Superconsumer AI Agents: Turn Your Greatest Customers Into AI Agents So You Never Go Astray

Without Superconsumer data, AI will automate mediocrity at scale.

Arrrrr! 🏴‍☠️ Welcome to a 🔒 subscriber-only edition 🔒 of Category Pirates. Each week, we share radically different ideas to help you design new and different categories. For more: Dive into an audiobook, listen to a category design jam session, or upgrade to a Founding subscription to ask the Pirate Eddie Bot your category design questions.

---

Dear Friend, Subscriber, and Category Pirate,

When Harley-Davidson announced in 2020 that Jochen Zeitz (a man with no motorcycle background) would take the helm as CEO, there was backlash.

What did he know about American culture, or Harley? (One of the most cult-like brands in the world). Nothing. He was the CEO of Puma in Germany.

Under Zeitz:

• People publicly destroyed their Harley bikes

• Influencers shared their outrage on social media

• Riders boycotted biker gatherings, like the infamous Sturgis rally

Longtime Harley Superconsumers—the riders with the eagle patches, road-worn leather, and bar-and-shield tattoos—were already uneasy. The pandemic had closed assembly lines and dealerships. Sales sagged.

Now, the company’s most sacred seat was going to someone without a motorcycle license.

Imagine running one of the most storied motorcycle brands without a visceral, lived connection to what it feels like to lean into a curve, throttle open, wind tearing at your jacket.

It was a deep category disconnect.

Zeitz brought a track record in shoe fashion and sustainability, not oil-stained knuckles or the thrill of leaning a V-twin into a sweeping curve.

He made moves that looked, to the Harley faithful, like heresy:

• Canceling bike projects they’d been waiting on

• Putting sustainability and DEI initiatives front and center

• A strategy aimed at wealthier, global buyers instead of the core American rider

Some called it “woke.” Others called it “corporate suicide.”

The reaction revealed something deeper than politics or product roadmaps.

Harley had drifted from its Superconsumers, the hardcore, lifestyle-committed riders who are custodians of the company’s soul.

For decades, Harley had been a master at selling more than motorcycles. It sold belonging, status, and a shared language of chrome and rumble. You didn’t just buy a Harley—you joined the club. That bond is why thousands of riders have Harley-Davidson tattoos. It’s also why any move that betrays that bond sets off alarms loud enough to rattle the Milwaukee headquarters.

Now, let’s compare Zeitz to a different executive.

When Gibson named Cesar Gueikian as CEO, musician Slash nailed the reaction:

“I can’t think of a more natural fit for CEO of Gibson than Cesar Gueikian. He’s one of, if not the smartest and most passionate Gibson guitar enthusiasts I’ve ever met… and he is a keen businessman; you can’t go wrong.”

Why does this matter?

Because Gueikian isn’t just running Gibson.

He’s a Superconsumer of guitars.

He lives the product. He bleeds the category. He knows what makes a Les Paul magic, what turns a guitar into a religion, what players obsess over. No one needs to explain why Gibson is a cult brand to Mr. Gueikian.

Because Cesar is already in the cult.

Now, contrast that with another outsider: Terry Semel, the CEO of Yahoo from 2001–2007.

The $5 Trillion Category Blindspot

In 2001, Yahoo was the internet. Then, they hired Terry Semel.

Terry was a Hollywood exec in Silicon Valley. Not a search guy. Not a Superconsumer. And because he didn’t get the behaviors driving digital media, Yahoo missed Google and Facebook.

Terry made a $5 trillion mistake.

• Google in 2002

• Facebook in 2006

Both deals were on the table, both ignored.

Imagine being the CEO who could have purchased both Google and Facebook but passed.

To lead a category, you have to love the category.

Not manage it.

Not study it.

Love it.

Be a Superconsumer of it.

Rarely can people groomed in the status quo see a different future. Never mind category design the markets of the future. And rarely can people who are not Superconsumers of the category lead the category.

Today:

• Google = $3T

• Meta = $2T

• Yahoo = $5B

That’s a $4,995,000,000,000 mistake. Why? Because Semel wasn’t a Superconsumer. And if you don’t know (or worse, aren’t the Superconsumer of your category), you cannot see the future.

The law is simple:

Superconsumers pull categories forward. Leaders who ignore them get left behind.

Lose your Superconsumers, lose your category.

Zeitz wasn’t the first leader to underestimate this truth, and he won’t be the last.

Trust erodes when leadership isn’t rooted in the lived experience of its best customers. Or when a company’s strategy starts chasing “new markets” without carrying its Supers along. Sales follow.

