What starting over on a new AI tool really costs

Every AI tool change restarts a hundred-hour setup, and the switch is rarely the one you chose. Here is the bill almost no one puts on the books.

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

Every new AI tool charges you a hundred hours to start over.
Made with Works

TL;DR

The Switching Tax is the hundred-hour cycle every AI tool change restarts: the setup, the prompts, the integrations, and the team’s habit all reset to zero. In AI the switch is rarely yours to choose, because models, platforms, and agency stacks change on someone else’s schedule. The exit is to hold your configuration in a layer a tool swap cannot reach, so a switch changes one component, not the whole business.


In this article

You did the work once. You picked the model, wrote the prompts, wired up the integrations, and got the team into the habit. Then the model changed, or the platform churned, or an agency arrived with a different stack, and you started over. That restart is one of the hidden costs of AI, the sixth of them, and almost no one puts it on the books, because in AI the switch is rarely a thing you do once and finish.

What is the real cost of switching AI tools

Far larger than the migration plan says, because the plan only counts the obvious part. The honest cost is the whole hundred-hour cycle restarting: the configuration, the prompts, the integrations, and the team’s muscle memory, all reset at once. Executives routinely budget a few weeks for an AI vendor switch, and the migrations that actually get attempted run long, get messy, or fail outright far more often than anyone planned for.

That underestimation is the tax itself. The gap between the four-week plan and the four-month reality is not bad luck, it is the built-in cost of a reset nobody priced in. The hard numbers back it up: the average business loses about $315,000 per platform migration, with 57 percent of IT leaders spending over a million dollars on migrations in a single year and 70 percent reporting developer burnout. A migration on a slide is a clean arrow from old tool to new tool. A migration in practice is re-doing the setup that made the old tool actually fit your business, and that setup was the expensive part the first time too.

The average business loses about $315,000 per platform migration, and 57 percent of IT leaders spend over a million dollars on migrations in a single year.
CIO Dive, 2025

So the real cost of switching is not the new subscription. It is paying the hundred-hour setup a second time, and a third, on a schedule you do not control.

Why do I have to start over every time an AI tool changes

Because the work that made a tool yours does not transfer to the next one, and in AI the tools change constantly. A model gets deprecated. A tool you depend on gets killed or churns out from under you. An agency brings a different stack. Each of those restarts the cycle, and the prompts, the integrations, and the context you built do not come along.

This is the part founders feel and rarely name: the switch is usually not your decision. And the true cost is not a subscription line. AI vendor lock-in is often a six-figure cost event for a single workload and a seven-figure one for a shared platform, because during migration you pay twice while you revalidate outputs, and the most expensive part is the hidden glue code that became permanent. Read that the right way and it is not a story about bad planning. It is a story about how much hidden setup lives inside a tool you have made your own, setup that has to be rebuilt the moment the tool underneath you changes.

During migration you pay twice while you revalidate outputs, and the most expensive part is the hidden glue code that became permanent.
StackAI, The Hidden Costs of Vendor Lock-In

It helps to see where this 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 Switching Tax lives at the very bottom, in the churn of tools changing underneath you, and every change at that bottom layer forces a rebuild before anything at the top can move again. You can spend a quarter starting over and have nothing new at the result layer to show for it.

How do I avoid getting locked into an AI vendor

Not by finding the one tool you will never have to leave, because that tool does not exist and the field will not let you stand still anyway. You avoid the trap by holding your configuration somewhere a tool swap cannot reset it, so the business stops depending on any single vendor staying exactly as it is.

That is a different move from the usual advice. “Stay flexible, keep your options open” still leaves the hundred hours of setup trapped inside whichever tool you are using this quarter, which means the next switch still resets you. The real answer is to separate the configuration from the tool: keep the prompts, the integrations, the context, and the way your business runs in a layer that persists, while the underlying model or app becomes something you can change without starting over. This is precisely the fork industry analysts now flag: as AI becomes a workflow layer, lock-in turns fundamental, and the organizations that did not build an abstraction layer are on a path of deepening dependency and rising exit cost.

This is also why the lock-in conversation and the switching conversation are two sides of one coin. Lock-in is the cost of staying. The Switching Tax is the cost of leaving. Both come from the same root, that your business setup is fused to a tool you do not control. We go deeper on the staying side in why locked-in AI dies with the vendor.

How do you stop paying the Switching Tax

Not with a tool you never leave or a migration you finally get right. By holding the configuration in a place a swap cannot reach, so when the model or the tool changes underneath you, the setup, the prompts, the integrations, and the team’s habit all stay put. The switch stops being a reset of the business and becomes a change of one component.

