The Two-Number Test for Every AI Decision

A software rep sends you a demo. The tool looks sharp. It has AI in the name, a clean dashboard, and three case studies from companies that sound a little bigger than yours. Everyone on the call nods. Someone says it feels like the future.

Then the real question never gets asked: which of your two numbers does this actually move, and by how much?

I ask that question for a living now, and I asked a version of it for years building software inside regulated businesses where a wrong call was expensive. The framing that survives every context is almost boring in how simple it is. A business has exactly two numbers it can change. Everything you will ever be sold is a tactic sitting under one of them.

There are only two numbers

Strip a business down to first principles and profit is one line of arithmetic:

Profit = Revenue − Expenses.

That is the whole machine. Money comes in, money goes out, and the gap is what you keep. It holds no matter your industry, your size, or how the pitch deck dresses it up. A plumbing company, a parts distributor, and a 12-person agency all run on the same equation. So when you are deciding whether to buy a tool, build a system, or automate a task, there are only two honest ways it can help you: it grows what comes in, or it shrinks what goes out.

I call this the two-number test. Before the features, before the demo, before the annual contract, you make the thing declare which number it moves. If it can’t, that is your answer.

Every tool you will ever be sold is a tactic under one of two numbers. Make it tell you which one before you make it a line item.

Most AI buying skips this step entirely. The tool gets bought because it is impressive, not because anyone traced it to revenue or expense. That gap between impressive and worth paying for is where most SMB technology budgets quietly leak.

Number one: revenue, and the leaks inside it

Revenue itself breaks down cleanly. It is price times volume, where volume is really your number of customers, times how often they buy, times how much they buy each time. Any tool that claims to grow revenue is pulling one of those levers:

• More customers. Marketing, outreach, referrals, new channels. A pure top-of-funnel play.
• More from the customers you already have. Upsells, recurring plans, higher purchase frequency. This is usually the cheapest revenue an SMB can get, because you already paid to acquire the relationship.
• Higher prices. Most owners underprice out of fear. If the value is real, a modest increase flows almost entirely to profit.

There is a fourth lever that gets missed because it doesn’t feel like growth: plugging the leaks. This is revenue you already earned that disappears before it reaches you. Churned customers who would have stayed with one follow-up. Invoices that never got collected. Quotes that were never sent. Leads that came in and sat in an inbox until they went cold. The customer wanted to pay you. The money was yours. A gap in the process ate it.

I bring this up first because it is the most overlooked and the fastest to fix. Chasing new customers is expensive and slow. Stopping the revenue that is already walking out the door is close to free, and it is exactly the kind of quiet, repetitive process work that automation is genuinely good at.

Number two: expenses, and the efficiency that hides inside them

The other number is expenses: your fixed costs, plus your variable costs times volume. You lower it two ways. You can cut the absolute number, or, more powerfully, you can lower the cost per unit of output so you can grow revenue without growing costs at the same pace.

That second one is the lever that matters most for a growing business, and it is where AI and automation actually earn their keep. The levers under expenses:

• Trim variable costs. Better supplier terms, less waste, fewer labor hours per job.
• Control fixed overhead. Kill the software seats nobody uses. Stop paying for tools that overlap.
• Do more with the same people. Automate the repetitive work so your existing team produces more without you adding headcount.
• Prune what loses money. Drop the low-margin customer, the loss-leading service, the offering that quietly drags the whole number down.

That third lever is the quiet giant. When a skilled, well-paid person spends two hours a day copying data between systems, chasing a status update, or rebuilding the same spreadsheet, you are paying senior wages for work that produces nothing new. Give those hours back and the same team handles more volume. That is capacity you didn’t have to hire for, which means it lifts revenue and holds down cost at the same time.

Why the two-number test protects you from shelfware

Here is the part the vendor won’t say out loud. A tool can be genuinely impressive and still move neither of your numbers. Most abandoned software wasn’t bad. It just never traced to a number anyone was accountable for, so no one noticed when it quietly did nothing.

AI has made this worse, not better. It is now easy to generate software and hard to generate software that fits a real business. The demos are more dazzling than ever, which makes the discipline of asking “which number, and by how much” more valuable than ever. The test is your defense against paying for the future and receiving a login you use twice.

Run any proposed tool or project through three questions:

1. Which number does this move, revenue or expense? If the honest answer is “it’s strategic” or “it’s the future,” you don’t have a case yet. You have a feeling.
2. Through which lever specifically? More customers, more from existing ones, higher price, plugged leak, or lower cost per unit. Name it. Vague benefits are where budgets go to die.
3. What is the rough size, and what does it cost all-in? Not to the decimal. Just enough to know whether you are chasing a real number or a rounding error. Count the full cost too, including the hours to set it up, the time to adopt it, and the ongoing babysitting, not just the sticker price.

If a tool can answer all three, it earns a real look. If it stumbles on the first, you just saved yourself a subscription and the slow guilt of a tool nobody opens.

The magic is where the two numbers overlap

The two numbers aren’t fully separate, and the most powerful moves live in the overlap.

Margin is the overlap in the small. Raise your price while lowering your cost per unit and you improve the same sale from both sides at once.

Operating leverage is the overlap that compounds. Once you have built efficient systems, each new dollar of revenue costs you far less to serve than the first dollars did. This is why efficiency work feels like a sibling to growth rather than an opposite. Automating a repetitive workflow lowers a cost today and raises your capacity to grow tomorrow. It shows up in both numbers.

One honest caveat so this stays a framework and not a sales pitch. Cash-flow timing, getting paid faster or paying suppliers slower, feels like it belongs on this list. It doesn’t create profit. It just makes the profit you already have more usable. Useful, real, worth doing. Not a third root number. Keep it in its lane so it doesn’t distract you from the two that actually move the machine.

What to do Monday

The framework is only worth anything if it changes what you do next. Here is the honest starting order for most SMBs, cheapest and fastest first:

1. Find and fix the leaks. Where is earned revenue disappearing before it reaches you? Follow-ups that don’t happen, invoices that don’t get collected, quotes that never go out. This is usually the highest return and the lowest cost.
2. Sell more to the people who already trust you. It is far cheaper than finding new ones.
3. Ruthlessly improve one repetitive process. Pick the task that eats the most skilled-person hours for the least new output, and systematize or automate it. This one compounds forever.

Then do the thing almost nobody does: measure the two numbers every month. Not a dashboard with forty metrics. Two numbers, and the gap between them. Watch which of your moves actually shifted them. You will learn more from that one habit than from any strategy book, because you will be running your business on the equation instead of on the pitch.

The next time something impressive lands on your desk with AI in the name, you already know the only question that matters. Which number does it move, and by how much? A tool that can’t answer that isn’t the future. It is a line item waiting to be cancelled.

Finding where a business is leaking revenue or paying skilled people to do repetitive work, then building the system that fixes it, is most of what I do at Joint Lab Solutions. The two-number test is where every one of those conversations starts.