Before You Spend Another Dollar on AI: The One-Workflow Rule

If you spent $25,000 on AI for your revenue team this year, which line in your P&L would prove it? Most owners cannot answer that question, and the published research says most never will. Here is a decision framework you can run before the money moves: one workflow, a proven leak, a baseline, one metric, one owner.

Every B2B owner I talk with this summer is somewhere in the same conversation. The board, the bank, the spouse, or the mirror is asking what the AI spending produced. The tools got bought in 2024 and 2025. The budget reviews are happening now. And the honest answer, in most companies, is a shrug backed by a usage dashboard.

That is a solvable problem, and solving it starts before the next purchase, at the moment you decide which workflow gets funded first. This article gives you the selection and measurement framework I use with clients: the One-Workflow Rule. You can run it yourself, this week, without hiring anyone.

The 2026 receipts: buying AI is common, proving it is rare

The research on this has hardened over the past year, and the numbers come from serious institutions rather than vendor decks.

MIT's finding is the headline, but the surrounding data tells the fuller story. S&P Global Market Intelligence found that 42 percent of companies abandoned most of their AI initiatives in 2025, more than double the 17 percent rate a year earlier, and that the average organization scrapped 46 percent of its AI proofs of concept before production. Deloitte's 2026 State of AI in the Enterprise survey of 3,000 leaders found that 74 percent of organizations want AI to grow revenue, while 20 percent have actually seen it happen. Gartner forecasts that more than 40 percent of agentic AI projects will be cancelled by 2027 without clear governance and ROI frameworks.

One more MIT finding deserves a budget owner's full attention. More than half of generative AI budgets flow to sales and marketing tools, yet the measurable returns in MIT's dataset concentrated in operational work with clean before-and-after numbers. Read that carefully. Revenue AI can absolutely pay. Revenue AI is simply the category where companies most often skip the measurement discipline that would prove it, because activity in a sales pipeline is easy to generate and hard to attribute. The spend goes where the excitement is. The proof lives where the discipline is.

None of this is an argument against AI in the revenue function. I install AI-supported revenue systems for a living, and the well-run ones pay for themselves. It is an argument against the way most first AI dollars get spent: tool first, workflow never, measurement optional. We covered why the tools themselves are rarely the bottleneck in AI Sales Tools Not Working? The Real Fix for B2B Teams. This article covers the decision that comes before any tool enters the building.

The behavior to stop: tool-first spending

The pattern shows up in nearly every stalled deployment I audit. A demo impresses someone. A subscription starts. The team is told to "use the AI." Six months later the renewal invoice arrives and nobody can say what changed, because nobody wrote down what was supposed to change, or what the numbers looked like before.

Tool-first spending feels like progress because something visible happened: licenses provisioned, training scheduled, activity generated. McKinsey's State of AI research found the inverse pattern among the companies reporting significant financial returns: they were about twice as likely to have redesigned the workflow before selecting any tool. The workflow decision comes first. The tool serves the workflow. Most buyers run that sequence backwards, and the failure statistics above are the bill.

The One-Workflow Rule

The rule itself fits in one sentence. Fund exactly one AI workflow inside the revenue function at a time, and only after it clears four gates in writing.

The rule carries one clause that makes it enforceable: the scale-or-stop date is written before kickoff. Ninety days is the right horizon for a first revenue workflow. On that date, the owner presents the baseline next to the current numbers and leadership makes one of two calls. Scale it, or stop it. A stopped workflow after one honest quarter is a cheap education. An unmeasured workflow running for two years is where the real money dies.

What the leak math looks like in practice

Here is the Gate 1 and Gate 2 arithmetic on the most common first workflow, using deliberately conservative illustrative numbers. Run it with your own.

A $10M industrial services firm gets 40 inbound inquiries a month. The CRM shows median time-to-first-touch is 26 hours, and 9 of the 40 never get a documented follow-up at all. Say your close rate on worked inbound is 15 percent and your average first order is $18,000. The 9 unworked leads alone represent about 1.35 expected wins a month, roughly $24,000 in expected first-order revenue, before counting what slow response does to the other 31. Annualized, the follow-up leak is priced in the hundreds of thousands, from your own records, in an afternoon of CRM queries.

Now the AI decision is no longer "should we buy an AI tool." It is "we have a six-figure follow-up leak, the baseline is 26 hours and 9 dropped leads a month, the metric is time-to-first-touch and percent of leads worked, and we will fund one governed workflow against it with a review on October 15." That sentence survives any budget meeting ever held. If you want help finding and pricing your leaks before choosing, the free Recursive Revenue Loop Builder scores the four feedback loops in your revenue system in about 12 minutes and names the weakest one, and we walked the five-point version of this diagnostic, leak by leak, in Pipeline Full, Revenue Flat?

The five workflows the rule usually selects

The right first workflow is the one attached to your largest provable leak, so the answer differs company to company. Across owner-led manufacturers, industrial services firms, and B2B teams in the $2M to $25M range, five candidates win the selection most often, because each one has a priceable leak, a clean baseline, and a metric that moves inside 90 days.

Notice what the list excludes: AI-written outreach at volume. Buyer-facing message generation is where trust goes to die when it runs without review, and it fails Gate 3 anyway, because attribution on outbound sequences is the murkiest measurement in the revenue function. If your diagnosis says the real problem sits upstream of any single workflow, in qualification, process, or team discipline, start with the questions in Revenue Root-Cause Self-Diagnosis before funding anything.

Where we stand

Our position at Get 'er Done is that AI earns a place in the revenue function one governed workflow at a time, with a human above the loop owning every decision that touches a buyer, and with proof measured per client, per record, in the client's own system. That last part is a design choice, and it comes from watching pooled vendor benchmarks evaporate under one pointed question from a CFO. A before-and-after comparison from your own CRM answers the question before it gets asked. The broader case for that operating posture is in From Human-in-the-Loop to Human-Above-the-Loop.

The compounding effect is real, too. The first workflow that clears its 90-day review does more than pay for itself. It leaves behind clean records, a written process, a trained owner, and a measurement habit, which is exactly the foundation the second workflow needs. Companies that run the rule twice are most of the way to a revenue system that improves on a loop instead of a pile of disconnected tools.

Run the rule this week: seven steps