Your CRM Isn't Obsolete. It's the Backbone of Your Agentic System
Every few months, a new wave of artificial intelligence tools arrives with a quiet implication: Maybe you don't need your CRM anymore. That hasn't actually been true in practice, and now the two companies most responsible for enterprise AI adoption have said so explicitly.
When Salesforce and Anthropic announced Claudeforce, they didn't position the CRM as a legacy system around which to route. They called it the harness, the trusted enterprise layer that gives Claude's reasoning somewhere real to land.
Their entire partnership is built on the premise that frontier AI only becomes actionable when it's pointed at the customer information and business context that companies have been building in Salesforce for decades. The CRM, in other words, isn't obsolete. It's the harness or the backbone of an agentic system. To realize the full value of agentic AI, the CRM must remain a living, connected, and continuously improving foundation for transformation.
Nothing reveals the previously overlooked gaps in companies' data architectures quite like AI agents. I discovered this firsthand when I led a team redesigning Salesforce from an agentic perspective. Early in the process, we quickly discovered the consequences of our back-end systems not being fully connected. As a result, missing context was leading the AI to produce incomplete answers, which threatened to make users question whether the agent was actually delivering value.
We had hit a problem many organizations encounter: users asked different AI tools the same question and got different answers because each tool could access different systems.
Of course, those users weren't expected to understand the back-end architecture. That's not their job. But they did need clear communication about what the AI could and could not see at any given time. Without that information, they might have assumed the mixed results meant the AI tool was bad. Adoption would have slowed or stalled altogether.
Visibility into the current state of the implementation helped us avoid that trap. We offered clear, consistent updates on what we were doing so everyone could understand why some things were working and others weren't. Meanwhile, my team doubled down on our efforts to see where our data lives, how it flows, and what context the agent can actually access.
It was a pretty clear lesson: AI agents still depend on a reliable data foundation, which is why the notion that AI will simply replace the CRM doesn't really add up.
AI Agents Change a CRM's Role, But Not Its Core Functions
Working on these implementations has changed how I think about CRMs. The interface might become less important, but the underlying functions like customer context, permissions, workflows, and operational data are just as crucial as they've always been.
Put simply, data has to live somewhere and remain accessible as agents perform work. That doesn't change. The CRM might become less visible in employees' day-to-day workflows, but it continues to provide the critical source of context that informs those workflows.
Take security and governance, for instance, which is crucial to any agentic AI project. CRM permissions and data hierarchies can carry through to agents, letting them know which information users are authorized to access. Employees might ask the agent for the information, but CRM permissions still determine what that agent and that human can access.
That distinction casts the CRM in a different light. Some tasks are still better handled by traditional CRM workflows or automation. And agents are best in situations where they provide genuine additional value.
A renewal alert that fires 90 days before contract expiration, for instance, doesn't need an agent. A standard workflow handles that reliably and cost-effectively every time. But when a sales rep needs a synthesis of 18 months of support tickets, product usage data, and open opportunities before stepping into a renewal conversation, that's where an agent genuinely earns its place.
The distinction isn't really about which approach is more sophisticated. It's about matching the right tool to the actual task.
Once organizations can distinguish, assign, and distribute tasks across agentic and human contributors, they start to understand the new value of the CRM. Instead of being where humans manually enter and retrieve information, it becomes the operational backbone that enables their AI-driven work.
The Path to Scale? Proving Value One Use Case at a Time
It's pretty common to see organizations go into agentic AI transformations with the belief that the technology is the most important factor. But in my experience, successful adoption depends just as much on disciplined business measurement and iteration.
One customer with whom I worked was looking to build a lead-development agent to research and score leads before sending them to the sales team. But even though the AI agent successfully accelerated the processing of leads, the organization soon discovered a different bottleneck: there weren't enough leads entering the pipeline in the first place.
That was a clear demonstration of how AI can expose the next constraint in business processes rather than magically solving the entire process. It's kind of like the old Whack-a-Mole arcade game. One bottleneck reveals another in an endless learning process.
That's why the more successful transformations on which I've worked have generally started with a single, measurable use case that followed a defined path that includes the following steps:
- Identify a specific business problem.
- Establish the current-state baseline.
- Define the expected business benefit.
- Make only the data or CRM changes necessary to support the use case.
- Launch and measure the result.
- Use what is learned to determine the next investment.
The ultimate goal is to realize the value of the investment by connecting AI consumption costs to tangible outcomes. Leadership should be able to say, "We spent $2,500 in AI credits and generated $50,000 in revenue."
That's the kind of statement that shows why the investment matters. And it's the kind of measurable win that provides a rational basis for deciding whether to scale.
Don't Choose Between the Foundation and the Future
Many organizations approach their agentic AI transformations with an apparent choice: invest in the CRM foundation or invest in agentic AI? But an either-or decision is not nearly as effective as taking a little from each and making them work together.
After all, data integrity lies at the heart of agentic AI. And while organizations don't need to perfect their entire data environments before experimenting with AI, they also can't ignore their foundational gaps.
The goal should be iterative readiness. That is, strengthening the foundation where a specific use case requires it, proving the value, learning from the deployment, and expanding from there.
That's how you determine whether your infrastructure is ready to support a new way of working. And your CRM remains at the heart of that discovery process.
Chelsea Monda is director of agentic transformation within the Salesforce practice at Perficient.