Share from 0:00
0:00
/

Jevons Paradox: Will Cheap AI Bury Your Leaders in Decks?

Innovation Unpacked: when checking ideas gets cheap, the jam moves to the people who decide.

Note: links to a free checklist and playbook at the end!


In 1865, Better Steam Engines Made Britain Burn More Coal. Cheap AI Will Do the Same to Your Strategy Team.

In 1865, an economist named William Stanley Jevons wrote a book called The Coal Question. He noticed something odd. Steam engines had gotten much more efficient. They needed less coal to do the same work. You’d think Britain would burn less coal.

It burned more. Cheaper steam power meant people used it for far more things. Total coal use went up, not down. Economists still call this the Jevons paradox.

In January 2025, the week a cheaper Chinese AI model called DeepSeek sent tech stocks down, Microsoft’s CEO Satya Nadella posted that the Jevons paradox had struck again. His point was that as AI gets more efficient and easier to get, we’ll use far more of it.

He’s right. And for strategy teams, that’s a problem nobody is planning for.

Here’s my thesis. AI is making it cheap to check an idea. So teams will check far more ideas and send far more of them up the chain. But the people who decide didn’t get cheaper. They have the same hours they had last year. The jam moves from the research step to the decision step. If you don’t put a filter in front of those people, cheap AI will make your strategy work slower, not faster.

How ideas get decided today

Think about how a company decides what to build. Someone has an idea. A team researches it, builds a deck and brings it to leadership. Leadership says yes, says no or asks for more.

That process used to be slow and costly. The slowness did a job nobody talks about. It kept the number of ideas low. Only the ideas someone cared enough about made it to a meeting.

AI changes that. In my view, research, market sizing and decks that used to take weeks can now take days. A tool can draft all three in a fraction of the time.

What happens when checks get cheap

Say AI cuts deck drafting time by about 60%. Teams don’t send the same number of decks faster. They send far more decks. If idea checks get 10 times cheaper, you might see 10 times as many land on the leadership calendar.

But leadership time didn’t get cheaper. There are still the same number of people in the room, with the same number of hours. So the jam moves. It used to be at the research step. Now it’s at the decision step.

More decks, the same number of good decisions. That’s the trap.

This is the same pattern Jevons saw. When something gets cheaper to use, people often end up using more of it, not less. Cheaper steam meant more coal. Cheaper idea checks mean more ideas. The total load goes up. It just lands somewhere new.

Why more slides don’t help

You might think more ideas is fine, as long as leaders can sort them. But look at what a deck is built to do.

A deck is built to persuade, not to test. Two decks about the same idea can reach opposite answers, and both can look convincing. Each one picks the numbers that make its case. Each one is shaped by the person presenting it.

When there are a few decks, leaders can dig into each one. When there are many more of them, they can’t. So they fall back on gut feel, politics or whoever presents best.

AI makes this worse, not better. It’s very good at making a deck look finished. Clean charts, confident sizing, a tidy story. So the look of a deck tells you less than it used to. The cheaper the decks get, the more you need a way to sort them that doesn’t depend on a meeting.

The fix is gates in code

Here’s the approach I use, from the Venture Proof method. Every idea runs through the same checks, written as rules, before a person ever reads it. The rules live in a sheet or a short script. The same inputs always give the same answer.

The checks ask five simple questions, in this order:

  1. What result does this idea produce, and for whom?

  2. What does that result cost today?

  3. What’s the lowest it could cost if nothing were wasted?

  4. Is the gap big enough to matter?

  5. Will anyone pay to close it?

Question 3 is the floor. It holds only what the work truly needs: compute, materials and the checks a law or a contract requires. No waiting, no handoffs, no extra sign-offs and no markup. Question 4 compares today’s cost with that floor. I call that ratio the Physics Gap.

Notice the order. The cheap questions come first. Question 5 needs interviews, so you only ask it about ideas that passed the first four.

If an idea fails a check, it stops. Nobody has to sit through the deck. The ideas that pass come with the math attached, so leaders can see why they passed. They can argue with the numbers instead of the slides.

This is how you use the Jevons effect instead of fighting it. You want more ideas, since cheap checks are a good thing. You just don’t want all of them reaching the people who decide.

A worked example

Here’s a made-up strategy team to show how it works. The numbers are examples, not client data.

Last quarter, the team reviewed 40 ideas. Only 12 got a clear yes or no. That’s 30%. The rest got “come back with more.” Leaders set aside 2 hours a week for idea reviews. Over 13 weeks, that’s 26 hours. At 45 minutes a review, there’s room for about 34 reviews.

Now the team has AI tools for research and decks. They expect about 90 ideas next quarter. That won’t fit.

So the team sits down with its leaders and turns each question into a pass or fail rule. They set the numbers before anyone sees this quarter’s ideas. One rule says today’s cost must be at least 10 times the floor, and the gap across a year must be at least $500,000. Another says at least 3 of 5 people who control the spend must say yes to a stated price. Those bars are this team’s choices. Yours will depend on your size and goals.