A real answer here has to clear a clear bar: hold the configuration so a tool swap does not reset it; reach the tools the business already runs on, so a new app is a connection rather than a rebuild; and absorb a model change without rebuilding the workflows above it. JynAI built Works, an AI Business OS, to clear exactly that bar. Here is the fit, plainly.

  • Pain: every tool change restarts the hundred-hour setup.
    Learns Your Business holds your customers, voice, pipeline, and history as durable context, so a six-month-old workspace runs on today’s Works without a re-setup.
    Gain: the configuration persists, so a swap is a change of one part, not a reset of the whole business.

  • Pain: a new model ships and you have to re-route everything through it by hand.
    Works Across Your Stack reaches 3,000+ apps via native integrations and Pipedream, and when a new model joins the pool, the existing workflows start using it without you touching anything.
    Gain: the model upgrade lands under your setup, not on your weekend.

  • Pain: the prompts, the integrations, and the team’s habit are all trapped inside a tool you do not control.
    Expert-Grade Workflows and Blueprints store the proven run in Works, not in the tool that ran it, so a vendor swap changes one component and the plays stay put.
    Gain: the hundred hours of setup you already paid survive the next switch.

  • Pain: you are the one who has to notice when something broke.
    Receipts logs every run, every action, and every outcome, so the activity record is yours regardless of which tool ran the step.
    Gain: the audit trail lives in Works, not in a vendor you may leave next quarter.

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 stopping the Switching Tax does not require a budget conversation first.

And this is not a theoretical fix. Machintel ran the same start-over cycle for years, paying the hundred-hour reset each time the ground moved, before the layer that holds the configuration was in place. Once it was, the team stopped starting over. We are biased about our own product, of course. The argument under it does not need us: if the expensive part of switching was always the setup trapped inside the tool, then the answer was never a better migration. It was holding the setup somewhere a swap cannot reach.

The Switching Tax is paid by people who keep their configuration where a tool change can reset it. The way out is to move it somewhere a tool change cannot touch. That is a different decision from picking a safer tool or planning a cleaner migration. It is deciding that the setup belongs to the business, not to the stack underneath it.

Stop starting over. Sign up for early access. Or get a rough number for your own AI Tax first with the AI Tax Calculator.

Common Questions

What is the real cost of switching AI tools?

Much larger than the migration slide suggests, because the slide prices the move but not the rebuild. Every part of the setup that made the old tool yours, the configuration, the prompts, the integrations, the team’s muscle memory, resets to zero and must be paid again. The data puts a number on that: the average business loses about $315,000 per platform migration, with 57 percent of IT leaders spending over a million dollars on migrations in a single year. That is not a migration cost. That is the cost of rebuilding the work the old tool had already absorbed.

Why do I have to start over every time an AI tool changes?

Because the work that made a tool yours does not transfer to the next one, and in AI the tools change constantly on someone else’s schedule. A model gets deprecated. A platform churns. An agency brings a different stack. Each of those restarts the cycle, and the prompts, integrations, and context you built do not come along. The switch is rarely your decision. The rebuild lands on you anyway.

How do I avoid getting locked into an AI vendor?

Not by finding the one tool you will never have to leave. You avoid the trap by holding your configuration somewhere a tool swap cannot reset it. Separate the prompts, the integrations, the context, and the way your business runs into a layer that persists, while the underlying model or app becomes something you can change without starting over. Lock-in is the cost of staying; the Switching Tax is the cost of leaving. Both come from the same root: your setup is fused to a tool you do not control. We go deeper on the staying side in why locked-in AI dies with the vendor.

How many times does the average business pay the Switching Tax?

As many times as the tools underneath them change, which in AI is not rare. Nearly three in four enterprises say losing an AI vendor would disrupt core operations, and models, agent platforms, and tools change on someone else’s schedule. Each change restarts the cycle. The tax is not a one-time migration event; it is a recurring cost that compounds every time the ground moves.

What is the Switching Tax?

The Switching Tax is the sixth of seven costs in the full AI bill for a founder-led business. It names the hundred-hour cycle that restarts whenever a tool changes: the setup, the prompts, the integrations, and the team’s habit all reset, and the reset is almost never the founder’s choice. Models get deprecated, platforms churn, agency stacks rotate, and each one triggers another rebuild. Nearly three in four enterprises say losing an AI vendor would disrupt core operations, which puts the cycle’s stakes in plain numbers. The full picture is in the AI Tax.

