden said the company works with frontier-model providers and open-source models, with the goal of abstracting model complexity so customers get the appropriate model for a desired outcome at the appropriate cost.
Why this matters for customer support teams
The shift to outcome-based pricing changes how support leaders should evaluate AI investments. When vendors charge only for resolved cases, the risk of paying for ineffective automation drops sharply. The 93% resolution rate from SharkNinja's agent and Wyndham's 25% reduction in handle time are not vendor promises - they are reported results from production deployments. For teams building a business case, the deployment timelines matter too: 30 to 45 days for out-of-the-box agents and six weeks for a custom contact center deployment reset expectations about what speed looks like. Professionals looking to build these skills can explore AI for Customer Support Courses or follow an AI Learning Path for Call Center Supervisors to understand how agentic workflows fit into existing operations.
Your membership also unlocks:
