For years, customer experience teams have been sold the same promise: put a chatbot on the front door, deflect some contacts, and call it transformation.
The problem is that most of those bots are flat.
They can retrieve a help article. They can recognize a keyword. They can point a customer toward a form, a phone number, or another queue. But when a customer has a real problem—an exception, a missing order, an account issue, or a complex request—they often reach the same dead end: “I can’t help with that.”
That is why I have started calling them flatbots.
At Dreamforce 2026, walking through Salesforce’s Agentic Enterprise City and speaking with brands including F1, Live Nation, Crocs, and SPECS, I saw a different ambition emerging. These organizations are not simply trying to make chat more conversational. They are trying to connect AI to the data, workflows, and actions required to actually solve a customer problem.
That distinction matters.
Different Deployment Methods, the Same Destination
The most impressive thing was not that every brand had deployed AI in exactly the same way. They had not.
The range of approaches was striking. Some were using prebuilt Agentforce interfaces and agents. Others were using Salesforce’s newer headless capabilities to create their own front ends on top of Salesforce data and workflows. Some were using AI-assisted—or “vibe-coded”—development to create new experiences for customers, frontline teams, managers, and leaders.
Different deployment methods. The same destination: personalize the experience, resolve more issues digitally, reduce avoidable escalations, and create commercial value from better customer interactions.
That is a much more meaningful ambition than chatbot containment.
From Digital Response to Digital Resolution
Fin, Salesforce’s most recent acquisition, brought a useful proof point to the conversation. It says its customers achieve an average 76% resolution rate through digital interactions.
That is an impressive vendor-provided figure, but the more important point is what it represents: a customer getting the outcome they need without repeating themselves, waiting in another queue, or beginning the journey again with a human agent.
Of course, a number alone does not settle the CX case. Leaders should ask what is being counted as resolved, what happens to repeat contacts, and whether customers are satisfied with the handoff when the AI cannot continue.
But it is a more useful measure than asking whether a bot simply responded.
Salesforce Wants CRM to Become the Layer Beneath the Experience
Salesforce’s wider Dreamforce strategy is built around this shift. Its AIforce proposition is that the traditional CRM interface matters less than the governed layer underneath it: customer data, permissions, business rules, workflows, security controls, and approved actions.
That opens up new possibilities. A customer-facing agent could use the same trusted context as a service representative. A service leader could work from a live operational view built around the problem in front of them. An employee could get the information and action they need in Slack, Claude, or an organization’s own interface rather than switching between multiple applications.
But this is where the excitement needs to meet reality.
Better AI Will Not Fix a Broken Customer Operation
A better AI model will not fix a broken customer operation.
If your customer data is fragmented, policies are ambiguous, workflows are disconnected, and escalation paths are poorly designed, AI will not magically create resolution. It may simply expose those problems more quickly—and at a larger scale.
That is why CX leaders now need to make a more deliberate choice about what they want from this next wave of technology.
Five Questions for CX Leaders
Do you want it easy? Start with a prebuilt agent and a well-defined, high-volume journey. A straightforward returns process, account update, or order-status workflow may be the fastest place to prove value.
Do you need it customized? Build around the workflows, terminology, channels, and decision boundaries that make your customer experience distinctive. The interface should fit the customer journey, not force the journey to fit the technology.
Do you want to vibe code new experiences? Then make sure the speed of development does not outrun security, accessibility, data quality, or operational accountability.
Do you want choice of model? Define which jobs require which models, what customer data they can access, and how you will evaluate their performance beyond fluent answers.
Do you want AI fully integrated into your estate? Then treat it as a transformation programme—not a chatbot project. That means connecting systems, clarifying ownership, setting action permissions, designing human handoffs, and giving CX a seat in AI governance.
The Flatbot Is Not Enough Anymore
The flatbot is not dead because chat no longer matters. It is dead because customers have outgrown it.
They do not want another polite digital assistant that tells them where to go. They want their problem understood, their context recognized, and the right action taken.
The vendors are rapidly giving CX teams more ways to build that future. The harder question is whether organizations are ready to do the work required to make it real.
