You're Reading the Wrong Curve

Why your AI investments look disappointing when you're doing everything right

Your board is asking why productivity isn't improving despite massive AI investment. Your CFO wants to see ROI. Your competitor just announced they automated 30% of their call center and you look behind.

Here's what Erik Brynjolfsson told me at Stanford that should change how you think about all of this: You're being evaluated using metrics designed for the wrong transformation, at the wrong point in the curve.

And that misreading is about to cost you everything.

The Pattern You're Inside

Erik has studied technology adoption for three decades. When electricity came to factories, productivity didn't jump. It declined for 20 years. Companies spent resources rewiring, retraining, experimenting. Massive inputs, flat outputs. Lower productivity.

Then they stopped just replacing steam engines with electric motors. They redesigned the entire factory around what electricity made possible. Productivity tripled.

The graph looks like a J. Down first, then vertical.

We're in that trough with AI. You're investing heavily, the capabilities are exploding, and the numbers aren't moving. Which makes it look like you're wasting money.

But that decline is what doing it right looks like.

The companies that quit during the trough - the ones that just automated a few tasks and called it done - those are the ones who actually wasted their investment. They got a small bump and stopped growing.

The ones who endured the J-curve? They got the exponential.

The Choice You're Making Without Realizing It

Erik's latest research tracked companies using AI for automation versus augmentation. The data is striking.

• Automation path: Cut costs, reduce headcount, boost short-term margins. Clean ROI this quarter. Looks great on your dashboard.

• Augmentation path: Redesign jobs to make humans more capable. Rising employment, higher productivity, better customer outcomes, lower turnover. Genuinely more profitable.

But augmentation requires living through a longer, deeper J-curve. You're not just implementing a tool. You're reinventing the entire organization. And every metric your board cares about punishes you for it.

So most companies choose automation. Not because it's better. Because it's defensible now.

Here's what that choice actually determines:

• In five years, do you have a workforce that's 20% smaller and marginally more efficient?

• Or do you have a workforce that can do things your competitors literally cannot do?

One path gives you cost savings. The other gives you a sustainable competitive advantage.

The companies stuck in the trough are the ones treating AI like previous IT rollouts (delegate it to IT). Define requirements, implement solution, measure efficiency gains. That worked for ERP systems. It will fail catastrophically for AI.

What false transformation looks like:

• Pilot programs that never scale

• AI pasted onto existing workflows

• Success measured purely in cost reduction

• echnology team owns it, business leaders wait for results

What genuine transformation looks like:

• Continuous experimentation across the organization

• Jobs redesigned around new capabilities

• Success measured in new things you can do, not just old things done cheaper

• Every leader treating this as their problem to solve

  The uncomfortable truth: Genuine transformation shows up first as chaos, exhaustion, and flat productivity numbers. If everything looks clean and ROI is immediate, you're probably just automating around the edges.

The Country of Geniuses

But here's what makes this moment different from electricity or steam engines: what's actually coming.

Dario Amodei, founder of Anthropic, has this thought experiment that Erik keeps coming back to. Imagine you could create a country of geniuses - Nobel Prize-level minds in physics, chemistry, economics, strategy, logistics, every domain that matters to your business. Now imagine making millions of instances of each one. And running them at 100x human speed.

That's not science fiction. That's the technical trajectory we're on. Maybe two years out, maybe five. But coming.

Stop and actually picture that for a second. A marketing genius who's read every campaign ever run, understands behavioral psychology at a PhD level, and can generate and test a thousand variations before lunch. A strategy genius who can war-game every competitive scenario simultaneously, accounting for variables you didn't know existed. A logistics genius who sees patterns across your entire supply chain that no human team could ever spot.

And they're all talking to each other. In real-time. Learning from each interaction.

That's not "10% more efficient at your current processes." That's a different game entirely.

This is why the automation versus augmentation choice matters so much. If you spend the next two years teaching AI to do your current jobs slightly cheaper, you'll have marginally lower costs when this hits. If you spend it learning how to combine human creativity with AI capabilities, building the organizational muscle to ask the right questions, to verify and iterate, to imagine possibilities that didn't exist before – then you'll be positioned to do things no one else can do.

What You Should Do Monday

Here's what I've been thinking about since talking to Erik: What geniuses do I actually need around my work?

Not "what tasks can I automate." What capabilities would transform what's possible?

Maybe you need a genius-level customer psychologist who can analyze every interaction pattern and help your team understand what customers actually want before they know themselves. Maybe you need a regulatory expert who can navigate complexity across fifty jurisdictions simultaneously. Maybe you need a creative director who can help your team explore a hundred directions for every brief.

The tools to start building this exist today. Platforms like n8n or Workato let you orchestrate multiple AI agents, give them specialized knowledge, let them collaborate with each other and with your team. You can prototype your country of geniuses this quarter.

Not to replace your people. To give them capabilities they've never had.

Start with one role. Build out the genius you wish you had on staff. Give it agency to bounce ideas around with your team. See what becomes possible when your smartest people have genius-level thought partners in domains where they're not experts.

That's what genuine augmentation looks like. Not incremental efficiency gains. Fundamental expansion of what your organization can do.

The J-curve is real. The trough is uncomfortable. Your metrics will punish you while you're in it.

But the exponential is coming regardless. The only question is whether you're building toward it or optimizing for looking good while it arrives.

The curve will turn. Make sure you're positioned for it.

This article comes from a podcast with Erik Brynjolfsson, check it out here in full: