What your AI stack really costs and why you cannot name one result

Most founders spend hundreds a month on AI and the spend hides four-fifths of its real cost. Here is the audit to run before the next renewal.

Operations
By Mark Choudhari · Jun 2, 2026 · 5 min read

Add up the AI on your bank statement. Now name one result.
Made with Works

TL;DR

The Subscription Graveyard is the stack of AI tools a founder pays for every month and cannot point to a single result from. The spend usually lands between $600 and $2,000 a month, and the subscription line is only a fifth to two-fifths of the true cost once you count finding, setting up, and maintaining each tool. The audit that makes the bill visible sorts every tool by result, not by price.

In this article

How do I audit how much I am spending on AI subscriptions

Start with the part that is easy to count and end with the part that hurts. The easy part is the statement: pull three months of charges and list every AI tool the business pays for, including the ones bundled inside other software and the ones a single person expensed and forgot. Most founders are surprised by the length of the list before they are surprised by the total.

The total usually lands in the $600 to $2,000 a month range across a founder-led AI stack, and we want to be honest about that number. It is what an audit tends to surface, a range, not a single cited statistic, because every stack is built differently. Treat it as the shape of the bill, not a fixed figure. It is also rarely all visible: across organizations, SaaS license waste now ranks as the single biggest IT-spend challenge, precisely because so much of it never gets reviewed.

SaaS license waste now ranks as the single biggest IT-spend challenge, because so much of it never gets reviewed.
CFO Dive, SaaS license waste tops IT spend challenges, 2025

Then comes the part the statement does not show. A ninety-nine-dollar-a-month tool is not a ninety-nine-dollar tool. Counted across a year with the time to find it, set it up, learn it, wire it into everything else, and keep it running, the subscription ends up only a fraction of the true cost, because the implementation, the integration, and the training routinely add up to several times the license fee. So a real audit has two columns. What you pay on the invoice, and the four-fifths around it that never gets invoiced. The second column is the AI Tax, and it is the larger one.

Which AI subscriptions should I cancel

The instinct is to cancel by price, starting with the most expensive line. That is the wrong sort. Sort by result instead, and cancel by what each tool produces rather than what it costs.

Put every tool into one of three piles. The first is tools you can point to a real result from, a job that gets done because the tool exists. Keep those. The second is tools you use but cannot tie to any result, faster activity that never shows up where the business is actually measured. The third is tools nobody opens at all. The third pile cancels itself the moment you look at it, and most audits find more of it than the founder expected. AI tools in particular are bought and abandoned fast: their annual subscriber retention runs around 21%, against roughly 31% for non-AI apps, which is the graveyard forming in real time.

AI apps retain only about 21% of subscribers a year, against roughly 31% for non-AI apps. The graveyard forms in real time.
TechNewsWorld, AI apps bring in cash but struggle to retain subscribers, 2025

The second pile is the hard one, and it is where the real money sits. A large share of purchased licenses simply sit idle, which is exactly why license waste tops the spend-challenge lists; scaled to a founder spending a few hundred a month, that same idle share is real recoverable money leaving for tools that are not earning their keep. The test is simple and uncomfortable. For each tool, write down the last specific result it produced. If you cannot, that is the answer. Cancelling it is not falling behind. The tool was already not moving the business, which is the definition of behind.

Why am I paying for AI tools nobody uses

Because the graveyard fills the way every graveyard does, one at a time, each addition reasonable on the day you made it. A peer swore by a tool. A consultant set up four custom things and then the engagement ended. A trial converted to paid while you were busy. None of those felt like a mistake in the moment. They are only a graveyard in aggregate, and aggregate is exactly what nobody looks at.

It helps to see where the cost actually sits. Think of AI in the business as three levels: the tools at the bottom, the tasks they perform in the middle, and the business results at the top. The Subscription Graveyard is what happens when money pours into the bottom and nothing arrives at the top. You are paying for tools, the tools are doing tasks, and the tasks are not adding up to a result. This is the gap behind one of the most quoted patterns of the year: most businesses now run AI somewhere, and yet the great majority of company AI pilots deliver no measurable return. The spend is real, the activity is real, and the result column is empty.