Here’s how two ideas run through it.

Idea A is an AI tool that writes first drafts of supplier contracts. Today, each contract costs about $1,500 in outside legal review and staff time. The floor is about $40, a lawyer’s required 15-minute check plus a few dollars of compute. $1,500 is about 37 times $40. The company signs about 800 of these a year, so the gap across a year is about $1,168,000. Four of five procurement heads say yes to a stated price. Idea A goes to leaders.

Idea B is an AI chatbot for the employee handbook. Today, each question costs about $12 in HR staff time. The floor is about $1. The ratio passes at 12 times. But there are only about 3,000 questions a year, so the gap across a year is about $33,000. Idea B stops. Its owner gets the reason. It can come back if the volume grows.

Before trusting the gate, the team runs last quarter’s 12 decided ideas through it. The gate agrees with leaders on 10. One miss was an idea leaders approved that failed the gap check. When they looked again, its gap was about half what the deck claimed. The other was an idea leaders turned down that passed every gate. It’s back on the list.

Then the new quarter runs through. Of 90 ideas, 25 stop at the result check, 20 at today’s cost, 10 at the floor and 20 at the gap. Only 15 need interviews. Six pass. Leaders review those 6 at 45 minutes each. That’s 4.5 hours out of 26, and every one of the 6 gets a clear yes or no.

What you get instead of a deck

The output isn’t a slide deck. It’s a go or no-go, with the reasons and the numbers behind it. Anyone can rerun it and get the same answer. If two people disagree, they can point to the exact input they think is wrong.

That changes the conversation. Instead of “I like this one,” you get “I think the floor on step 3 is too low.” That’s a disagreement you can settle in a few minutes. Change the input, rerun the gate and look at the answer.

The counter-case

The strongest pushback goes like this. “Strategy is judgment. You can’t put judgment in a spreadsheet.”

I agree that the final call is judgment. The gate doesn’t make it. It handles the questions that have a number for an answer. What does this cost today? What’s the floor? Will anyone pay? Those don’t need a senior leader’s time. They need inputs. Leaders still make the call on the ideas that pass. They just make it on 6 ideas with the math attached, not 90 decks.

The second pushback is that rules will kill good, odd ideas. A rule can be wrong. That’s why the back-test matters, and why every stopped idea comes back with a reason. If a good idea keeps failing one gate, that’s a sign the rule needs work. Fix the rule in the open. Don’t go around it in a hallway.

The third pushback is that people will game the inputs. Some will. But a made-up number in a sheet is easier to catch than a made-up story in a deck. Every input is in plain view, and anyone can challenge it. That’s a better place to have the fight.

What to do Monday

Look back at last quarter.

  1. Count the ideas your team reviewed.

  2. Count how many got a clear yes or a clear no.

  3. Count how many hours leaders spent in those reviews.

  4. If fewer than half got a clear answer, the jam is at the decision step. That’s a rule of thumb, not a law. You don’t need more ideas. You need a filter in front of the people who decide.

  5. Write down the five questions every idea should answer before it gets a meeting.

  6. With your leaders, give each question a pass or fail rule. Set the numbers now, before anyone sees this quarter’s ideas.

  7. Put the rules in a sheet, one row per idea and one formula per gate.

  8. Run last quarter’s decided ideas through it. Look hard at every miss.

  9. Send every new idea through the gate before it gets a meeting. Give each stopped idea its reason.

  10. Check the gate once a month. Raise the bar if too many pass. Check the inputs if almost nothing does.

Where this breaks

This has limits.

First, the inputs can be weak. If today’s cost is a guess, label it as one. A gate fed by guesses gives you a confident wrong answer.

Second, the bars can drift. If more ideas pass than leaders can review, the jam comes back. Tune the rules once a month.

Third, some ideas don’t fit the math yet. A new idea may have no clear cost today. Give those a separate, small path, so they don’t vanish without a trace.

Fourth, a gate is only as good as the leaders’ willingness to use it. If people can skip it by knowing the right person, it won’t hold.

The close

Jevons saw that better engines didn’t save coal. They made people use more of it. Cheap AI will do the same to idea checks. You’ll get more ideas than ever.

That’s good news, as long as they don’t all land on the same few calendars. Put the cheap questions in code. Save the people who decide for the ideas that pass.

Get the free download

Grab the free checklist, with the playbook included. The checklist gets you started this week. The playbook walks you through the full strategy team example, gate by gate.

We build these gates in the Innovation Unpacked Mastermind.

If you want to work through your own, book a call.


Sources


Book an appointment: Click here
Email me: mike@pjtbd.com
Call me: +1 678-824-2789
Follow me on 𝕏: https://x.com/mikeboysen
Articles - jtbd.one - De-Risk Your Next Big Idea
Website - pjtbd.com

Slaying the Sacred Cows of Innovation

Discussion about this video

User's avatar

Ready for more?