You don't have an AI problem. You have a friction problem.

The advice arriving in every leader's inbox right now sounds roughly the same: rebuild your company on AI. Put a copilot on everything. Move fast or get left behind.

So companies are spending. And most of that spending isn't showing up in any number that matters. Survey after survey lands in the same place — a large majority of AI initiatives never produce measurable ROI. The technology works. The strategy does not.

I have been building deep learning systems for nearly a decade, and the pattern is always the same. When AI disappoints, it is almost never because the models were too weak. It's because they were pointed at the wrong problem.

So I want to offer a reframe that has held up across every industry I have worked in, from government to media to enterprise software. You don't have an AI problem. You have a friction problem.

Find the friction → measure it → fix it.

The instinct, when AI is the shiny new tool, is to ask: where can we use this? That question is backwards. It starts from the solution and goes looking for a place to apply it — which is how you end up with a chatbot nobody uses, bolted onto a problem nobody had.

The better question is: where is my team stuck? Where does the work pile up? Which tasks are high-volume, repetitive, and draining your most capable people of the hours they should be spending on creative thinking, strategy, and human judgment?

That's friction. And friction is mappable.

Find the friction

Take any workflow your team runs every day and break it into its actual steps. Then score each step on two axes:

• How often and how expensive is it? (volume × time)
• How much real human judgment does it need?

Plot those two and you get four zones:

• Automate — high volume, low judgment. This is the takeout: the repetitive, rules-based work that's eating hours and burning your people out. AI is genuinely great here.
• Augment — high volume, high judgment. Keep the human in the seat, but give them leverage: draft the first pass, surface the context, do the lookup. The person still decides.
• Ignore — low volume, low judgment. Real, but rare. Automating it costs more than the friction it removes. Leave it alone.
• Redesign — low volume, high judgment. The hard, high-stakes calls. These don't get automated; they get protected — supported with better information, never replaced.

The work of an AI strategy isn't adding AI — it's figuring out how to free up your people for the interesting stuff.

Measure the friction

We have run this play before. This isn't a thought experiment. While at Atlassian, our founders spent three years applying exactly this map to customer support — a high-volume, high-friction operation.

We decomposed the workflow into steps. We found the mechanical work that AI is well suited for. And we improved the tools for the humans handling the judgment-heavy steps.

The result: $60M+ in documented savings, 92%+ CSAT, and three years running in production.

Fix the friction

I want to dwell on that CSAT number, because it's the part most "we added AI" stories quietly skip. Customer satisfaction went up. That's not a coincidence — it's the whole thesis. When you take the mechanical work off your team's plate, they stop drowning, and they spend their attention on the moments that actually need a person. Done right, friction-mapping doesn't trade quality for cost. It improves both.

Here's the reframe anyone who runs a team feels in their bones: the output of this work isn't "labor removed." It's capacity created.

Every hour your team isn't spending on rote, repetitive work is an hour redeployed to the work that grows the business — the judgment, the relationships, the creative problem-solving no model can do. You are not shrinking the team. You are aiming it at what it's actually good at.

And if you think in financial terms: that freed capacity is margin you didn't have before. The savings are real and they compound. But the reason your best people will thank you is that they got their time back.

Finding the friction before you spend a dollar is the problem we started YourCadre to solve — so we built it into a tool.

You name a company, it researches public sources, finds the single highest-volume workflow worth automating, breaks it into steps, and puts a real number on the opportunity in about two minutes — including, honestly, the steps it would leave alone. We use it to think critically about the opportunities hiding inside an org.

Try it out yourself — just name your company. Or contact us and I'll run one with you live.

What's the workflow in your org that everyone knows is broken, but no one has ever measured end-to-end? That's where I'd start.

Pri Oberoi is a co-founder and Chief AI Officer of YourCadre, where they help companies find the friction in their operations and put AI exactly where it pays off. They have spent nearly a decade building deep learning systems across government, media, and enterprise software.