Why turning to your enterprise applications for an AI strategy is a well-intentioned mistake — and what to do instead.
Every enterprise software vendor has an AI story right now. Salesforce has Agentforce, ServiceNow has AI Agents, Zendesk has their suite of intelligent automation, and if you’ve been in a sales meeting with any of these vendors in the last twelve months, you’ve probably heard some version of the same thing: Turn on our AI capabilities. Your data is already here. Let’s put it to work.
It’s a compelling pitch, but it’s also an incomplete one, and building your AI strategy around it is one of the most well-intentioned missteps I’m seeing businesses make right now.
Give Credit Where It's Due
Let’s be fair to the vendors; they’re not wrong that AI can create real value inside their platforms. A sales team using AI-powered lead scoring in Salesforce can absolutely prioritize their pipeline better, a support team using Zendesk’s AI can route and resolve tickets faster, and a ServiceNow deployment with intelligent workflows can meaningfully cut resolution times and reduce manual triage. Those are real, tangible gains that nobody should dismiss.
But the problem starts when a business looks at one of those platforms and says, “This is our AI strategy.” At that point, they’re confusing a feature with a foundation, mistaking what an application can do with what the enterprise actually needs, and that distinction matters more than most leaders realize.
The Data Reality Nobody Wants to Confront
Think about what actually happens when a sales team tries to use AI to forecast revenue or prioritize accounts. The CRM has deal data, activity logs, and contact records, which is useful, but the most predictive signals almost always live somewhere else, in ERP systems tracking order history and fulfillment patterns, in marketing platforms capturing engagement and intent data, in support systems revealing customer health, in financial systems showing payment behavior, or in product telemetry showing actual usage.
I see this every single day in our client work: no single application holds the full picture. The average enterprise runs somewhere between 100 and 300+ applications, and the data that would make AI truly transformative. The kind of data that lets you anticipate churn before it shows up in a support ticket, identify cross-sell opportunities based on operational patterns, or optimize supply chain decisions using demand signals from multiple channels is scattered across the entire technology landscape. This isn’t a knock on any particular vendor; it’s just the reality of how businesses operate. Data accumulates where work happens, and work happens everywhere.
Application AI vs. Enterprise AI
This is where the conversation needs to shift, because there’s a meaningful difference between what I’d call application-level AI and enterprise-level AI, and conflating the two leads to real strategic misalignment.
Application-level AI operates within the boundaries of a single platform; it uses the data that lives there to automate tasks, surface insights, and improve workflows inside that tool’s scope. It’s valuable, but inherently limited by what the application can see. Enterprise-level AI, on the other hand, operates across the business, drawing on data from every relevant source, CRM, ERP, marketing, support, finance, product, and operations, to deliver insights and actions that no single application could generate on its own. That’s the kind of AI that actually augments how people work across the entire organization.
The difference isn’t academic either. We’re working with clients right now who are running into hard constraints with Agentforce, such as architectural limits on agents and topics per deployment, behavioral inconsistencies driven by LLM variability, and cost unpredictability with consumption-based pricing. And even a perfectly executed Agentforce deployment can only reason over what Salesforce knows, so if the most important context for a decision lives outside the CRM, which it almost always does, the AI is working with an incomplete picture.
The same story plays out with ServiceNow. They’re investing heavily in their Workflow Data Fabric and agentic AI capabilities, and the vision is compelling, but the implementation complexity is real. Connecting to external sources introduces licensing costs, integration overhead, and a search experience that doesn’t match what you get with native data. The pattern is consistent across all of these vendors: each one is optimizing AI for their own ecosystem, which is a smart product strategy for them, but it’s not a comprehensive AI strategy for you.
Data Strategy Is the Strategy
So if enterprise AI requires enterprise data, the question that actually matters is: how do you get there?
It starts with treating data as a strategic asset rather than a byproduct of application usage, which means investing in a data foundation that can unify structured, semi-structured, and unstructured data from across the business — not to replace your applications, but to create a layer where AI can reason over the full context of your enterprise.
This is where platforms like Snowflake’s AI Data Cloud become relevant, not as a replacement for your CRM or ITSM tool, but as the connective tissue that brings enterprise data together in a governed, secure, and AI-ready environment. There’s a principle gaining real traction here that I think matters: bring the models to the data, not the data to the models. Through capabilities like Cortex AI, organizations can run frontier AI models — including Claude and GPT — directly against their enterprise data without moving it into separate environments or sacrificing governance controls. The SAP-Snowflake partnership, with its zero-copy bidirectional data sharing, shows what this looks like in practice: enterprise application data becoming accessible for AI without the traditional ETL complexity, and similar integrations with platforms like Google Drive, Workday, Box, and Zendesk through Snowflake Openflow are extending this pattern further.
I want to be clear that the point isn’t Snowflake as the only path forward — the point is that the architecture matters. You need a place where data from across the enterprise can converge, be governed, and be made available to AI workloads, whether those workloads are generating forecasts, powering agents, or informing strategic decisions.
What This Means for Your Team
This isn’t just a technology conversation, right? It’s an organizational one. When your AI strategy is anchored to a single application, you end up with pockets of intelligence — a sales team with AI-powered insights that don’t connect to what operations sees, a support team with automated workflows that can’t account for what marketing knows about the customer, a finance team making projections that don’t reflect what’s happening on the ground. Those silos don’t just limit the AI; they limit how your people work together.
When your AI strategy is anchored to your data strategy, though, you create the conditions for what we at 7Rivers call the augmented enterprise — an organization where human insight and AI work together across every business function, drawing on the same unified foundation of data. That’s the real opportunity: not AI that works in one tool, but AI that works across the business.
Getting Started (Without Boiling the Ocean)
None of this means you need to pause everything and rebuild your data infrastructure from scratch — the path forward is usually more practical than that.
Start with business outcomes, not technology. Identify the highest-value use cases across your five core business spheres — customer, partner, employee, operations, and product — and map the data those use cases actually require. More often than not, you’ll find the data spans multiple systems, which tells you something important about where your foundation needs to be.
Assess your current data landscape with eyes wide open. Where does your most valuable data live? How much of it is accessible, governed, and AI-ready? Where are the gaps? This is the foundation for any real AI roadmap, and the range of what I’ve seen is enormous. I worked with a manufacturer in the Midwest whose IT systems were so siloed that we ran a POC and they immediately realized there was no bridge between their operational data, their finance data, and their people data. On the other hand, I’ve worked with financial services firms whose data posture is so mature that they’re deploying conversational analytics on day one. The starting points are wildly different, and knowing where you stand is what makes the roadmap real.
Build the data foundation in parallel with application-level wins. You don’t have to choose between using Agentforce today and building a long-term enterprise data strategy — do both, but make sure the short-term wins feed into the larger architecture rather than creating more silos.
Invest in the people side. The skills gap in enterprise AI is real, and teams need to understand not just how to use AI tools, but how to think about data as a cross-functional asset. That’s a cultural shift as much as a technical one. And honestly, AI is going to accelerate the gap between the haves and the have-nots — people who know how to work with data today can now do it ten times faster, and the ones who don’t are going to fall further behind.
The Bottom Line
The enterprise software vendors building AI into their platforms are doing important work, and those capabilities have genuine value that should be part of your toolkit. But they’re not your AI strategy — they can’t be, not when the data that matters most lives across a dozen or more systems and the real power of AI comes from connecting all of those dots.
Your AI strategy is a data strategy. Full stop. Everything else is a feature.