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 starting over on a new AI tool really costs

Every AI tool change restarts a hundred-hour setup, and the switch is rarely the one you chose. Here is the bill almost no one puts on the books.

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

Every new AI tool charges you a hundred hours to start over.
Made with Works

TL;DR

The Switching Tax is the hundred-hour cycle every AI tool change restarts: the setup, the prompts, the integrations, and the team’s habit all reset to zero. In AI the switch is rarely yours to choose, because models, platforms, and agency stacks change on someone else’s schedule. The exit is to hold your configuration in a layer a tool swap cannot reach, so a switch changes one component, not the whole business.


In this article

You did the work once. You picked the model, wrote the prompts, wired up the integrations, and got the team into the habit. Then the model changed, or the platform churned, or an agency arrived with a different stack, and you started over. That restart is one of the hidden costs of AI, the sixth of them, and almost no one puts it on the books, because in AI the switch is rarely a thing you do once and finish.

What is the real cost of switching AI tools

Far larger than the migration plan says, because the plan only counts the obvious part. The honest cost is the whole hundred-hour cycle restarting: the configuration, the prompts, the integrations, and the team’s muscle memory, all reset at once. Executives routinely budget a few weeks for an AI vendor switch, and the migrations that actually get attempted run long, get messy, or fail outright far more often than anyone planned for.

That underestimation is the tax itself. The gap between the four-week plan and the four-month reality is not bad luck, it is the built-in cost of a reset nobody priced in. The hard numbers back it up: the average business loses about $315,000 per platform migration, with 57 percent of IT leaders spending over a million dollars on migrations in a single year and 70 percent reporting developer burnout. A migration on a slide is a clean arrow from old tool to new tool. A migration in practice is re-doing the setup that made the old tool actually fit your business, and that setup was the expensive part the first time too.

The average business loses about $315,000 per platform migration, and 57 percent of IT leaders spend over a million dollars on migrations in a single year.
CIO Dive, 2025

So the real cost of switching is not the new subscription. It is paying the hundred-hour setup a second time, and a third, on a schedule you do not control.

Why do I have to start over every time an AI tool changes

Because the work that made a tool yours does not transfer to the next one, and in AI the tools change constantly. A model gets deprecated. A tool you depend on gets killed or churns out from under you. An agency brings a different stack. Each of those restarts the cycle, and the prompts, the integrations, and the context you built do not come along.

This is the part founders feel and rarely name: the switch is usually not your decision. And the true cost is not a subscription line. AI vendor lock-in is often a six-figure cost event for a single workload and a seven-figure one for a shared platform, because during migration you pay twice while you revalidate outputs, and the most expensive part is the hidden glue code that became permanent. Read that the right way and it is not a story about bad planning. It is a story about how much hidden setup lives inside a tool you have made your own, setup that has to be rebuilt the moment the tool underneath you changes.

During migration you pay twice while you revalidate outputs, and the most expensive part is the hidden glue code that became permanent.
StackAI, The Hidden Costs of Vendor Lock-In

It helps to see where this 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 Switching Tax lives at the very bottom, in the churn of tools changing underneath you, and every change at that bottom layer forces a rebuild before anything at the top can move again. You can spend a quarter starting over and have nothing new at the result layer to show for it.

How do I avoid getting locked into an AI vendor

Not by finding the one tool you will never have to leave, because that tool does not exist and the field will not let you stand still anyway. You avoid the trap by holding your configuration somewhere a tool swap cannot reset it, so the business stops depending on any single vendor staying exactly as it is.

That is a different move from the usual advice. “Stay flexible, keep your options open” still leaves the hundred hours of setup trapped inside whichever tool you are using this quarter, which means the next switch still resets you. The real answer is to separate the configuration from the tool: keep the prompts, the integrations, the context, and the way your business runs in a layer that persists, while the underlying model or app becomes something you can change without starting over. This is precisely the fork industry analysts now flag: as AI becomes a workflow layer, lock-in turns fundamental, and the organizations that did not build an abstraction layer are on a path of deepening dependency and rising exit cost.

This is also why the lock-in conversation and the switching conversation are two sides of one coin. Lock-in is the cost of staying. The Switching Tax is the cost of leaving. Both come from the same root, that your business setup is fused to a tool you do not control. We go deeper on the staying side in why locked-in AI dies with the vendor.

How do you stop paying the Switching Tax

Not with a tool you never leave or a migration you finally get right. By holding the configuration in a place a swap cannot reach, so when the model or the tool changes underneath you, the setup, the prompts, the integrations, and the team’s habit all stay put. The switch stops being a reset of the business and becomes a change of one component.

