The $10K Test

Everyone loves to start AI projects.

But a lot of projects shouldn’t be started at all. A good chunk of them will just sit there collecting digital dust, and piling up cost with no real return. Not because the tech failed. But because someone said "yes" to a project that should've been killed right out of the gate.

How do you separate "interesting" ideas from valuable ones? That's where the $10K Test comes in.

It's a simple 3-question filter I use to decide which projects get the green light and which get killed fast – without guilt.

Today, I'll show you how it works.

The Problem with "Interesting" AI Ideas

Most "successful" AI projects aren't actually that successful. They're participation trophies dressed up as business wins.

I see this everywhere.

• Companies celebrating AI chatbots with "94% satisfaction rates" that free up zero additional resources.

• Teams patting themselves on the back for "20% efficiency gains" that don't translate to a single dollar saved.

• Execs showing off AI dashboards that look impressive but change exactly zero decisions.

The AI world has developed a collective delusion: If the technology works, the project succeeded.

That's backwards.

If you’ve been following my blog, you know that the 10K Rule helps you spot high-value problems in your business. It’s a great initial filter to identify which problems actually have the potential to give you a meaningful payback.

But just because a problem is worth solving doesn’t mean every solution deserves to be built.

That’s where the AI $10K Test comes in.

It’s a simple, ruthless filter I use to kill weak solution ideas — and green-light the few worth building.

Let me show you how it works.

The $10K Kill/Keep Tree

AI success is less about achieving high accuracy or user satisfaction scores, and more about recurring, measurable business value that compounds over time.

Companies getting this right don't celebrate their chatbots' cleverness.
They celebrate the $20K in monthly support work those chatbots eliminated.

So before you write a single line of code or book a single meeting, run every AI idea through these 3 questions:

Let me break down why each question matters:

Question 1: Will this generate at least $10K+ in recurring value?

This is your first gate.

Depending on your ambition level, you could target yearly, quarterly, monthly, or even weekly paybacks. If you can't hit your defined threshold, you're probably looking at an "interesting" project, not a valuable one.

Because of the inverse economics of AI, you need to prove value more than ever.

Talking about proving value…

Question 2: Can you prove that value in 90 days or less?

Speed isn’t optional — it’s the essence.

If it takes longer than a quarter to show results, you’re probably starting too big or solving the wrong problem.

And yes — I know that 90 days sounds unrealistic to big organizations. But with the pace of AI today, you’ll need to get used to it sooner rather than later.

Quick wins build momentum. Long, uncertain projects kill it.

Question 3: Does it augment an existing process (not replace or bypass it)?

The fastest path to impact is to plug AI into something that already works.

Starting from scratch is slow, risky, and expensive. The best AI projects don’t reinvent — they upgrade. Look for live workflows or products where a small augmentation unlocks big returns.

Not sure how to start? Revisit the principles for effective workflow augmentation.

If your idea passes all three tests? Green-light it. Build lean. Move fast.
Fail even one? Kill it or go back to the drawing boards. No regret.

Common Traps That Fail the Test

Of course, passing the test isn’t always black-and-white.

Plenty of ideas land in “maybe” territory. They sound smart, but fall apart under pressure.

Here are some common patterns that feel strategic but fail the $10K Test:

• ❌ "It'll boost productivity by 20%" — Sounds impressive, but where's the $10K? If you can't translate percentage gains into dollar amounts, it's not ready.

• ❌ "We need this data analysis dashboard" — One-time insights don't generate recurring value. Unless it changes how decisions get made every month, it's just expensive reporting.

• ❌ "It'll help us stay competitive" — Fear-driven projects rarely have clear ROI. If you can't prove value in 90 days, you're building insurance, not impact.

• ❌ "We'll revolutionize customer service with AI agents" — Starting from zero instead of augmenting your existing support process. Recipe for a two-year project with uncertain adoption.

• ❌ "Let's automate our entire workflow" — Ambitious, but rarely works as planned. Instead, increase integration and automation over time. Still got a $10K case? There you go!

These ideas feel important but can't answer the basic question: "What specific problem will this solve, and how much is that problem costing us right now?"

If you can't answer that in one sentence, you don't have a $10K opportunity – you have a research project.

What to Do With Borderline Ideas

I see teams agonizing over these borderline cases, trying to massage the numbers or adjust the scope until it "technically" qualifies. Don't do that.

Borderline projects are distractions in disguise.

You have two choices:

Option 1: Rescope it ruthlessly Strip it down until it clearly passes all three tests. Maybe that "full workflow automation" becomes "augment the most painful 10 minutes of the process." Maybe that "comprehensive analytics platform" becomes "one critical report that saves 2 hours daily."

If you can't rescope it to clearly pass, move to Option 2.

Option 2: Kill it early This isn't giving up—it's protecting your focus. Every borderline project you pursue is energy you're not spending on a clear winner.

Now, if your idea does survive the Kill/Keep Tree, that's when you get granular. In my consulting work, I use a detailed $10K Scorecard to estimate the actual value potential—factoring in implementation costs, adoption curves, and risk multipliers.

But if f it doesn't pass the three basic questions first, it never even gets scored. The tree is your first filter. Everything else comes after.

Conclusion

The bottom line is: Your goal isn't to use AI everywhere. It's to use AI where it creates undeniable value.

The $10K Test helps you find those spots and ignore everything else.
It's honest. And in AI, honesty is ROI.

Kill the weak ideas fast.
Build the strong ones lean.

Your future self (and your budget) will thank you.

Keep innovating (profitably)!

See you next Friday,
Tobias