As someone who's spent years watching CX professionals wrestle with the impossible balance between personalization and scale, I found myself captivated by what Steve Hammond shared at the Salesforce Agentforce World Tour 2025. Hammond, EVP and GM of Agentforce Marketing, didn't just describe a product roadmap. He painted a picture of a fundamental shift happening right now in how brands connect with customers, and more importantly, he offered a pragmatic path forward for leaders who feel overwhelmed by the pace of change.
The energy on the show floor was palpable. Marketing leaders weren't asking if they should adopt AI agents. They were asking how to start, where to begin, and what they could accomplish. That urgency tells you everything you need to know about this moment. But to understand why this shift feels so significant, we need to look at the problem marketers have been trying to solve for the past two decades.
The 20-Year Personalization Problem Finally Has an Answer
For two decades, marketers have chased the same dream: true personalization that makes customers feel genuinely understood. The challenge has always been execution. How do you capture meaningful data and interpret it accurately? How do you create the right content and deliver it at the right moment across multiple channels, all at scale?
Hammond explained that previous approaches forced marketers into educated guessing. "We've had different technologies that can do some kind of personalization, but a lot of that is based off of trying to derive based on your click patterns. It's inferred, it's derived, it's kind of forced," he said.
What makes AI agents fundamentally different is their ability to understand intent directly rather than inferring it from behavior patterns. Large language models can now interpret meaning, tone, and intent directly from customer interactions. Rather than creating hundreds or thousands of content variations and hoping to match them correctly to audience segments, marketers can now break content into atomic elements—libraries of images, copy, and calls to action—and let AI assemble infinite variations dynamically based on individual customer profiles and behaviors.
The Pandora case study Hammond referenced during the keynote demonstrated this perfectly. Instead of manually crafting experiences for each channel, brands can deploy agents that design experiences on the fly across email, web, mobile applications, SMS, and WhatsApp. This dynamic assembly approach solves both the content creation bottleneck and the matching problem that have plagued personalization efforts for years.
Yet even the most sophisticated AI can only be as good as the data it has access to, which brings us to the foundational infrastructure making this transformation possible.
Data 360: The Foundation That Makes Agentic Marketing Possible
When I asked Hammond about the infrastructure enabling this transformation, he pointed to Data 360 as the critical foundation. For AI to make intelligent decisions, it needs a unified view of customer data, even when that data lives across different systems and data lakes.
"What you're doing with Data 360 is harmonizing a common view of your customer along with other relevant data in order to facilitate optimized decision making for personalization, for AI, for customer experience."
This zero-copy approach means data can remain in its original locations while AI gains visibility across all sources. That centralized decision-making capability is what allows agents to learn from every channel and continuously improve their understanding of what content works best for individual customers.
Without this unified view, Hammond emphasized, AI struggles to determine whether data is relevant or whether interactions are successful. The fragmentation that has characterized most enterprise data architectures becomes a critical barrier to AI effectiveness. But even with the right infrastructure in place, many organizations still hesitate to deploy AI in customer-facing roles. Understanding why reveals as much about organizational psychology as it does about technology.
The Fear Factor and How Salesforce Addresses It
Despite the excitement Hammond witnessed on the show floor, he acknowledged that fear of the unknown remains the biggest barrier to adoption. Legal concerns, regulatory worries, and the risk of embarrassment or fines keep many organizations from putting AI on their websites.
These concerns aren't irrational. A single AI misstep that goes viral can damage brand reputation built over decades. The stakes feel impossibly high, particularly for regulated industries or consumer-facing brands with millions of customers.
Salesforce has built comprehensive safety controls to address these concerns. The platform protects customer data by ensuring large language models never access sensitive information or learn from it. LLMs operate with what Hammond called "goldfish memory," processing queries and then deleting the data immediately. No personally identifiable information passes through the system, and all responses undergo evaluation to prevent insensitive or harmful content from reaching customers.
Beyond data protection, Salesforce enables gradual rollouts. Marketing teams can deploy AI to specific audience segments, starting with just 10% of one audience group, testing results, and scaling up incrementally. This phased approach with built-in guardrails gives organizations the confidence to experiment without risking their brand reputation.
Hammond was quick to point out that these safety measures aren't just about risk mitigation. They're about enabling action.
"With these safety controls in place, with the ability to have moderation and control over the ramp up and deployment, these safety nets are there, and it's really time for companies to start trying more and more of this."
The implication was clear: the infrastructure exists to protect brands, but hesitation itself has become a competitive liability.
This becomes even more apparent when you look at the traffic patterns Hammond has been observing across the customer base.
The Traffic Shift No One Saw Coming
Hammond shared a striking statistic that should make every CX leader pay attention: companies are now reporting that 10% to 20% of their web traffic comes from AI agents rather than humans clicking through pages. In some cases, that figure reaches 40%.
"The end consumer expectation is a far more curated experience," Hammond said. "They're looking for that agent to do the work for them, to discover and look for the patterns they're asking to find and come back with the answers. If a brand isn't providing that experience for customers, their customers are going to go to those who are."
This shift represents a fundamental change in how customers interact with brands. It's no longer enough to have a well-designed website with good SEO. Brands need to provide experiences that AI agents can navigate and extract value from on behalf of customers.