The walls are closing in on data silos. The rise of ChatGPT as the front-end to our AI world is pushing them to open. And legacy B2B and SaaS vendors are about to learn a painful lesson about who really owns customer data. Albeit not without some battles.
Fight it like Salesforce + Slack? Or embrace is like HubSpot?
The Salesforce-Slack Moment That Changes Everything
Salesforce just fired the first shot in what will become the defining battle of enterprise software’s next decade. By blocking AI rivals like Glean from accessing Slack data, they’ve crystallized the central question every SaaS vendor will face: Is the data in your app YOUR data, or your customer’s data?
The answer will determine which vendors survive the AI revolution and which become relics of the pre-AI era.
Per Jamin Ball:
So what happened? The TLDR: Slack is making it significantly harder, or in some cases outright blocking, customers from using their own Slack data (stored conversations) the way they might want to. There are now tighter restrictions around storing, indexing, or copying Slack data accessed via the API. They’ve also introduced strict new rate limits: for methods like
conversations.historyorconversations.replies, you’re now limited to 1 request per minute, capped at 15 messages per call.And most critically, they’re blocking third-party platforms from indexing, copying, or storing Slack data, even if the customer wants that to happen.
What MCP Really Means (And Why Some Legacy Vendors Are Terrified, and The Rest Are Embracing It)
Model Context Protocol (MCP) isn’t just another technical standard. It’s the HTTP of AI—a simple, universal way for AI systems to connect to any data source. Think of it as the moment when every website could suddenly talk to every browser, instead of requiring proprietary plugins.
For customers, MCP means freedom. Their AI assistant can pull data from Salesforce, analyze it alongside Slack conversations, cross-reference it with Notion docs, and generate insights from their entire business ecosystem—all through a single interface.
For legacy SaaS vendors, MCP represents an existential threat to their most valuable asset: data lock-in.
Some are embracing it as the future, like HubSpot. But many classic B2B and SaaS leaders will likely fight it.
The Great Data Hostage Crisis
Here’s the uncomfortable truth about enterprise software: Most “sticky” B2B and SaaS companies aren’t sticky because they’re great products. They’re sticky because they’re data prisons.
They stay because moving 10 years of customer interactions, deal history, and custom workflows to another platform is a $2M, 18-36 month nightmare.
This data hostage model has worked brilliantly for two decades. But AI changes everything.
When customers can access their Salesforce data through ChatGPT, analyze it with Claude, and generate insights without ever opening the actual Salesforce interface, what exactly are they paying $150-$300/user/month for?
A Coming Customer Revolt?
The shift won’t happen overnight. Legacy enterprise buyers—your typical VP of Sales who’s been using Salesforce since 2015—will keep paying for the familiar.
But new customers? They’re different. They’ve grown up with AI. They expect their tools to work together seamlessly. They’ll evaluate vendors on one simple question: “Can I use this data with my AI?”
The answer better be yes.
The New Buyer Checklist:
- Does this vendor support MCP or similar open protocols?
- Can I use ChatGPT/Claude as my primary interface?
- Will my data be accessible if I want to switch vendors?
- Does this integrate with my AI workflow, or force me into theirs?
Vendors who answer “no” to these questions will find themselves eliminated from consideration before the first demo.
The False Choice: Integration vs. Innovation
Legacy vendors are making a fatal mistake. They’re treating AI integration as a zero-sum game—either customers use their AI features, or they lose the customer entirely.
But that’s not how customers think. They don’t want to choose between your AI and their AI. They want their AI to work with your data.
Smart vendors understand this. They’re racing to implement MCP, building robust APIs, and positioning themselves as AI-native infrastructure rather than AI competitors.
The winning strategy isn’t “use our AI.” It’s “use our data with whatever AI you prefer.”
Why Data Portability Becomes Competitive Advantage
This creates a counterintuitive dynamic: The vendors who make it easiest to leave will be the hardest to leave.
When customers can use their preferred AI tools with your data, they’re more likely to choose you in the first place. When they can easily export their data, they’re more confident investing deeply in your platform.
It’s the same psychological principle that made AWS dominant. Customers trusted AWS partly because they knew she could leave if needed. That trust led to deeper, stickier relationships.
The Three-Year Timeline
Here’s how this plays out:
Year 1 (2025): Early adopters start requiring MCP support in RFPs. Legacy vendors dismiss this as “a small segment” and double down on their integrated AI features.
Year 2 (2026): The small segment becomes the majority of new deals. Vendors who moved early start winning competitive battles purely on data accessibility. Legacy vendors begin panicked efforts to retrofit openness.
Year 3 (2027): Data portability becomes table stakes. Vendors who built walls around their data find themselves with aging customer bases and declining new acquisition. The market has moved beyond them.
The Salesforce Test Case
Salesforce’s decision to block Slack data access will be studied in business schools as either brilliant or catastrophic—depending on what happens next.
If they can provide superior AI experiences within their ecosystem, they might justify the walls. If they can’t, they’ve just painted a target on their back for every competitor building MCP-native alternatives.
The early signs aren’t promising. Salesforce’s AI features are… fine. But “fine” doesn’t justify data imprisonment when customers can get “excellent” AI experiences elsewhere.
What This Means for Every B2B / SaaS Vendor
Every B2B and SaaS company needs to ask themselves: Are we a data company or a workflow company?
Data companies (like Salesforce, HubSpot, or Workday or even LinkedIn) are most vulnerable. Their value prop is largely “we store and organize your important business data.” If customers can access that data through better interfaces, what’s left?
Workflow companies (like Slack, Notion, or Figma) are more defensible. Their value is in the collaborative experience, not just the data storage. Even if customers can query their data externally, they still need the platform for daily work.
The smartest vendors are evolving from data companies to workflow companies. They’re making their data radically accessible while doubling down on experiences that can’t be replicated elsewhere.
The Open Data Manifesto
Here’s what winning looks like in the MCP era:
- Embrace radical transparency: Your customers’ data belongs to your customers, period.
- Compete on experience, not lock-in: Make your daily workflow so valuable that customers choose you even when they can easily leave.
- Become AI-infrastructure: Position yourself as the best place for AI to access domain-specific data, not as an AI competitor.
- Lead with openness: Make MCP support a core differentiator, as HubSpot has, not a reluctant afterthought.
The Bottom Line
The data hostage model is dying. Customers are demanding their data freedom. And AI is giving them the tools to get it.
Vendors can fight this trend and slowly lose relevance, or they can embrace it and build the next generation of enterprise software.
The choice is binary. And the window for choosing is closing fast.
The question isn’t whether your customers will demand data portability. The question is whether you’ll give it to them voluntarily, or whether they’ll take it by force—by choosing your competitors.
The MCP revolution isn’t coming. It’s here. And it’s time to pick a side.
