The honest question a founder asks at renewal time is not whether a tool is good. It is whether any of this AI spend is an asset, or whether it is money down the drain. After a couple of years of subscriptions, the suspicion is fair: the bills recur, the tools change, and there is rarely anything you could point to and call a thing the business owns. JynAI built Works, an AI Business OS, on the opposite premise, that your AI spend should compound into something you keep. This piece is about the line between an asset and an expense, and a single test for telling which of your AI dollars holds value.

It depends on one thing: whether what you built into the tool stays when the tool changes. Accounting already draws this line precisely. The work of configuring and implementing a system is capitalized as a balance-sheet asset and amortized over a useful life of two to five years, while a subscription is operating expense, money that buys access for a period and leaves no residual value when it stops. Configuration is capital. A subscription is rent.

That distinction is the whole question in miniature. The hours you spend teaching a system your customers, your voice, and your workflows are either capital or rent, depending entirely on whether the system keeps them. If it does, you built an asset that appreciates. If it forgets the day you leave, you were renting all along, whichever line of the budget the spend sat on.

Configuration and implementation are capitalized as a balance-sheet asset. A subscription is operating expense.
Wall Street Prep, 2023

The part that carries forward is the configured, contextual knowledge, if the system is built to keep it. That accumulated, intangible layer is exactly where value lives in a modern business. Intangible assets are now about 92 percent of the market value of the S&P 500, up from 17 percent in 1975, a near-total inversion in where corporate value sits.

Read against your AI spend, that statistic is a directive. The value of what you are building is not the software license. It is the context the software accumulated about your business, the configuration you tuned, the institutional knowledge you moved out of your head and into the system. That is the appreciating asset. The license is just the meter. So the question for any AI dollar is whether it is thickening that intangible layer, or just paying for access to a tool that keeps none of it.

Intangible assets are now about 92 percent of the market value of the S&P 500, up from 17 percent in 1975.
Ocean Tomo, 2025

Because most of it is rent, and most rent is wasted. The subscription model buys access, not ownership, and the waste in that model is enormous: 30 to 40 percent of SaaS licenses go unused, and companies reclaim only 5 to 15 percent of the waste they identify, with SaaS spend growing three to five times faster than overall IT budgets. That is money leaving every month and leaving nothing behind.

The feeling of rented time is accurate because the arrangement is rented time. You pay, you get access, and the moment you stop paying, the access and everything inside it is gone. There is no equity in a subscription. A tool that forgets is rented time you never own, and the reason your AI can feel like a treadmill is that a treadmill is exactly what a stack of forgetful subscriptions is: motion you pay for that takes you nowhere you get to keep. The recurring cost of that forgetting, paid on every tool switch, is the reset tax.

You make it last by spending where the value compounds, on a setup that keeps the context and configuration instead of renting you access to it. The test to run on every AI dollar is the one accounting implies: will this still be worth something next year, or am I renting time I never own. The work that holds value is the work the system remembers.

This is the same instinct businesses already apply to their data: treated as an asset you keep and build on, it appreciates; treated as something disposable, it never gets the chance. The same is true of your AI configuration. The hours you put into teaching a system your business are capital if the system keeps them and rent if it does not, and the difference shows up not this quarter but next year, when one founder is still building on what they set up and another is setting it up again.

What a setup that compounds looks like

The bar any honest answer has to clear is residual value: the spend has to keep paying after you stop spending. That is the bar Works was built to clear. A few concrete pieces of how it does it:

• The context becomes the asset: Works learns your customers, voice, pipeline, and history as a side effect of running the business on it, and that accumulated context feeds every workflow, agent, and chat. The pain it relieves is starting over; the gain is an intangible asset that appreciates the longer Works runs, instead of a subscription that buys access and keeps nothing.
• The investment survives the upgrade: A six-month-old workspace runs on today’s Works without a re-setup. When the models underneath change, your configuration stays put; new models, connectors, and workflows land inside the setup you already built. That is capital compounding, not rent recurring, the same reason a setup that keeps its context does not pay the reset tax.
• It clears the bar at a founder’s price: The full capability set sits at the $49 Pro tier, so building an asset is not gated behind an enterprise contract. We built this for ourselves first: six teams running in ninety days, after years of fragmented AI experiments that left no asset behind because each tool kept nothing. The operation we built kept appreciating; the throwaway tools evaporated.

Invest where it compounds. Get early access. Or get the Compounding brief first, the short read on which AI spend holds value and which evaporates.

The test to carry away is simple. On every AI dollar, ask: will this still be worth something next year. Configuration is capital. A subscription is rent. The work that holds value is the work the system remembers.

Is AI a sunk cost or an investment for my business?

AI spend is an investment when it leaves something behind and a sunk cost when it leaves nothing. Accounting makes the split concrete: implementation and configuration are capitalized as balance-sheet assets amortized over two to five years, while a subscription is operating expense with zero residual value. The hours you spend teaching a system your business follow the same rule: capital if the system keeps them, rent if it forgets the day you leave. The fuller case is in how the foundation gets stronger.

How do I build so that capability stacks instead of resetting?

You build on a foundation that keeps what it learns, so each new capability lands on top of the last instead of replacing it. That means keeping context and configuration in one durable layer rather than scattered across tools that each start over. The mechanism, how a foundation compounds rather than resets, is covered in the foundation that gets stronger; the cost of getting it wrong is the reset tax.

What of my AI work actually carries forward when I switch tools?

The part that carries forward is the configured, contextual knowledge, but only if it lives in a layer you own rather than inside the tool itself. Intangible assets now make up roughly 92 percent of S&P 500 market value, and the same principle applies at the business level: the context, configuration, and institutional knowledge are where the value sits. If that layer is trapped inside a single platform’s database, none of it follows you out. More on the memory that compounds in memory that compounds.

What is residual value and why does it matter for AI spend?

Residual value is what an investment is still worth after you stop paying. For AI, it is the test that splits asset from rent: a tool that keeps your configuration and context leaves something behind when you leave; a tool that forgets leaves nothing. If the residual value is zero, you were renting, regardless of how the spend was categorized in the budget.

Why does AI feel like a treadmill?

The treadmill feeling is accurate because the arrangement is rent: you pay, you get access, and the moment a tool changes, the setup and what it learned resets with it. Studies show 30 to 40 percent of SaaS licenses go unused, meaning the drain starts before a tool even earns its place. The exit is not a better subscription. It is a layer that accumulates rather than expires.

1. Wall Street Prep, “Capitalized Software Costs,” 2023
2. Ocean Tomo (a part of J.S. Held), “Intangible Asset Market Value Study,” 2025
3. Certero (Joseph Lobo), “Five Surprising Stats About SaaS Waste in Enterprise IT,” 2025
   
   

Keep reading:

• The reset tax, the cost of forgetting
• The foundation that gets stronger
• Memory that compounds
• Compounding AI