The Discovery Tax and what choosing AI really costs

The hours you lose deciding which AI to use never show up on a subscription line, and they come out of the one thing you cannot buy back.

Technology
By Mark Choudhari · Mar 26, 2026 · 6 min read

AI Economic Growth
Made with Works

TL;DR

The Discovery Tax is the time and attention a founder spends deciding which AI to use, before any work gets done. With dozens of new tools launching daily and the average business already running several, the real cost is not the subscriptions. It is the evaluating, comparing, and re-deciding that never ends and comes out of a founder’s scarcest resource, attention.


In this article

Thirty new AI tools were listed today. There will be thirty more tomorrow. Somewhere in that stream is the one that might finally move your business, and somewhere in your week is the time you will spend trying to find it. That time is the first hidden cost of AI, and almost no one puts it on the books. It is the first of seven taxes in the AI Tax, the real price of AI for a founder-led business once you count everything past the subscription line. This piece is about that first tax, because it is the one you are paying right now, whether or not you have noticed.

Why do I keep evaluating new AI tools every few weeks?

Because the field is built to make you. New tools arrive faster than any one person can read, so the shortlist is never finished and the deciding never closes. The open-source hub Hugging Face alone now hosts more than two million models, with thousands added every week. The typical business already runs a handful of AI tools side by side. Adoption is no longer the question. Choosing is.

The cycle is familiar. A tool launches. A peer mentions it on a call. A post promises it changed their entire workflow. Now it is on your list. You read, you compare, you start a trial, you give it twenty minutes, and you move on, a little more tired and no closer to a decision. None of those twenty-minute sessions feels expensive. Added up across a quarter, they are one of the most expensive things you do.

More than two million models now live on a single open-source hub, with thousands added every week.
Hugging Face, State of Open Source AI, Spring 2026

This is not a discipline problem. It is the design of the moment you are operating in. Staying current with AI has quietly become a standing job, and you are the one doing it.

How do I choose the right AI tool when there are too many options?

You stop trying to. Choosing more carefully, by trialing more tools, makes the problem worse, not better, because each new tool adds load without adding a result. A Boston Consulting Group study of nearly 1,500 workers found productivity rose when people used three or fewer AI tools, then fell once they used four or more. The founder doing the most evaluation, keeping the most current, running the most trials, is often the one getting the least out of AI.

Productivity rose when people used three or fewer AI tools, then dropped once they used four or more.
Fortune, BCG study, 2026

So “pick the best tool” is the wrong target. There is no best tool when the field reshuffles every Monday. The better question is not which of these thirty options wins. It is whether you should be the one holding the shortlist at all.

It helps to see where this cost sits. Think of AI in your business as three layers: the tools at the bottom, the tasks they do in the middle, and the business results at the top. The Discovery Tax lives at the very bottom, in the churn of swapping and re-evaluating tools, and that churn never rises to the top as a result. You can spend your whole week down there and have nothing at the outcome layer to show for it. The fuller version is in the Three-Layer Pyramid.

How do I know if an AI tool will actually help my business?

Most of the time, you cannot know in a twenty-minute trial, and that is the trap. A demo shows you a capability. It does not show you whether that capability survives contact with your data, your team, and your actual process. So you are left guessing, and guessing pushes you back into more evaluation, which is the tax again.

The honest answer is that a tool helps your business when it does a job your business needs, end to end, against your real context, not when it looks impressive in isolation. That is a much higher bar than “is this tool good,” and it is a bar almost no standalone tool clears on its own, because the tool does not know your business. It only knows its own feature.

This is the gap behind one of the most quoted numbers of the year. While the large majority of companies are using AI somewhere, an MIT study of 300 initiatives found roughly 95% report no measurable return on it. That is not because the tools are bad. It is because choosing and wiring up tools is not the same as getting a result, and most of the effort is still stuck at the choosing.

How much time am I losing just deciding which AI to use

We are deliberately not going to hand you a precise hours-per-week figure, because an honest one does not exist and we will not invent it. What the evidence does show is the shape of the cost: a daily firehose of new tools, a stack of several already in use, and a measurable penalty for adding more. The deciding is continuous, it competes with the work that actually grows the business, and it produces fatigue, not output. Heavy AI users report markedly higher burnout, and a lot of that is the cognitive load of running and second-guessing the tools rather than the work itself.