There is a quieter reason too. Paying ninety-nine dollars a month feels like staying current, like insurance against falling behind. So the bill keeps renewing on a feeling rather than a result, and “I might need it” is the most expensive sentence in the stack.

How do I run the audit and stop filling the graveyard

The audit is the work, so here it is as a checklist you can run before the next renewal. Pull three months of statements and list every AI charge, including the bundled and the forgotten. Next to each, write the last specific result it produced for the business. Sort into the three piles: produces a result, used but no result, never opened. Cancel the never-opened immediately. For the no-result pile, set one renewal cycle as the deadline to tie the tool to a result or cut it. Add up what is left and ask the only question that matters, which is not “how much am I spending” but “what is all of this producing.”

The deeper fix is not a better spreadsheet of cancellations. It is to stop buying AI on faith and start buying on proof of the work, so the spend and the result live in the same place and the question is never “is this worth ninety-nine dollars” but “here is what it produced this month.”

What a system that shows you the result actually does

If the graveyard is money in the bottom level and nothing at the top, the answer has to do the opposite. It has to run the work across the tools you already pay for instead of adding another tool to the pile. It has to ship a result, not faster activity. And it has to show, per run, what the spend actually produced, so the result column stops being empty.

JynAI built Works, an AI Business OS, to do exactly that. Here is the fit, plainly.

  • The pain: the spend keeps fragmenting across more tools nobody fully uses. The mechanism: Works Across Your Stack runs across 3,000-plus apps, so the work happens inside the tools you are already paying for rather than adding another seat to the pile. The gain: the stack stops growing and the spend starts showing up in results instead of in new line items.
  • The pain: you pay for tools but cannot point to what any of them produced. The mechanism: Receipts logs every workflow run, every action, every outcome, versioned and exportable, so the result column stops being empty and the question changes from “is this worth ninety-nine dollars” to “here is what it produced this month.” The gain: the spend has something to show for itself, and the audit you have been avoiding becomes the proof you can actually use.
  • The pain: the graveyard fills because each new tool looks cheap on the day you buy it. The mechanism: Business-Aware Setup arrives already knowing the business from your LinkedIn, site, and files, so there is no blank-page integration project and no new-tool setup tax every time you start. The gain: the additions stop landing because the system you have is already doing the work.
  • The pain: 95% of generative AI pilots deliver no measurable return, and most founders cannot say theirs is different. The mechanism: Works runs work in three modes, Strategy to plan, Action to execute across connected tools, Automation to run hands-free, and every run is logged so a specific result is always attributable to a specific run. The gain: you are no longer in the majority that cannot name a result.

The affordability is the part that makes this honest rather than aspirational. The full capability set is available at the $49 tier, not behind an enterprise contract, so the founder paying two hundred a month for tools that never produce a result can reach the same ground for forty-nine dollars and get something to point to.

And we are not theorizing. Machintel ran two years of fragmented AI experiments, a graveyard of its own, paying for tools that kept the spend high and the result column empty. Six teams were running on Works in ninety days once the graveyard was replaced with a system that showed us what the spend was actually doing [VERIFY owned]. We are biased about our own product, of course. The argument under it does not need us: if the graveyard is money in the bottom level and nothing at the top, then adding another tool to the pile was never the answer. Running the work across the tools you already pay for, and logging what each run produced, is.

Stop subscribing without proof. Sign up for early access. Or size your own bill first with the AI Tax Calculator [VERIFY - calculator deliverable to_build; confirm live URL or hold this line].

Common questions

How do I audit how much I am spending on AI subscriptions?

Pull three months of bank statements and list every AI charge, including tools bundled inside other software and anything a single person expensed and forgot. Next to each, write the last specific result it produced for the business, not a faster draft, a result. Sort into three piles: produces a result, used but no result, never opened. The never-opened pile cancels itself immediately. Most founders are surprised by the length of the list before they are surprised by the total.