A real answer here has to clear a clear bar: hold the configuration so a tool swap does not reset it; reach the tools the business already runs on, so a new app is a connection rather than a rebuild; and absorb a model change without rebuilding the workflows above it. JynAI built Works, an AI Business OS, to clear exactly that bar. Here is the fit, plainly.

  • Pain: every tool change restarts the hundred-hour setup.
    Learns Your Business holds your customers, voice, pipeline, and history as durable context, so a six-month-old workspace runs on today’s Works without a re-setup.
    Gain: the configuration persists, so a swap is a change of one part, not a reset of the whole business.

  • Pain: a new model ships and you have to re-route everything through it by hand.
    Works Across Your Stack reaches 3,000+ apps via native integrations and Pipedream, and when a new model joins the pool, the existing workflows start using it without you touching anything.
    Gain: the model upgrade lands under your setup, not on your weekend.

  • Pain: the prompts, the integrations, and the team’s habit are all trapped inside a tool you do not control.
    Expert-Grade Workflows and Blueprints store the proven run in Works, not in the tool that ran it, so a vendor swap changes one component and the plays stay put.
    Gain: the hundred hours of setup you already paid survive the next switch.

  • Pain: you are the one who has to notice when something broke.
    Receipts logs every run, every action, and every outcome, so the activity record is yours regardless of which tool ran the step.
    Gain: the audit trail lives in Works, not in a vendor you may leave next quarter.

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 stopping the Switching Tax does not require a budget conversation first.

And this is not a theoretical fix. Machintel ran the same start-over cycle for years, paying the hundred-hour reset each time the ground moved, before the layer that holds the configuration was in place. Once it was, the team stopped starting over. We are biased about our own product, of course. The argument under it does not need us: if the expensive part of switching was always the setup trapped inside the tool, then the answer was never a better migration. It was holding the setup somewhere a swap cannot reach.

The Switching Tax is paid by people who keep their configuration where a tool change can reset it. The way out is to move it somewhere a tool change cannot touch. That is a different decision from picking a safer tool or planning a cleaner migration. It is deciding that the setup belongs to the business, not to the stack underneath it.

Stop starting over. Sign up for early access. Or get a rough number for your own AI Tax first with the AI Tax Calculator.

Common Questions

What is the real cost of switching AI tools?

Much larger than the migration slide suggests, because the slide prices the move but not the rebuild. Every part of the setup that made the old tool yours, the configuration, the prompts, the integrations, the team’s muscle memory, resets to zero and must be paid again. The data puts a number on that: the average business loses about $315,000 per platform migration, with 57 percent of IT leaders spending over a million dollars on migrations in a single year. That is not a migration cost. That is the cost of rebuilding the work the old tool had already absorbed.

Why do I have to start over every time an AI tool changes?

Because the work that made a tool yours does not transfer to the next one, and in AI the tools change constantly on someone else’s schedule. A model gets deprecated. A platform churns. An agency brings a different stack. Each of those restarts the cycle, and the prompts, integrations, and context you built do not come along. The switch is rarely your decision. The rebuild lands on you anyway.

How do I avoid getting locked into an AI vendor?

Not by finding the one tool you will never have to leave. You avoid the trap by holding your configuration somewhere a tool swap cannot reset it. Separate the prompts, the integrations, the context, and the way your business runs into a layer that persists, while the underlying model or app becomes something you can change without starting over. Lock-in is the cost of staying; the Switching Tax is the cost of leaving. Both come from the same root: your setup is fused to a tool you do not control. We go deeper on the staying side in why locked-in AI dies with the vendor.

How many times does the average business pay the Switching Tax?

As many times as the tools underneath them change, which in AI is not rare. Nearly three in four enterprises say losing an AI vendor would disrupt core operations, and models, agent platforms, and tools change on someone else’s schedule. Each change restarts the cycle. The tax is not a one-time migration event; it is a recurring cost that compounds every time the ground moves.

What is the Switching Tax?

The Switching Tax is the sixth of seven costs in the full AI bill for a founder-led business. It names the hundred-hour cycle that restarts whenever a tool changes: the setup, the prompts, the integrations, and the team’s habit all reset, and the reset is almost never the founder’s choice. Models get deprecated, platforms churn, agency stacks rotate, and each one triggers another rebuild. Nearly three in four enterprises say losing an AI vendor would disrupt core operations, which puts the cycle’s stakes in plain numbers. The full picture is in the AI Tax.