Put plainly: the Discovery Tax is not a line item, which is exactly why it is so easy to keep paying. It comes out of the scarcest thing a founder has, which is attention, and it comes out every week.

There is a fair objection here, and it deserves a straight answer. Evaluating tools is normal due diligence, and one round of it is healthy. The Discovery Tax is not that round. It is the recurring re-deciding every few weeks as the landscape shifts, the diligence that never gets to finish because the field will not hold still. Keep the one-time diligence. The repeat is the part that is optional.

What a system that does the deciding for you looks like

The way out is not a sharper shortlist or a better comparison spreadsheet. It is to stop being the person who has to keep one. Any real answer has to clear a clear bar: it has to know your business well enough to pick the right approach for a job, run that approach against your real context rather than a demo, and absorb the next thirty tools so you never have to rank them. The question you ask changes from “which of these thirty tools is right” to “what do I want done.”

JynAI built Works, the AI Business OS for founders and their teams, to clear exactly that bar. Here is the fit, plainly.

  • Pain You are the one holding the shortlist, re-deciding every week.
    Keeps Getting Better keeps 100+ models in a pool and auto-selects the right one per step, and new models, connectors, and workflows land inside the setup you already have.
    Gain The deciding happens underneath you, not on your calendar.

  • Pain A demo never tells you whether a tool fits your business.
    Business-Aware Setup reads your LinkedIn, site, and files into a workspace Works already understands, so the choice is made against your real context, not a blank trial.
    Gain The fit question is answered before you would have started a trial.

  • Pain Every new tool adds load without adding a result.
    Expert-Grade Workflows ship 500+ plays built on EOS, MEDDIC, ABM, and PLG, so the work starts from a proven approach instead of a tool you are still evaluating.
    Gain You move at the result layer, not the tool layer.

The price proof is the part that makes this honest for a founder-led business: the full capability set unlocks at the $49 Pro tier, not a six-figure engagement. And the first-party version is plain. Machintel spent close to two years on fragmented AI experiments before building the thing that ended the deciding, and reached six teams in 90 days where two years of choosing had not. The number that mattered was not how many tools the team tried. It was 90 days against two years.

If you only take one thing from this: the tools will keep coming, every day, forever. The tax on choosing them is the optional part.

Stop the FOMO. Stop evaluating. Sign up for early access. Or estimate your own AI Tax first with the AI Tax Calculator.

Common Questions

Why do I keep evaluating new AI tools every few weeks?

Because new tools arrive faster than anyone can read, so the shortlist never closes. The cure is not more discipline; it is to stop being the person who has to keep the shortlist. The full taxonomy of where this cost sits is in the seven AI taxes, added up.

How do I choose the right AI tool when there are too many options?

You change the question. Adding more tools past three or four reduces the gains, so “pick the best tool” is the wrong target in a field that reshuffles weekly. The better move is to let a system that knows your business pick the approach for each job, which is the shift covered in the Three-Layer Pyramid.

How do I know if an AI tool will actually help my business?

A twenty-minute trial shows you a capability, not whether it survives your data, your team, and your real process. A tool helps when it finishes a job your business needs end to end against your context. That is a far higher bar than “is this tool good,” and most standalone tools cannot clear it alone.

Is evaluating AI tools just normal due diligence?

One round is. The Discovery Tax is the recurring re-evaluation every few weeks as the landscape shifts, the diligence that never finishes. Keep the one-time look; the repeat is the cost you can stop paying.

Why do most companies see no measurable return on AI?

Because choosing and wiring up tools is not the same as getting a result. An MIT study of 300 initiatives found roughly 95% reported no measurable return, and most of that effort is still stuck at the choosing and connecting, not at the outcome. The recurring re-wiring that follows is its own cost, covered in why your AI keeps breaking.