Which AI subscriptions should I cancel?

Sort by result, not by price. For each tool, write the last specific result it produced. If you cannot name one, that is the answer. Cancel the never-opened immediately. For the used-but-no-result pile, set one renewal cycle as the deadline to tie the tool to a result or cut it. Cancelling is not falling behind; the tool that was not producing a result was the falling behind. The stack simplifies once you cancel by result rather than by cost.

Why am I paying for AI tools nobody uses?

Because the graveyard fills one at a time, and each addition looked reasonable on the day it was made. A peer swore by a tool. A consultant set up four custom things and the engagement ended. A trial converted to paid while you were busy. AI tools retain only about 21% of subscribers a year, against roughly 31% for non-AI apps, so abandonment is built into how AI tools are used. The graveyard is not a willpower problem. It is what happens when you buy on faith rather than proof.

What is the Subscription Graveyard?

The Subscription Graveyard is the pile of AI tools a founder pays for monthly and cannot attribute a single business result to. Spend typically lands between $600 and $2,000 a month once the full stack is counted, but the subscription line captures only a fifth to two-fifths of the true cost: finding, setting up, wiring, and maintaining each tool adds several times the license fee over the year. AI tools retain only about 21 percent of subscribers annually, which is the graveyard forming in the retention data. The Phase 4 question is the moment a founder finally asks what any of it actually did.

What is the true cost of a ninety-nine dollar AI tool?

The ninety-nine dollars is the part that shows on the statement. Once you add the time to find and evaluate it, the hours to set it up and wire it to the rest of the stack, the training, and the ongoing maintenance, implementation typically adds several times the license fee over a year. A tool that costs $99 a month but cannot be tied to a single business result is not a $1,188 annual expense. It is a $2,500-plus investment that returned nothing, and the real audit is not the dollar figure, it is the empty result column next to it.

Recent Posts
AI Economic Growth
The DIY Tax and what building AI yourself really costs
Getting AI working is one cost. Keeping it working is the tax.
The Discovery Tax and what choosing AI really costs
Integration Tax
The Integration Tax and why your AI keeps breaking

What your AI stack really costs and why you cannot name one result

Most founders spend hundreds a month on AI and the spend hides four-fifths of its real cost. Here is the audit to run before the next renewal.

Operations
By Mark Choudhari · Jun 2, 2026 · 5 min read

Add up the AI on your bank statement. Now name one result.
Made with Works

TL;DR

The Subscription Graveyard is the stack of AI tools a founder pays for every month and cannot point to a single result from. The spend usually lands between $600 and $2,000 a month, and the subscription line is only a fifth to two-fifths of the true cost once you count finding, setting up, and maintaining each tool. The audit that makes the bill visible sorts every tool by result, not by price.

In this article

How do I audit how much I am spending on AI subscriptions

Start with the part that is easy to count and end with the part that hurts. The easy part is the statement: pull three months of charges and list every AI tool the business pays for, including the ones bundled inside other software and the ones a single person expensed and forgot. Most founders are surprised by the length of the list before they are surprised by the total.

The total usually lands in the $600 to $2,000 a month range across a founder-led AI stack, and we want to be honest about that number. It is what an audit tends to surface, a range, not a single cited statistic, because every stack is built differently. Treat it as the shape of the bill, not a fixed figure. It is also rarely all visible: across organizations, SaaS license waste now ranks as the single biggest IT-spend challenge, precisely because so much of it never gets reviewed.

SaaS license waste now ranks as the single biggest IT-spend challenge, because so much of it never gets reviewed.
CFO Dive, SaaS license waste tops IT spend challenges, 2025

Then comes the part the statement does not show. A ninety-nine-dollar-a-month tool is not a ninety-nine-dollar tool. Counted across a year with the time to find it, set it up, learn it, wire it into everything else, and keep it running, the subscription ends up only a fraction of the true cost, because the implementation, the integration, and the training routinely add up to several times the license fee. So a real audit has two columns. What you pay on the invoice, and the four-fifths around it that never gets invoiced. The second column is the AI Tax, and it is the larger one.

