The Real Threat AI Poses to SaaS: It’s Not What You Think
SaaS is dead! AI will replace everything!
Not so fast, says Ali Ghodsi, CEO of Databricks. His company just hit $5.4 billion in annual revenue, growing 65% year-over-year. But his perspective on AI’s actual threat to SaaS businesses is far more nuanced—and frankly, more alarming—than the “vibe-coded replacements” meme suggests.
The Core Insight
The conventional wisdom is wrong. Companies aren’t going to rip out their Salesforce or SAP installations and replace them with AI-generated alternatives. These are systems of record—critical databases storing essential business data. Moving them is painful, risky, and unnecessary.
“Why would you move your system of record? You know, it’s hard to move it,” Ghodsi notes.
The model makers aren’t even trying to replace these databases. They’re not offering storage infrastructure. Instead, they’re going after something more fundamental: the user interface itself.
Here’s the real threat: once the interface becomes natural language, the product becomes invisible. Like plumbing.
Why This Matters
For decades, the biggest moat for enterprise software companies wasn’t their features—it was the trained workforce. Millions of people globally became Salesforce specialists, ServiceNow experts, SAP consultants. Learning these complex interfaces was a career investment.
“Millions of people around the world got trained on those user interfaces. And so that was the biggest moat that those businesses have.”
That moat is evaporating. When anyone can interact with any system through natural language, the interface expertise becomes worthless. The switching costs that protected enterprise software vendors for decades suddenly collapse.
Databricks itself is benefiting from this shift. Their Genie product—an LLM interface for their data warehouse—is driving growth because queries that once required specific technical language can now be asked in plain English. Ghodsi uses it to ask why revenue spikes on particular days, a question that previously needed custom reports.
Key Takeaways
AI won’t replace systems of record. Your company’s Salesforce data isn’t going anywhere. The database stays.
AI will replace user interfaces. Natural language interfaces make every product equally accessible to non-experts.
Interface expertise is depreciating. The career path of becoming a “Salesforce specialist” or “SAP expert” is under threat.
SaaS companies that embrace LLM interfaces can grow. Databricks is proof—AI products drove $1.4B of their revenue.
But the door opens to AI-native competitors. Products designed for agent interaction from the ground up may outperform legacy systems with LLM interfaces bolted on.
New infrastructure is emerging. Databricks launched Lakebase, a database designed specifically for AI agents. Eight months in, it’s generating twice the revenue their data warehouse did at the same age.
Looking Ahead
Ghodsi’s framing resets the SaaS-versus-AI debate. The question isn’t whether AI will destroy SaaS businesses—it’s which SaaS companies will adapt and which will become invisible plumbing operated by someone else’s AI agent.
The companies that win will be those that recognize the interface is no longer their moat. Instead, they’ll compete on data quality, integration depth, and API design. They’ll build for agents first, humans second.
For enterprise software buyers, this is actually good news. Vendor lock-in through interface complexity is weakening. When your AI can talk to any system, the power shifts back to the customer.
For enterprise software sellers? Time to rethink what actually differentiates your product when the interface becomes a commodity.
Based on analysis of “Databricks CEO says SaaS isn’t dead, but AI will soon make it irrelevant” from TechCrunch