This year's Dreamforce was undeniably dominated by one feature: Agentforce. With it, Salesforce is showcasing an AI-powered platform designed to automate complex business tasks. On the surface, it sounds revolutionary. But let’s take a step back—are businesses really ready to hand over the reins to fully autonomous agents in sales, marketing, and service?
Agentforce, powered by the Atlas Reasoning Engine and integrated with Data Cloud, promises intelligent decision-making and enhanced efficiency. In my initial experience with it, I can see the potential. But potential was the key theme of Dreamforce. Over three days of celebrity appearances, industry gurus, and elaborate AI demos, there was a sense of hype that often overshadowed the practical realities. While the promises are big, there are key concerns and challenges that were left largely unaddressed.
The reality is that businesses still have to trust AI with critical processes on a large scale—and many don’t, for good reason. For any leader, this moment is frustrating. There are some clear, immediate AI use cases, like case deflection and AI-generated summaries, that should be a no-brainer for companies. The potential for cost savings is significant, especially for enterprises. Even more compelling is the realization that automating thought and decision-making can lead to direct margin increases post-implementation. This means that businesses must adopt these agents to stay competitive, or risk being left behind.
However, this leads to a major disruption in business strategy. Many companies, especially those that didn’t foresee this rapid shift, are unprepared. The technical and operational roadmaps of countless organizations will need to pivot quickly to accommodate AI-driven changes. But here’s the crux of the issue: not every business will be able to adopt Agentforce easily. Regulatory limitations, particularly around security and trust, will act as major hurdles.
One of the most insightful stories I heard at Dreamforce revolved around deploying digital agents in highly regulated industries, like finance and healthcare. These industries have specific requirements for human agents, such as verifying whether the person they're speaking with is actually an authorized customer or representative. With AI, personally identifiable information (PII) has to be masked before it can be processed by a model, meaning digital agents can't handle customer authentication and authorization on their own. This is a huge limitation. You can’t just transfer these processes to a digital agent.
The lesson here is that, just like before, companies will need to approach this transition thoughtfully. Scaling fleets of digital agents will require careful architecting, with expertise in navigating the interplay between people, processes, and technology—similar ODNOS approaches solutions with our Nexus design methodology. Digital agents have limitations, and those who rush to adopt them without considering these constraints will hit roadblocks.
Early adopters will see gains, but they need to tread carefully, especially if they’re managing rapid growth and accumulated technical debt. This is something Salesforce clearly anticipated, which is why Dreamforce was filled with big-picture promises rather than detailed, substantive updates. Salesforce’s Data Cloud consolidation and movement of off-platform features to the core Force.com platform are strategic moves, designed to tackle the challenges of scaling AI.
Here’s why this shift is important:
Salesforce is migrating to external data centers, which likely give them the capability to handle the data explosion AI will bring. Whether this is a full migration or an augmentation of their own data capabilities is to be seen.
It’s moving to a consumption-based licensing model for that augmented data/compute need, more in line with AWS's pricing structure.
They are consolidating functional areas into the Force.com platform. This reduces the need for complex maintenance across expanded technical domains, such as older acquisitions like Marketing Cloud or Demandware, and even sector-specific models like NPSP.
These changes are necessary for Salesforce's Trust Layer to manage secure connections with Large Language Models (LLMs) and for the control necessary to be able to mask data properly.
Salesforce is making a big bet with its $1B AI investment. But the critical question isn’t whether the tech can perform; it’s whether the data security and trust will hold up at scale. That’s the big issue facing business leaders right now. Dreamforce was all about showcasing possibilities, demonstrating that Salesforce is positioning itself for the AI age. But the underlying message, though not explicitly stated, was clear: Salesforce has invested heavily in overhauling its platform, and now it’s time for its customers to follow suit—or risk falling behind.
For businesses, this raises a bigger dilemma. When an entire platform evolves beneath your feet, you’re left with no choice but to adapt. And while for some, this will be a net positive, for others, the prospect of being locked into Salesforce's vision of the future is unsettling. Flexibility to innovate beyond what the platform allows might be more critical to their success than implementing features like chat summaries. Dreamforce may have been about exciting possibilities, but it also reinforced an ongoing concern: when you invest everything in a single platform, you’re bound by its evolution—whether you're ready or not.
Comments