Joshua Kushner of Thrive Capital sent his first-ever letter to investors in August 2026. Written personally by a VC who has deployed billions of dollars into AI investments, the letter carried a single warning: "The AI opportunity is enormous. But letting excitement erode investment discipline would be a serious mistake."

It sounds like a story confined to Silicon Valley investors. Yet the same mistake repeats every week at a much smaller scale — inside the solo businesses and small teams adopting AI subscription services.

When the Tool Arrives Before the Problem

The process of adopting an AI tool usually unfolds in this order: you see the tool in someone's review or a YouTube video. You read that it multiplies productivity several times over. You start a free trial, and two weeks later you upgrade to a paid plan.

Nowhere in this sequence does the question "what task am I actually stuck on right now?" ever come up.

When the problem isn't written down first, the tool ends up manufacturing the problem instead. Subscribe to Claude, and pressure builds to "write with AI." Use Midjourney, and you start steering toward "content that needs images." Work quietly begins reorganizing itself around whatever the tool happens to do well. It happens silently, the moment the order of adoption gets reversed.

Pin Down the Stuck Step, Then Pick the Tool

Adoption that starts with the problem follows a different order. Pick the one task that's currently slowest. Write down why it's slow and exactly which step it stalls at. Only then check whether AI can actually intervene at that specific step.

Take a solo business owner who writes a newsletter every week: whether the slowdown is in gathering material, drafting, or editing calls for an entirely different tool. For someone stuck at gathering material, a text-generation tool is no help. For someone stuck at drafting, an image tool is irrelevant.

Once that distinction exists, deciding not to use a tool also becomes possible — because you can now judge, "AI wouldn't solve the problem I actually have at this step." Without that distinction, subscriptions keep piling up while the actual bottleneck stays exactly where it was.

Applying Kushner's Investment Discipline to Tool Adoption

The investment discipline Kushner describes is a procedure for checking underlying assumptions before making a call: What real market problem does this company solve? Is it solving that problem fast enough? Does this investment hold up independent of the AI trend?

Carried over to tool adoption, it becomes: What specific step of my work does this tool actually handle? Is the quality it delivers at that step good enough compared with doing it myself? Would this subscription hold up even without the AI trend?

That last question cuts the deepest. Attach the condition "independent of the AI trend," and a good number of the tools you were about to adopt fail the test. Would you have chosen this tool if the trend didn't exist? If the answer is no, the tool isn't solving a work problem — it's chasing a trend.

One Line Defining the Problem Filters Out the Wrong Subscriptions

Make it a habit to write the problem down in a single line before adopting anything, and the pace of your subscription decisions slows. It feels inefficient, but it's the opposite. Learning a wrongly adopted tool, re-tuning your workflow around it, and confirming that it doesn't actually work all cost time — and that time is more expensive than the subscription fee itself.

There's no reason the checklist venture investors apply to decisions worth hundreds of millions of dollars shouldn't work on a $20–30-a-month subscription call. The smaller the scale, the easier it is to wave through without scrutiny. The low price, the easy cancellation, and the abundance of alternatives all lower the bar. Which is exactly why small subscription decisions pile up and end up disrupting the workflow anyway.

Still, writing the problem down first is harder than picking a tool. Answering "what's actually stuck right now" with any specificity requires stepping back from your own work. AI doesn't do this part for you. It's work you can do without any tool at all — and if you've decided to use one, it's the work that has to come first.