OpenAI Is Turning the Enterprise Front Door Into an App Store
For most of the last two decades, the story of enterprise software came down to a single question: who owns the system of record? The companies that could answer it, Oracle, SAP, ServiceNow, became the backbone of the Fortune 500. You bought their software, it went to work inside your operations, and over the years it became so woven in that leaving felt unthinkable. The moat was never really the code. It was the fact that the company's work lived there.
This week, two of those companies did something that quietly inverts that story. On Tuesday, Oracle announced that it would bring its applications and data into ChatGPT and build new Oracle products powered by OpenAI's models. The next day, ServiceNow announced that it would bring OpenAI's models and agents into its own workflows and let OpenAI's agents call ServiceNow workflows from inside ChatGPT and Codex. In other words, two of the most entrenched enterprise software companies on earth plugged their products into a chat window within 48 hours.
It's strange, if you stop to think about it. These are the companies that spent decades being the front door. Oracle is where a huge share of the large companies' records live. ServiceNow is where their IT work gets routed. Their entire business model is being the place where work happens. And yet they are now embedding themselves inside the chat interface of a company barely six years old. The firm that owns the interface is moving into the position they spent decades building.
Both companies frame this, not surprisingly, as a partnership. And the short term reach argument is real, so I want to grant it. If ServiceNow's workflows are reachable from ChatGPT, more people will touch them in contexts the application never had. If Oracle's data is queryable through a conversation, the software gets used in ways a form field never allowed. That's genuinely valuable, and no one on either side is making a mistake today. The reach is real, and the consolidation is something enterprise buyers actually want.
It's also worth reading as an industry wide response rather than a one off. Oracle and ServiceNow did not separately stumble into the same deal. They are answering the same pressure from the same direction. Buyers are starting to ask whether their assistant can do the work, and the answer is increasingly routed through ChatGPT. So the enterprise software companies with the deepest workflows are racing to make themselves callable from that assistant, because being unreachable from the front door is starting to look like a risk of its own. When an entire category moves in the same direction at the same time, it is usually a sign that the pressure is real and structural, not a passing fad.
But the question that matters is a different one, and it is a question about time. When an application lives inside a front door, the front door starts to define the relationship over time. The customer stops thinking I use ServiceNow and starts thinking I ask my assistant, which happens to call ServiceNow. That sounds like a small shift. It isn't.
This pattern has happened before, and everyone in software watched it happen to consumer applications. When software moved from a standalone disk to running inside a browser, and then into an app store, the company that owned the interface came to control the relationship, the discovery, and the payments. The apps still mattered. They shipped, they got used, they made money. But their leverage relative to the platform quietly shrank for years, and by the time the shift was obvious, the position had already moved. The lesson was never that the apps died. It was that they stopped being the front door.
It's not as clean as the partners are being fooled, and I don't think it is. Oracle and ServiceNow have assets OpenAI doesn't: deep domain workflows, years of accumulated data, a compliance posture that took a decade to build, and an installed base that already trusts them. They aren't handing over the front door. They're renting their depth into it. And there is a rational, near term reason to do that. A buyer who manages one interface instead of ten is a buyer who moves faster, and a partner whose work shows up in more places is a partner that grows. The deal makes sense by any reasonable standard.
The risk is subtler than that, and it sits on the partner's side. It is that the front door learns the workflows. Every time an OpenAI agent calls a ServiceNow workflow through ChatGPT, the interface sees what the work looks like, what it touches, and how it flows. The application provides the capability, but the front door accumulates the map. And a map of how a company actually runs is one of the most valuable assets in this whole market. It is the thing that, once you have it, makes you look less like a channel and more like a competitor.
It's worth noting that this isn't a one way street for OpenAI either. OpenAI is betting that the front door is the durable asset, but a front door with nothing behind it is just a browser. The reason Oracle and ServiceNow are valuable to OpenAI is that they carry the depth. If that depth stops flowing, the front door gets shallow, and a shallow front door is exactly the kind of thing a better model can commoditize. So OpenAI is, in a real sense, dependent on the partners too. The two companies are building a shared asset, and the open question is which of them ends up owning more of it.
This is the same dynamic I wrote about when the big labs cut prices a few days ago, seen from the other side. The model layer is being commoditized. Prices are collapsing, and open weight models are pulling the floor down further. So the durable value is migrating to the layers a model can't easily take: distribution, customer relationships, workflow, trust. What OpenAI is doing with Oracle and ServiceNow is a visible instance of a company trying to own exactly those layers. It isn't selling tokens. It's buying the front door.
For anyone building software on top of these platforms, this is the clearest signal yet about where the gravity is. The interface the customer talks to is becoming the most valuable real estate in the stack. If your product is a capability the front door can call, you're a component, and components get repriced. If your product is the front door, you're the platform, and the platform takes the margin. The distinction matters more than it has in years, because the companies deciding the distinction right now are the ones with the most distribution and the most money. And the uncomfortable part is that the distinction is no longer theoretical. It is being decided in boardrooms this quarter, by the two companies with the deepest workflows, choosing which front door to build themselves into.
I don't think Oracle and ServiceNow are making a mistake, and I don't think OpenAI is about to absorb them into nothing. Both sides get something real in the near term, and these are rational deals by any reasonable standard. But the direction of the shift is clear enough to name out loud. Value in enterprise software is moving from the application to the interface, from the system of record to the front door.
For twenty years, the companies that owned where work happened owned the enterprise. The front door is changing hands now, and the companies that built the old one are the ones letting the new one in.