AI agents are supposed to reshape enterprise work, but early adopters at Dreamforce are offering a reality check: automation can accelerate service, yet customers still need people at the moments that matter most. AT&T, Southwest Airlines, CVS, and Crocs are all putting Salesforce’s Agentforce to work—but each is drawing deliberate boundaries around human judgment.
Agentforce adoption is growing, but selective
Salesforce launched Agentforce in 2024 with a promise of digital workers that can handle business tasks without constant human direction. At this week’s event, the company said roughly 30,000 customers use the platform. Against a base of about 150,000 customers, that’s around 20% adoption—meaningful, but still far from a fully autonomous enterprise.
“AI can make us much more efficient,” said John Miller, VP of consumer and business solutions at AT&T. “But we really want to make sure that there’s still that human interaction.”
Where brands are holding the line
AT&T has reduced the time to upgrade a customer’s phone from about 30 minutes to roughly 10 minutes using AI. But the company intentionally keeps cancellations with human employees. The goal is not just to save a customer; it’s to understand why they are leaving and to respond with options an agent may not be equipped to offer.
Southwest Airlines takes a parallel approach. It uses chatbots for servicing tasks such as flight changes and transactions. Megan Rauber, a technology manager at Southwest, said humans remain on call for “right here, right now” problems—the urgent, emotionally charged moments where hospitality and reassurance matter.
- AT&T automates routine upgrades but keeps cancellation conversations human.
- Southwest automates transactions but reserves empathy-heavy moments for staff.
- Crocs runs consumer-facing bot Rivet plus internal promo tools, but maintains human oversight.
Crocs: the hidden complexity is data
Crocs is deploying AI agents on both sides of the customer journey. Its consumer-facing chat and voice bot, Rivet, assists shoppers, while internal agents help developers create promotions and categorize products. Still, Feliz Papich, SVP of digital technology and experience, said the organization keeps a human in the loop as it learns the capabilities and limits of automation.
Papich cautioned that the rush to build slick agent interfaces can obscure the harder problem: the data underneath. If the data foundation isn’t reliable, even the best-designed agent fails against customer expectations. For marketing and CX teams, that’s a useful reminder that an AI agent rollout is as much a data engineering project as a technology deployment.
Why this matters for marketers
Consumer trust in AI drops sharply as autonomy increases. DEPT research cited in the report found that 65% of shoppers who have used AI are comfortable with it through the recommendation stage, but only 15% are comfortable with preparation or transaction steps, and just 7.5% with fully autonomous actions. Another finding: 56% of recent AI shoppers chose a different brand than they originally planned.
The pattern suggests a practical framework for marketing and customer-experience leaders: automate high-frequency, low-stakes tasks; keep high-empathy, high-commercial-risk moments human; and treat data quality as a gate before launching agent workflows. Efficiency wins are real, but they don’t eliminate the need for relationship-building.
Source: Adweek
