The Bill Jumps the Month It Starts Working
Every AI tool has two prices. The one on the page, and the one you actually pay once you can't rip it out.
Every AI tool you'll ever buy has two prices. The one on the pricing page, and the one you actually pay once it works. The gap between them isn't an accident. It's the plan.
Here's how it goes. You sign up on the starter tier because the starter tier is cheap and the demo was good. You wire it into your business. It takes a few weeks, maybe a month, to get it doing something real: the follow-up that goes out on its own, the records that sync, the intake form that files itself. And it works. You stop thinking about it. That's the dangerous part, because "you stop thinking about it" is precisely when a business becomes dependent on a thing.
Then month four arrives, and you hit a wall you never saw on the pricing page. A record cap. A row limit. A monthly task ceiling. A contact tier. The email says you've hit your plan's limit and your automation has stopped. Or worse, it's still running but silently dropping everything over the line. And the button to fix it doesn't take you to a plan that costs a little more. It takes you to one that costs five times more.
The tool is cheapest when it's useless and most expensive the month it finally pays for itself.
I build these systems for small businesses, so let me tell you how this feels from the inside. I put a client on a platform's entry plan once. Fine for a shop their size, the math worked, everyone was happy. That platform's free and cheap tiers cap you at a thousand records. A thousand sounds like plenty until you notice that every job, every customer, every automated log entry eats a record, and a working business generates records the way a running engine burns fuel. You don't cross a thousand because you got greedy. You cross it because the thing worked and the business grew, which was the entire point.
That's the tell. The cap isn't sized to stop abuse. It's sized to catch success. It sits just past the point where a real operation starts leaning on the tool, so the exact moment you can't rip it out without pain is the exact moment the invoice jumps.
Nobody in the sales process mentions this, and technically they aren't lying. It's all on the pricing page if you scroll to the comparison grid and read the faint gray row about "records" or "tasks" or "seats included." But you're not reading that when you sign up. You're reading the top of the page, where it shows the number you want to see. The real number is parked, deliberately, in the row you skim.
And the jump between tiers is never linear. Watch this pattern across almost every SaaS and AI tool on the market: the plans go $20, then $99, then "contact sales." That's not three plans. That's a cheap one to get you in, a painful one to catch you when you grow, and a wall where the pricing stops being published at all, because now they're negotiating against your dependency. The second-tier price isn't set by what the software costs to run. It's set by what it costs you to leave.
Here's what makes it nastier than a straight price hike. When a usage bill goes up, at least you know why: you used more. The tier cliff doesn't work like that. You didn't do anything different in month four than you did in month three. You just crossed an invisible line, and the price of standing still tripled. You're paying more for the same thing, and the reason is that you can no longer afford to stop.
The measurement problem hides inside this one too. Most owners have no idea how many records, rows, or tasks their operation will generate in a year, because they've never had to count. So when a vendor asks "how many contacts do you have," the honest answer is "I don't know," and the honest answer means you can't price the tool you're about to depend on. You find out the real number the day it bills you for it.
So here's the move, and it costs you five minutes before you sign anything. Don't ask what the plan costs. Ask what the plan above it costs, and find the exact limit that forces the jump. Then ask yourself a blunt question: if this tool works as well as they're promising, will I blow past that limit? If the answer is yes, then the price you should be comparing is the next tier up, because that's the one you'll actually be on by the time it matters. The starter price is a number for a version of your business that isn't using the tool. Price the version that is.
And ask it out loud, to the salesperson, in these words: "What's the first limit I'll hit if this works, and what does it cost to cross it?" Then watch. An honest vendor has the number ready. A vendor selling you a cliff gets vague, steers you back to the entry price, or tells you not to worry about that now. The vagueness is the answer. The whole model depends on you not asking until you can't leave.
I've moved clients off a tool for exactly this reason before we ever went live, once I ran the math and saw the month-four cliff coming. The tool wasn't bad. It was priced to be cheap right up until it was load-bearing, and load-bearing is the only state that matters. A tool you don't depend on is a toy. A tool you depend on is leverage the vendor holds over you. Know which one you're buying, and know what it costs at the tier where it becomes the second one.
The cheap tier isn't the deal. It's the bait.