It’s not just pilots anymore: AI agents are in production, servicing real customer interactions at scale. The quickest-moving players are squarely in the consumer-facing industries, where surges of demand and service expectations, and slim margins reward automation which can talk, act, and hand off cleanly to humans.
New data in Salesforce’s Agentic Enterprise Index and learnings from its customers suggest that the tables have turned: first-mover companies more than doubled their use of AI agents in terms of creations they made (up 119%) during the first six months of this year, with service organizations leading the way as customer-service conversations handled by agents ballooned. For the average business, agent-led service chats had grown over 2,000% in six months, and there was an average 22-fold increase in agent-led conversations.
Where AI agents are scaling most quickly
Among those, Salesforce’s index identified three industries that are racing ahead: travel and hospitality, retail, and financial services. Travel and hospitality led the surge, with agent actions increasing at a monthly rate of 133% on average in the first half of the year; retail was second, at 128%; and financial services increased by 105% per month.
The common element: high-volume, repeatable interactions that are well suited to autonomous workflows. Booking alterations, returns and exchanges, claims, and fraud checks — these are the kind of routine tasks where AI agents excel repeatedly and leave human experts free to handle exceptions.
Why consumer industries acted first on AI agents
Consumer brands rise or fall on response times and personalization. They also contend with seasonal spikes and a long tail of routine questions that doesn’t merit adding permanent headcount. AI agents take on the surge, triage quickly, and pick up context before routing complex cases to humans — reducing cost-to-serve without degrading experience.
Another accelerant: data. They are already flush with a wealth of operational and customer history, otherwise known as orders, itineraries, loyalty profiles, or account events — rich soil for retrieval-augmented agents that can converse as well as perform. In recent years, McKinsey has repeatedly listed customer operations among the highest-value pools for AI — these sectors are currently demonstrating that.
How AI agents are currently being used in business
Customer service is still the breakout use case. Salesforce research:
- FAQs
- Order status
- Refunds (regular refund checks)
- Conversation summaries
- Knowledge lookups for reps
- Personalized recommendations
In sales, agents write emails, add to-dos, or schedule meetings. Internally, agents answer calls, gather records, set up follow-ups, and launch workflows between systems.
Adoption is pulling customers toward, not pushing them away: when offered the choice, 94% of customers opt into agent interactions. Crucially, that doesn’t replace people. That increase in routing actual humans was even more pronounced — passing from 22% of calls to 32% over a period of several months, indicating that agents are getting better at determining when they need someone with expertise on the line and sending it there.
Workforces are adapting quickly too. Global worker use of AI is up 233% since late last year, according to the Salesforce Workforce Index. Service leaders anticipate that 50 percent of cases will be resolved using AI by 2027, which is consistent with month-over-month growth in employee–agent conversations and downstream agent actions that Salesforce has seen.
From the front lines, real-world examples
In retail and fintech, Klarna announced its customer service AI assistant now takes care of most chats, doing the work of hundreds of full-time agents — a figurative benchmark for what at-scale containment can mean. Big retailers are integrating agentic assistants in apps to handle returns, check inventory, and offer something like post-purchase support — with Walmart and IKEA among those experimenting with AI-driven shopping experiences or service.
Travel platforms are using planning and servicing agents in combination. Expedia and Booking.com unveiled AI trip planners that create itineraries and in some cases start rebooking flows when plans do change. The agents are being dispatched by airlines and hotel groups to answer queries on disruptions, issue vouchers, and handle loyalty-program questions within minutes.
In banking, table stakes are consumer-grade assistants. Bank of America’s Erica has surpassed the billion-interaction milestone, and more recent agentic tools — utility belts for humans inside call centers — help reps with summarization, knowledge retrieval, and compliance prompts. On fraud and disputes, agents gather data, verify an identity, and write up a case file before an expert reviews the edge cases.
What leaders measure next as AI agents scale up
The KPI collection is also progressing beyond just containment. Executive teams are monitoring:
- Agent-led resolution rates
- Cost-to-serve
- First-contact resolution
- Speed to escalation
- CSAT after handoff
- Error budgets for autonomous actions
For regulated industries, model provenance, audit trails, and policy compliance are table stakes.
It’s difficult to ignore the financial case. Salesforce predicts agentic AI could generate nearly $450 billion in value by 2028, in terms of revenue and efficiency gains, based on the company’s research — and says almost eight out of ten CEOs believe AI will significantly change his or her industry within three years. And with retail, travel, hospitality, and financial services already amplifying monthly growth in agent actions, the competitive gulf will expand rapidly — to the advantage of companies that design for human-AI teamwork right from day one.