Harley’s revenue: Down 11% to $5.2B
Motorcycle shipments: Down 17%

Harley’s stock got crushed.

Superconsumers are not a “nice to have.” You need Supers in your:

• Customer base

• CRM and data analysis

• Sales training and activation

• Marketing and lightning strikes

• In your C-suite, board of directors, and investor base

It’s the difference between creating category-defining loyalty and engineering your own backlash.

To be fair to Harley, it is hard enough to find a qualified CEO (much less a CEO who is also a Superconsumer). But great CEOs spend more than half their time with customers. Legendary CEOs have their Supers on speed dial.

Thanks to AI, a CEO can talk with Superconsumers every day.

AI gives you the ability to capture your Supers’ voice, instincts, desires, behaviors, and emotional drivers like never before.

Every conversation, data point, transaction, and social media post from a Super can be uploaded as training material to AI. Sadly, too much of the hype around AI and consumers and market research is not centered on Supers. Companies are feeding AI the wrong inputs of average customers and generic market research.

Do this, and you won’t just get bland ideas.

You’ll get the Obvious Onslaught.

Soon, the market will be flooded with AI “insights” trained on generic internet scraping, unsegmented customer inputs, and average survey data.

This Obvious AI will always default to the safe and stale:

• “Drop the price.”

• “Make the logo bigger.”

• “Copy what’s trending.”

It’s not malicious. The machine just doesn’t know any better.

Which is why the companies that win in the AI era will be the ones who bottle their Supers first. They’ll feed their AI the instincts, preferences, quirks, and Non-Obvious insights of the people who spend the most, evangelize the loudest, and shape the future of the category.

If you’re in the C-suite (or aspiring to get there) and get hit with the Obvious Onslaught, your career is at risk of sinking to the bottom of the ocean.

Harley showed how costly that can be.

While AI can help you avoid the Harley Davidson problem, be careful to dodge the Obvious Onslaught with Obvious AI. The best way to avoid it is to train your AI with Supers, not average customers. However, this is harder than you might think due to two sad truths.

• Too many CEOs barely spend any time with Superconsumers

• Too few AI advisors get why Superconsumers matter

Let’s dive into both.

The First Problem: CEOs Barely Spend Any Time With Supers

Most executives don’t know their Superconsumers.

This is especially true of large companies. A Harvard Business School study of 27 CEOs who ran large companies with an average revenue of $13 billion found the CEOs worked an average of 62.5 hours per week.

Here’s how they spend their time:

• 46% of time with direct reports

• 32% with other internal senior leaders

• 13% with business partners

• 3% with customers

• 6% other

3% with customers.

3% with customers.

3% with customers.

3% with customers.

3% with customers.

3% with customers.

🤯

CEOs spend 78% of their time talking to their people. And they spend more time with consultants than with customers. Let’s assume that Superconsumers are 10% of total customers and CEOs aren’t prioritizing them. This means that CEOs spend 0.3% of their time with Supers.

That’s 11 minutes per week.

Or 1.6 minutes per day.

Let us be clear: That is mental.

Per Grok, the average person spends 8-14 minutes per day sitting on the toilet (for an average of 12 minutes per day). This means the average $13 billion CEO spends 8x more time per day 💩ing than with Superconsumers.

That’s more than mental.

It’s disdain.

If you ever needed (more) proof that CEOs don’t give a damn about customers, there it is.

One of the big reasons why incumbents eventually fall to startups is that startup CEOs spend exponentially more time with their customers. For example, the Delphi.ai CEO sent an email asking if customers wanted to talk for 15 minutes. Pirate Eddie took him up on it. The CEO asked what seemed like scripted questions in a more perfunctory process. Maybe Pirate Eddie was boring. Maybe the CEO was told to follow a script. But we applauded him for the effort.

More CEOs should do the same.

Based on 30 years of experience, our guesstimate is that successful enterprise startup CEOs spend 40-60% of their time with customers.

Many of the most legendary CEOs are themselves Superconsumers of the category, who started their own companies to solve their own problems.

• Heidi Zak was frustrated with ill-fitting bras.

• Apoorva Mehta was tired of wasting time on grocery runs.

• Justin Gold needed fuel for long bike rides.

These founders created ThirdLove ($750MM valuation), Instacart ($13B market cap), and Justin’s ($280MM in sales), respectively.

No professional CEO can keep up with a Superconsumer CEO solving their own existential problem. A 3% CEO will always have their posterior pummeled by those with a 100% obsession with customers. Because a 100% CEO focuses on customers’ problems. Their opportunities. Their needs.

