There’s an incredible amount of excitement right now surrounding the latest breakthroughs in AI intelligence. To our frontier model partners and customers:
Welcome to the Enterprise.
It’s a bold new world, where models play a critical role as one part of a reliable enterprise-grade architecture that delivers real outcomes for customers. We’re glad many are seeing the power of an operating system for humans and agents to work together.
Recent industry announcements — including the move toward platforms designed to manage agents and provide shared business context — do more than just move the needle. They validate the very vision we’ve been building for decades. It’s a powerful confirmation: The world is finally realizing that raw model intelligence is just the beginning. The real challenge, and the real opportunity, is how those agents are built, governed, and run within the heart of a business.
Validating the Agentic Enterprise
We’ve said it from the start: An agent is only as good as the data it can access and the trust it can maintain. It’s rewarding to see the industry move toward the realization that agents need a “context layer” and deep integration across siloed systems like CRM and ticketing tools.
At Salesforce, we didn’t have to connect these silos — we’ve been the system of record, system of engagement, and system of work for our customers for twenty-five years. What others are now identifying as the “opportunity gap”— the space between model potential and real-world deployment?
That’s Agentforce.
We’ve known for a while that a system of intelligence alone is nothing without the critical layers that surround it. To operate an agentic enterprise you need a:
- System of Engagement: The interface to connect with your customers and employees across the channels of your choice.
- System of Agency: In the enterprise, intelligence isn’t enough. You need behavior that is predictable, compliant, and safe. Support for an open ecosystem of applications and models, and the power to not just answer a question, but to trigger a workflow, update a record, and collaborate across silos natively. And, of course, you need governance and observability built in.
- System of Work: A trusted platform with apps trusted by, and tailored for, your department and your industry.
- System of Context: Without Data 360 (formerly Data Cloud), your AI is operating in a vacuum. To be effective, an agent must have a 360-degree view of the customer, grounded in the system of record.
Powering the Frontier: AI-Native Pioneers
We’re proud to partner with the very trailblazers who are defining the frontier. These AI-native companies don’t just build intelligence; they use the Agentforce 360 platform as their operational backbone to scale that intelligence.
- Anthropic The pioneers of Claude rely on Slack as their primary interface for collaboration and productivity. By using Slack as their “operating system,” Anthropic has saved $4.5 million and cut three days off complex deal cycles.
- Google partners with Salesforce and is deploying Agentforce across their portfolio including Google Fiber, YouTube, and Waymo, rather than building competing solutions from scratch.
- NVIDIA partners with Salesforce to connect Cobalt security robots at Salesforce Tower to Agentforce. By using NVIDIA’s visual AI (Cosmos Reason), the robots act as “eyes” that can detect physical issues—like a door left open—and automatically trigger a maintenance or security ticket within Salesforce.
- OpenAI partners with Salesforce and uses the Agentforce 360 Platform and Slack. OpenAI’s frontier models, including GPT-5, empower enterprises to build sophisticated agents within the Salesforce platform.
Outcomes, Not Outputs
I have a deep belief in our agentic approach, but that belief would mean nothing without customer trust and results. Customers from around the world, in different industries, for different use cases, are driving massive value with Agentforce today. This isn’t just a tech trend; it’s a fundamental shift in how work gets done:
- Absa Bank Group – Absa’s Fraud & Debit Order Agent was deployed in just 5 weeks and delivers 88% faster turnaround for debit order reversals with a 50% resolution rate for fraud cases.
- AGIBANK (Brazil) – AGIBANK’s Call Deflection Agent handles 9,000 conversations per week with an 80% call deflection rate, significantly reducing costs and the need for human agents.
- Big Brothers Big Sisters uses Agentforce to reduce the administrative burden on staff, allowing them to focus on creating more meaningful matches, twice as fast.
- CaixaBank serves more than 20 million customers and uses Agentforce to provide real-time insights and answers during calls, helping CaixaBank lead the way in AI-powered banking within a highly regulated industry.
- Engine has reduced sales rep research time by 40%, delivering savings in the millions.
- Falabella provides 24/7 assistance to global shoppers, resolving 60% of requests autonomously via WhatsApp.
- FedEx uses the Agentforce 360 Platform to orchestrate 15 million daily package deliveries across 220+ countries.
- Heathrow is using Agentforce to resolve 90% of chats without the need to transfer to a live agent.
- iFood (Brazil) – iFood’s WhatsApp Sales Agent “Diana” achieved a 2.7% conversion rate from cold leads (5x-7x improvement over human reps) and is projected to reach 4+ million leads, earning it the title of “one-million-dollar Agentforce agent.”
- Indeed is using Agentforce to support 24/7 assistance for the 3.5 million employers who use their platform.
Lennar‘s after-hours Sales Agent “LISA” handles overnight inquiries, helping up to 30% of Lennar’s inbound leads make appointments, even when a live agent isn’t available. - LIV Golf has deployed two AI agents using Agentforce. “Chip,” the Fan Caddie, engages and entertains LIV Golf fans while they stream events on their devices. Agent Caddie tees up real-time shot intelligence and rich statistics for LIV Golf’s award-winning broadcasters.
- OpenTable is using Agentforce to scale personalized dining experiences, deflecting an estimated 70% of Tier 1 restaurant support cases and saving $2M annually.
- Prudential is freeing up their wholesalers from manual tasks like note entry with Data 360 unifying Financial Services Cloud, Tableau, and external data lakes to power their Agentforce pilot, giving wholesalers hours back each week.
- RBC Wealth Management advisors save 60 minutes per meeting while RBC cut data management costs by 50%, boosting efficiency and capacity to serve more clients while building trusted relationships.
- Smartsheet has transformed their customer and employee experience with Agentforce. Median time to resolve customer issues has been reduced by 9 days, deflections have increased from 28% to 47%, and employees at Smartsheet save 54 minutes per day.
- Southwest Airlines is using Agentforce to deflect an estimated 20% of its 20 million annual customer service requests, reducing costly phone volumes and removing friction for travelers.
- The Adecco Group is using Agentforce to help 300 million annual applicants find their dream jobs faster, with total precision.
- Vivint’s Agentforce is delivering $8M savings while scaling to 50% of authenticated users.
- Williams-Sonoma, Inc. went from idea to agent in 28 days and is now automating over 60% of chat inquiries.
These aren’t experiments or pilots of a new platform; hundreds of thousands of customers around the world run their businesses on Salesforce. And we’re helping every one of them transform into an agentic enterprise.
The Road Ahead
To our partners building the next generation of intelligence: We welcome your innovation. We’re glad to see the industry pivot toward the governance and business context that we’ve made central to the Salesforce platform for years.
But as we look toward the future, we must keep customer trust at the forefront. A model can be brilliant, but a platform must be trusted. It must be grounded in the operational backbone that runs the world’s most successful businesses. Because everyone wants to talk about vibe coding, but no one wants to talk about vibe operating.
And your data? It should never belong to anyone but you.
We’ve been the system of record for a long time. And we’ve been building agentic enterprises since the GPT models became public. We’re glad the rest of the industry is finally arriving at the same conclusion.
A frontier is a wonderful place to explore, but the enterprise is where the work gets done.
Let’s get to work.
