Enterprise AI Spending Just Overtook Consumer. Your Most Expensive Support Problem Never Saw a Dime.

The lines crossed. CNBC reported last week that OpenAI CFO Sarah Friar told investors the majority of the company’s revenue now comes from enterprise customers. The year opened at a 60-40 split in favor of consumers. Then enterprise AI spending accelerated faster than OpenAI itself expected, and the business side pulled ahead for good.

That is a genuinely big signal. Companies, not consumers, now fund the largest AI vendor on the planet. But here is the uncomfortable truth hiding inside that milestone. Almost none of that enterprise AI spending reaches the support problems that drain the most money from enterprise budgets.

Where Enterprise AI Spending Actually Goes

Follow the purchase orders and a pattern appears quickly. Copilots for knowledge workers. Chatbots for password resets and billing questions. Agents that draft documents, summarize meetings, and write code. Every one of these deals shares a common trait. The work involved is text.

That makes sense, because language models eat text for breakfast. If a problem can be typed, an AI system can usually read it, reason about it, and respond. So enterprise AI spending flows toward the problems that fit the tool. Procurement teams buy what demos well, and text demos beautifully on a conference room screen.

Meanwhile, the most expensive support cases in your company contain no text at all.

The Problem That Never Made the Budget

Picture the support cases that actually hurt. A production line sensor throws an error nobody can interpret over the phone. A restaurant’s payment terminal blinks a pattern the manager cannot describe. A medical device beeps in a clinic three hours from the nearest technician. These are physical problems. They live in the real world, not in a ticket field.

Today, most companies handle them the same way they did in 2005. An agent asks questions. The customer struggles to answer. The agent gives up and schedules a visit. A technician drives out, often to discover the fix takes four minutes, or that the right part sits back at the warehouse.

Every one of those visits costs hundreds of dollars, sometimes more than a thousand. First-time fix rates across field service have barely moved in years, even as AI budgets exploded. We wrote about this gap before, in Where the Money Goes When You Cut 4,000 Support Agents. Companies celebrate the savings from automating text tickets while the physical escalations keep bleeding out the back door.

And the milestone makes the imbalance sharper, not softer. If the majority of AI revenue now comes from enterprises, then enterprises are collectively placing an enormous bet that intelligence alone resolves problems. It does not. Intelligence plus eyes resolves problems.

Intelligence Without Eyes Is Half a Support Strategy

Ask a simple question about any AI system your company bought this year. Can it see what your customer is looking at? For nearly all of them, the answer is no. They read tickets about the jammed machine. They summarize the case history of the jammed machine. They cannot look at the jammed machine.

This is why visual context is the highest-leverage line item missing from most enterprise AI spending plans. The technology is not exotic. Your customer already holds a camera. A remote video support session turns that phone camera into your support team’s eyes. The agent sees the blinking terminal, the miswired cable, the error code on the screen, live, in the first ten minutes of the case.

The economics follow immediately. Some visits stop existing, because the agent guides the fix on camera. The visits that survive get cheaper, because the technician arrives knowing the model number, the symptom, and the right part. Diagnosis moves from the driveway to the first phone call.

The Line Worth Crossing Next

OpenAI crossing the enterprise line tells you where the market’s attention went. It does not tell you where your margin leaks. For any business that supports physical products, the leak sits in dispatches, repeat visits, and the long minutes agents spend asking customers to describe things a camera would show instantly.

So here is a practical suggestion for your next budget cycle. Keep the copilots. Keep the chatbots doing tier-one triage. Then take a small slice of your enterprise AI spending, point it at visual context, and measure what happens to truck rolls and first-time fix rates for a quarter. Tools like Viewabo make this a low-friction test. The customer taps a link, no app install, and your agent sees the problem within seconds.

The companies that win the next phase of AI in support will not be the ones with the most licenses. They will be the ones that matched the tool to the problem. Text tools for text problems. Eyes for physical ones.

Enterprise AI spending just proved it can grow faster than anyone predicted. The question for support leaders is simpler and older than any model release. When your customer stands in front of a broken machine, can anyone on your team actually see it?