Recent Posts
AI Economic Growth
The DIY Tax and what building AI yourself really costs
Integration Tax
The Integration Tax and why your AI keeps breaking
AI Economic Growth
Why your team never reaches the AI payoff

The Discovery Tax and what choosing AI really costs

The hours you lose deciding which AI to use never show up on a subscription line, and they come out of the one thing you cannot buy back.

Technology
By Mark Choudhari · Mar 26, 2026 · 6 min read

AI Economic Growth
Made with Works

TL;DR

The Discovery Tax is the time and attention a founder spends deciding which AI to use, before any work gets done. With dozens of new tools launching daily and the average business already running several, the real cost is not the subscriptions. It is the evaluating, comparing, and re-deciding that never ends and comes out of a founder’s scarcest resource, attention.


In this article

Thirty new AI tools were listed today. There will be thirty more tomorrow. Somewhere in that stream is the one that might finally move your business, and somewhere in your week is the time you will spend trying to find it. That time is the first hidden cost of AI, and almost no one puts it on the books. It is the first of seven taxes in the AI Tax, the real price of AI for a founder-led business once you count everything past the subscription line. This piece is about that first tax, because it is the one you are paying right now, whether or not you have noticed.

Why do I keep evaluating new AI tools every few weeks?

Because the field is built to make you. New tools arrive faster than any one person can read, so the shortlist is never finished and the deciding never closes. The open-source hub Hugging Face alone now hosts more than two million models, with thousands added every week. The typical business already runs a handful of AI tools side by side. Adoption is no longer the question. Choosing is.

The cycle is familiar. A tool launches. A peer mentions it on a call. A post promises it changed their entire workflow. Now it is on your list. You read, you compare, you start a trial, you give it twenty minutes, and you move on, a little more tired and no closer to a decision. None of those twenty-minute sessions feels expensive. Added up across a quarter, they are one of the most expensive things you do.

More than two million models now live on a single open-source hub, with thousands added every week.
Hugging Face, State of Open Source AI, Spring 2026

This is not a discipline problem. It is the design of the moment you are operating in. Staying current with AI has quietly become a standing job, and you are the one doing it.

How do I choose the right AI tool when there are too many options?

You stop trying to. Choosing more carefully, by trialing more tools, makes the problem worse, not better, because each new tool adds load without adding a result. A Boston Consulting Group study of nearly 1,500 workers found productivity rose when people used three or fewer AI tools, then fell once they used four or more. The founder doing the most evaluation, keeping the most current, running the most trials, is often the one getting the least out of AI.

Productivity rose when people used three or fewer AI tools, then dropped once they used four or more.
Fortune, BCG study, 2026

So “pick the best tool” is the wrong target. There is no best tool when the field reshuffles every Monday. The better question is not which of these thirty options wins. It is whether you should be the one holding the shortlist at all.

It helps to see where this cost sits. Think of AI in your business as three layers: the tools at the bottom, the tasks they do in the middle, and the business results at the top. The Discovery Tax lives at the very bottom, in the churn of swapping and re-evaluating tools, and that churn never rises to the top as a result. You can spend your whole week down there and have nothing at the outcome layer to show for it. The fuller version is in the Three-Layer Pyramid.

How do I know if an AI tool will actually help my business?

Most of the time, you cannot know in a twenty-minute trial, and that is the trap. A demo shows you a capability. It does not show you whether that capability survives contact with your data, your team, and your actual process. So you are left guessing, and guessing pushes you back into more evaluation, which is the tax again.

The honest answer is that a tool helps your business when it does a job your business needs, end to end, against your real context, not when it looks impressive in isolation. That is a much higher bar than “is this tool good,” and it is a bar almost no standalone tool clears on its own, because the tool does not know your business. It only knows its own feature.

This is the gap behind one of the most quoted numbers of the year. While the large majority of companies are using AI somewhere, an MIT study of 300 initiatives found roughly 95% report no measurable return on it. That is not because the tools are bad. It is because choosing and wiring up tools is not the same as getting a result, and most of the effort is still stuck at the choosing.