Which AI subscriptions should I cancel

The instinct is to cancel by price, starting with the most expensive line. That is the wrong sort. Sort by result instead, and cancel by what each tool produces rather than what it costs.

Put every tool into one of three piles. The first is tools you can point to a real result from, a job that gets done because the tool exists. Keep those. The second is tools you use but cannot tie to any result, faster activity that never shows up where the business is actually measured. The third is tools nobody opens at all. The third pile cancels itself the moment you look at it, and most audits find more of it than the founder expected. AI tools in particular are bought and abandoned fast: their annual subscriber retention runs around 21%, against roughly 31% for non-AI apps, which is the graveyard forming in real time.

AI apps retain only about 21% of subscribers a year, against roughly 31% for non-AI apps. The graveyard forms in real time.
TechNewsWorld, AI apps bring in cash but struggle to retain subscribers, 2025

The second pile is the hard one, and it is where the real money sits. A large share of purchased licenses simply sit idle, which is exactly why license waste tops the spend-challenge lists; scaled to a founder spending a few hundred a month, that same idle share is real recoverable money leaving for tools that are not earning their keep. The test is simple and uncomfortable. For each tool, write down the last specific result it produced. If you cannot, that is the answer. Cancelling it is not falling behind. The tool was already not moving the business, which is the definition of behind.

Why am I paying for AI tools nobody uses

Because the graveyard fills the way every graveyard does, one at a time, each addition reasonable on the day you made it. A peer swore by a tool. A consultant set up four custom things and then the engagement ended. A trial converted to paid while you were busy. None of those felt like a mistake in the moment. They are only a graveyard in aggregate, and aggregate is exactly what nobody looks at.

It helps to see where the cost actually sits. Think of AI in the business as three levels: the tools at the bottom, the tasks they perform in the middle, and the business results at the top. The Subscription Graveyard is what happens when money pours into the bottom and nothing arrives at the top. You are paying for tools, the tools are doing tasks, and the tasks are not adding up to a result. This is the gap behind one of the most quoted patterns of the year: most businesses now run AI somewhere, and yet the great majority of company AI pilots deliver no measurable return. The spend is real, the activity is real, and the result column is empty.

There is a quieter reason too. Paying ninety-nine dollars a month feels like staying current, like insurance against falling behind. So the bill keeps renewing on a feeling rather than a result, and “I might need it” is the most expensive sentence in the stack.

How do I run the audit and stop filling the graveyard

The audit is the work, so here it is as a checklist you can run before the next renewal. Pull three months of statements and list every AI charge, including the bundled and the forgotten. Next to each, write the last specific result it produced for the business. Sort into the three piles: produces a result, used but no result, never opened. Cancel the never-opened immediately. For the no-result pile, set one renewal cycle as the deadline to tie the tool to a result or cut it. Add up what is left and ask the only question that matters, which is not “how much am I spending” but “what is all of this producing.”

The deeper fix is not a better spreadsheet of cancellations. It is to stop buying AI on faith and start buying on proof of the work, so the spend and the result live in the same place and the question is never “is this worth ninety-nine dollars” but “here is what it produced this month.”

What a system that shows you the result actually does

If the graveyard is money in the bottom level and nothing at the top, the answer has to do the opposite. It has to run the work across the tools you already pay for instead of adding another tool to the pile. It has to ship a result, not faster activity. And it has to show, per run, what the spend actually produced, so the result column stops being empty.

JynAI built Works, an AI Business OS, to do exactly that. Here is the fit, plainly.

