The 1:1 Ratio

Most AI maturity assessments ask people to rate themselves. That's the problem. People systematically overestimate how embedded AI is in their operations, and the survey instrument has no way to correct for it. A team that uses ChatGPT occasionally for drafting and has a Copilot license rolled out will describe itself as "advanced" or "scaling" in any self-assessment framework, because those frameworks don't have a grounding mechanism. They're measuring narrative sophistication, not operational reality.

There's a better diagnostic, and it's a single ratio: AI token spend divided by salary spend. Pull both numbers, divide, and you know where your organization actually sits on the deployment spectrum. The best AI-embedded companies in 2026 are approaching 1:1. Most companies that describe themselves as AI-forward are at 1:1,000 or worse. And most of them haven't checked.

What the ratio measures

The metric works because token spend is a proxy for computational cognitive work being done by AI. Salary spend is the cost of human cognitive labor. The ratio measures how much AI labor is running in parallel with human labor. You can't inflate it with a survey response. Either the API is being called at volume or it isn't. That fact lives in a billing dashboard, not in a judgment call.

What different ratios look like in practice:

At 1:1,000, which is most companies in 2026 regardless of what the board deck says, some employees use ChatGPT occasionally. There may be a Copilot license rollout. AI isn't meaningfully in the operating model. The ratio reveals this immediately; the maturity survey often doesn't.

At 1:100, you have a meaningful but bounded deployment. One or two workflows are genuinely automated. The teams running those workflows are saving real time. But the deployment hasn't spread because the first implementation wasn't designed to demonstrate replicability, or because there's no internal champion connecting the proof of concept to adjacent use cases.

At 1:10, AI is embedded in several core workflows, probably including customer-facing ones. Someone in the organization tracks AI costs as a meaningful budget line item. Productivity impacts are measurable and have been measured. This is where genuine operational change starts to be visible.

At 1:1, AI is co-running the operational core. People are managing AI outputs and handling exceptions, not doing the underlying work. EXL's 2026 research calls organizations at this level "AI leaders" and documents the outcomes: 26% cost reduction, 27% revenue increase, 22% margin improvement. These aren't projections. They're measured actuals from organizations that got there.

Why the usual metrics measure the wrong thing

The alternatives to this metric all have the same structural problem: they measure effort and intent, not volume of AI work actually being done.

Survey-based maturity models ask people to evaluate themselves. They produce distributions that cluster toward "advanced" because the self-assessment is a social act and because the frameworks don't penalize confident overstatement.

Use-case counts treat every deployment as equivalent. A use case that saves one person thirty minutes a week counts the same as one that eliminated a cost center or transformed a revenue-generating process. The count tells you how many places AI has been introduced. It doesn't tell you how much AI labor is running.

Board-deck AI slides are the most detached from reality. They measure how well a company can present AI strategy to investors, which is a separable skill from how much AI is actually in the operating model. A company can have three slides on AI and a 1:1,000 ratio. It happens constantly.

Token spend is different because you can't narrate your way to a higher number. Either the integrations exist, the workflows are running, and the APIs are being called at volume, or they aren't. The billing record is the ground truth. There's no way to describe your way to a higher number on the invoice.

How to run the calculation

How to self-assess:

Pull total AI API spend for the last thirty days across all connected models. That means OpenAI, Anthropic, Azure OpenAI, Google Vertex, and any other inference endpoints your organization runs through, including SaaS applications that expose their own AI spend in their billing view. Add them up.

Divide by monthly payroll cost.

The ratio locates you on the deployment spectrum. If you're at 1:1,000, you're in exploration mode regardless of how you'd characterize your AI strategy. If you're at 1:100, you have a proof of concept that hasn't been designed for scale. If you're at 1:10, you're in the minority of organizations that have genuinely changed how work gets done.

To improve the ratio, the question isn't "how do we deploy more AI tools." Tools without workflow integration don't move the ratio. The question is which workflows should AI be co-running, and what organizational and data conditions need to be met before that's possible. That's a different planning question than "which AI tools should we evaluate." It's an operational redesign question, and the ratio is how you track whether the redesign is working.

Check it before the next planning cycle

Run the calculation before the next planning cycle. Not as a point of pride or a benchmark to report externally, but as a grounding mechanism. The number will tell you something about your actual state that your self-assessment won't. If there's a gap between where you thought you were and where the ratio puts you, close the gap in the planning, not in how you describe the situation.

If the number surprises you, it's doing its job. If it doesn't, you probably haven't checked.

---

If you want that question answered for your specific situation, the Forge Playbook does it. Answer a few questions about your business and we'll put together a tailored outline of which workflows are worth automating and what a realistic budget looks like for each. Free, no obligation, takes about three minutes.

Get your free Forge Playbook →