How much time am I losing just deciding which AI to use

We are deliberately not going to hand you a precise hours-per-week figure, because an honest one does not exist and we will not invent it. What the evidence does show is the shape of the cost: a daily firehose of new tools, a stack of several already in use, and a measurable penalty for adding more. The deciding is continuous, it competes with the work that actually grows the business, and it produces fatigue, not output. Heavy AI users report markedly higher burnout, and a lot of that is the cognitive load of running and second-guessing the tools rather than the work itself.

Put plainly: the Discovery Tax is not a line item, which is exactly why it is so easy to keep paying. It comes out of the scarcest thing a founder has, which is attention, and it comes out every week.

There is a fair objection here, and it deserves a straight answer. Evaluating tools is normal due diligence, and one round of it is healthy. The Discovery Tax is not that round. It is the recurring re-deciding every few weeks as the landscape shifts, the diligence that never gets to finish because the field will not hold still. Keep the one-time diligence. The repeat is the part that is optional.

What a system that does the deciding for you looks like

The way out is not a sharper shortlist or a better comparison spreadsheet. It is to stop being the person who has to keep one. Any real answer has to clear a clear bar: it has to know your business well enough to pick the right approach for a job, run that approach against your real context rather than a demo, and absorb the next thirty tools so you never have to rank them. The question you ask changes from “which of these thirty tools is right” to “what do I want done.”

JynAI built Works, the AI Business OS for founders and their teams, to clear exactly that bar. Here is the fit, plainly.

  • Pain You are the one holding the shortlist, re-deciding every week.
    Keeps Getting Better keeps 100+ models in a pool and auto-selects the right one per step, and new models, connectors, and workflows land inside the setup you already have.
    Gain The deciding happens underneath you, not on your calendar.

  • Pain A demo never tells you whether a tool fits your business.
    Business-Aware Setup reads your LinkedIn, site, and files into a workspace Works already understands, so the choice is made against your real context, not a blank trial.
    Gain The fit question is answered before you would have started a trial.

  • Pain Every new tool adds load without adding a result.
    Expert-Grade Workflows ship 500+ plays built on EOS, MEDDIC, ABM, and PLG, so the work starts from a proven approach instead of a tool you are still evaluating.
    Gain You move at the result layer, not the tool layer.

The price proof is the part that makes this honest for a founder-led business: the full capability set unlocks at the $49 Pro tier, not a six-figure engagement. And the first-party version is plain. Machintel spent close to two years on fragmented AI experiments before building the thing that ended the deciding, and reached six teams in 90 days where two years of choosing had not. The number that mattered was not how many tools the team tried. It was 90 days against two years.

If you only take one thing from this: the tools will keep coming, every day, forever. The tax on choosing them is the optional part.

Stop the FOMO. Stop evaluating. Sign up for early access. Or estimate your own AI Tax first with the AI Tax Calculator.

Common Questions

Why do I keep evaluating new AI tools every few weeks?

Because new tools arrive faster than anyone can read, so the shortlist never closes. The cure is not more discipline; it is to stop being the person who has to keep the shortlist. The full taxonomy of where this cost sits is in the seven AI taxes, added up.

How do I choose the right AI tool when there are too many options?

You change the question. Adding more tools past three or four reduces the gains, so “pick the best tool” is the wrong target in a field that reshuffles weekly. The better move is to let a system that knows your business pick the approach for each job, which is the shift covered in the Three-Layer Pyramid.

How do I know if an AI tool will actually help my business?

A twenty-minute trial shows you a capability, not whether it survives your data, your team, and your real process. A tool helps when it finishes a job your business needs end to end against your context. That is a far higher bar than “is this tool good,” and most standalone tools cannot clear it alone.

Is evaluating AI tools just normal due diligence?

One round is. The Discovery Tax is the recurring re-evaluation every few weeks as the landscape shifts, the diligence that never finishes. Keep the one-time look; the repeat is the cost you can stop paying.

Why do most companies see no measurable return on AI?

Because choosing and wiring up tools is not the same as getting a result. An MIT study of 300 initiatives found roughly 95% reported no measurable return, and most of that effort is still stuck at the choosing and connecting, not at the outcome. The recurring re-wiring that follows is its own cost, covered in why your AI keeps breaking.