The Signal Behind the Spending

Ramp's AI Index measures something straightforward: actual money. Instead of asking company leaders what they plan to spend on AI or how excited they are about it, the index pulls real transaction records from corporate credit cards and payment systems used by more than 70,000 U.S. companies on Ramp's platform. Published monthly, it tracks what fraction of those companies are paying for AI services. The result is one of the clearest windows into what businesses are actually doing with AI, not what they say they plan to do.

This methodological choice matters. Over the past three decades, I have watched sentiment surveys consistently overstate adoption during hype cycles and understate it during quieter phases when companies are building infrastructure behind the scenes. Spend data is harder to misread: money moved from an account, a contract was signed, a service was activated. It is harder to fake than a response on a survey.

The Top 1%: A Category of Their Own

The most striking finding in the index is the spending profile of what Ramp calls the "AI-pilled" firms — the top one percent of companies on the platform by AI expenditure. These businesses spend $7,500 per employee per month on AI tools and services, according to Ramp's data. That is a monthly figure, not an annual budget divided up.

To put that in context: $7,500 per employee per month adds up to $90,000 per employee per year, spent entirely on AI subscriptions and services. For a 50-person team, that is $4.5 million annually — roughly what you would pay in salary for several senior engineers, except this money goes toward buying AI capability instead of hiring people.

Who are these firms? The index does not name them, but the spending level tells you something structural. At $7,500 per person per month, companies are not buying a basic ChatGPT subscription and a few Microsoft Copilot licenses. They are running custom AI models, processing huge volumes of AI requests through APIs, or using specialized business software built on foundation models — or combinations of all three. In other words, AI is core to how these businesses operate, not a side tool.

The spending figure does come with a caveat worth keeping in mind. It measures spending per employee overall, but that masks variation: a 10-person AI startup with a large monthly bill for model access will look identical in this metric to a 10-person AI team inside a bigger company with a dedicated budget. The index compresses that difference.

Subscription Penetration as the Real Signal

Beyond the outlier cohort, the more broadly useful number is subscription penetration — what fraction of all 70,000-plus firms use at least one paid AI service in a given month. This running count tells you how far AI adoption has spread across the full spectrum of U.S. businesses, not just tech pioneers but companies of all sizes and types that use Ramp's platform.

This matters because Ramp's customer base is heavily weighted toward small and mid-market companies — the segment that tends to be overlooked in tech coverage, which focuses on big corporations and hyperscalers. Penetration data from this group offers a more realistic picture of whether AI tools are becoming normal business infrastructure or whether they remain concentrated among early adopters.

We have seen this pattern before. In the mid-2000s, cloud computing adoption followed nearly the same path: a handful of startups ran everything on AWS from the start, a broader wave of tech-savvy small and medium businesses moved to the cloud within two to three years, and the rest of the business world spent a decade or more migrating data and applications off their own servers. The Ramp index, tracked month by month, is essentially a tracker of where AI is on that same adoption curve — and how fast it is climbing.

What the Data Shows and What It Misses

Transaction data has blind spots. It only captures spending that flows through Ramp's payment system. Companies using other corporate card providers, making large payments directly from bank accounts, or running their own AI infrastructure — leasing GPUs, hosting open models like Llama or Mistral on their own servers, or building custom tools — do not show up here. Big cloud bills from Amazon, Microsoft, or Google might also be undercounted if they are paid through separate agreements rather than Ramp's platform.

The index also covers only U.S. companies, which limits what you can conclude about global AI adoption but makes it more precise for understanding the American business market. For investors, policy makers, or companies trying to size the U.S. AI market, a panel of 70,000 firms is large and detailed enough to be reliable.

What the Spending Really Means

The $7,500-per-employee-per-month figure for top spenders signals something important: at that level, AI is no longer a line item under "software subscriptions." It is a central investment in business capability, sitting alongside computing power, networks, and security in the core infrastructure budget.

For most companies not yet at that level, the more relevant question is simpler: what does the subscription penetration curve look like across the full 70,000 firms, and how quickly is it rising? That monthly count — the shift from zero paid AI services to at least one — is where the real adoption story is written, one transaction at a time.

The Ramp AI Index does not tell you which tools are winning, which vendors are struggling, or whether the productivity gains justify the cost. Transaction data alone cannot answer those questions. What it does provide is a relatively straightforward, regularly updated picture of how widely and intensely U.S. companies are adopting commercial AI — grounded in what they actually pay for rather than what they claim to intend.