  • The pain: the spend keeps fragmenting across more tools nobody fully uses. The mechanism: Works Across Your Stack runs across 3,000-plus apps, so the work happens inside the tools you are already paying for rather than adding another seat to the pile. The gain: the stack stops growing and the spend starts showing up in results instead of in new line items.
  • The pain: you pay for tools but cannot point to what any of them produced. The mechanism: Receipts logs every workflow run, every action, every outcome, versioned and exportable, so the result column stops being empty and the question changes from “is this worth ninety-nine dollars” to “here is what it produced this month.” The gain: the spend has something to show for itself, and the audit you have been avoiding becomes the proof you can actually use.
  • The pain: the graveyard fills because each new tool looks cheap on the day you buy it. The mechanism: Business-Aware Setup arrives already knowing the business from your LinkedIn, site, and files, so there is no blank-page integration project and no new-tool setup tax every time you start. The gain: the additions stop landing because the system you have is already doing the work.
  • The pain: 95% of generative AI pilots deliver no measurable return, and most founders cannot say theirs is different. The mechanism: Works runs work in three modes, Strategy to plan, Action to execute across connected tools, Automation to run hands-free, and every run is logged so a specific result is always attributable to a specific run. The gain: you are no longer in the majority that cannot name a result.

The affordability is the part that makes this honest rather than aspirational. The full capability set is available at the $49 tier, not behind an enterprise contract, so the founder paying two hundred a month for tools that never produce a result can reach the same ground for forty-nine dollars and get something to point to.

And we are not theorizing. Machintel ran two years of fragmented AI experiments, a graveyard of its own, paying for tools that kept the spend high and the result column empty. Six teams were running on Works in ninety days once the graveyard was replaced with a system that showed us what the spend was actually doing [VERIFY owned]. We are biased about our own product, of course. The argument under it does not need us: if the graveyard is money in the bottom level and nothing at the top, then adding another tool to the pile was never the answer. Running the work across the tools you already pay for, and logging what each run produced, is.

Stop subscribing without proof. Sign up for early access. Or size your own bill first with the AI Tax Calculator [VERIFY - calculator deliverable to_build; confirm live URL or hold this line].

Common questions

How do I audit how much I am spending on AI subscriptions?

Pull three months of bank statements and list every AI charge, including tools bundled inside other software and anything a single person expensed and forgot. Next to each, write the last specific result it produced for the business, not a faster draft, a result. Sort into three piles: produces a result, used but no result, never opened. The never-opened pile cancels itself immediately. Most founders are surprised by the length of the list before they are surprised by the total.

Which AI subscriptions should I cancel?

Sort by result, not by price. For each tool, write the last specific result it produced. If you cannot name one, that is the answer. Cancel the never-opened immediately. For the used-but-no-result pile, set one renewal cycle as the deadline to tie the tool to a result or cut it. Cancelling is not falling behind; the tool that was not producing a result was the falling behind. The stack simplifies once you cancel by result rather than by cost.

Why am I paying for AI tools nobody uses?

Because the graveyard fills one at a time, and each addition looked reasonable on the day it was made. A peer swore by a tool. A consultant set up four custom things and the engagement ended. A trial converted to paid while you were busy. AI tools retain only about 21% of subscribers a year, against roughly 31% for non-AI apps, so abandonment is built into how AI tools are used. The graveyard is not a willpower problem. It is what happens when you buy on faith rather than proof.

What is the Subscription Graveyard?

The Subscription Graveyard is the pile of AI tools a founder pays for monthly and cannot attribute a single business result to. Spend typically lands between $600 and $2,000 a month once the full stack is counted, but the subscription line captures only a fifth to two-fifths of the true cost: finding, setting up, wiring, and maintaining each tool adds several times the license fee over the year. AI tools retain only about 21 percent of subscribers annually, which is the graveyard forming in the retention data. The Phase 4 question is the moment a founder finally asks what any of it actually did.

What is the true cost of a ninety-nine dollar AI tool?

The ninety-nine dollars is the part that shows on the statement. Once you add the time to find and evaluate it, the hours to set it up and wire it to the rest of the stack, the training, and the ongoing maintenance, implementation typically adds several times the license fee over a year. A tool that costs $99 a month but cannot be tied to a single business result is not a $1,188 annual expense. It is a $2,500-plus investment that returned nothing, and the real audit is not the dollar figure, it is the empty result column next to it.