(FYI, we can not believe that we have to explain to CEOs that they must spend time with customers.)

Some companies are run by missionaries with vision, and many companies are run by mercenaries with spreadsheets.

If you’re not a Super of your category, you’re working off layers of interpretation (agencies, research firms, survey data). All of these sand down the edges and turn real human insight into bland averages.

But that’s only one problem.

The Second Problem: Too Few AI Advisors Get Why Superconsumers Matter

A dominant mindset within AI circles is “the more data, the better.”

There is truth to this, of course, as computation scales exponentially. And having a clear north star and strong reinforcement learning from human feedback greatly improves the quality of the output.

But “garbage in, garbage out” still applies to AI.

AI can’t completely solve a missing Superconsumer mindset among CEOs. AI can’t convert a mercenary CEO into a missionary. But AI consumers are coming, which will dramatically reduce the barriers to spending time with consumers.

(Over time, much buying and selling will be done by AI agents, not humans.)

However, the niche of AI consumers is also falling prey to the Obvious Onslaught.

Because most AI advisors don’t get the following:

• Supers want different benefits, but average consumers dislike change.

• Supers love to buy the category, but average consumers buy on autopilot.

• Supers are happy to pay more for more outcomes, but average consumers want to pay less.

Average consumers want you not to rock the boat, to dumb everything down, and to lower the price. But Superconsumers want you to lead them in creating exponential outcomes via different benefits, which they are happy to pay a higher price for. Yet, too many companies are uploading more and more average customer input with the hopes that it improves the quality of AI consumers.

Even McKinsey, one of the most famous consulting firms, is running headfirst into this problem.

The Synthetic Average Consumer Curse

For a century, companies paid McKinsey millions for its human expertise.

Elite consultants parachute in.

Synthesize complexity.

Map out strategy.

And deliver those famous, polished decks. Now, AI can do a “good enough” version of that in minutes. McKinsey has already deployed more than 12,000 AI agents to assist with everything from research summaries to writing in the official “McKinsey tone of voice.” The agents sound like real consultants, speak in coherent sentences, and can answer almost any question you ask.

(Remember: AI is a lot about existing knowledge regurgitation. That’s why legacy consultants, teachers, and knowledge workers are in deep trouble.)

As one of McKinsey’s senior partners admitted, the tech gets you to a pretty good, average answer.

Mediocre expertise is being automated out of existence, which means two things happen:

1. The illusion problem – If you don’t know better, an AI-generated deck or dataset looks professional, persuasive, and “right.” But without the lived context of actual experts (the kind who have been in the trenches and seen the movie before), it’s just plausible mediocrity at scale.

2. The value inversion – Distinctive expertise, built from deep pattern recognition and years of immersion, becomes more valuable than ever.

That’s the same risk with synthetic consumers. If you train your AI on average survey data, average buying data, random internet scraping, or unsegmented customer inputs, it will give you answers that feel complete but reflect the middle of the bell curve.

Feed AI mediocrity, and you will suck at scale.

It will recommend product changes your Supers will hate.

It will green-light campaigns that feel “off POV” to your most loyal customers.

It will suggest price moves that alienate the people most willing to pay more.

It will give you more radically obvious answers.

You’ll get Harley-level backlash (only faster) because AI lets you scale mistakes. An AI trained on the middle (obvious) will give you more obvious. An AI trained on your Superconsumers will give you the (non-obvious) advantage.

This is why the companies that win in the AI era will bottle their Supers first.

They’ll feed AI the instincts, preferences, and non-obvious insights of the people who spend the most, evangelize the loudest, and shape the category’s future.

You can create Non-Obvious connections about your Superconsumers by training AI.

The voice of the Superconsumer used to be a luxury that was only available to:

1. Founders who were Supers themselves – Steve Jobs who wanted simple and elegant technology, Jack O’Neill who wanted to surf longer and invented the wetsuit, or Justin Gold who created Justin’s Nut Butters so he could carry almond butter on long bike rides.

2. Companies with tiny customer sets – B2B firms selling to fewer than 20 accounts, where you must know your buyers inside and out.

3. Cultures built on fanatical customer focus – Costco, with its $1.50 hot dog and soda combo, is essentially saying to Supers, “Bulk value pricing is our brand. Our promise to you is carefully curated, high-quality products at a great price if you buy in bulk. And our $1.50 hot dog is our signal to you that we are keeping our promise. We’ll lose $80M a year before we raise this price.”

If you weren’t in one of those three categories, you settled for what most companies still do today.

Layers of interpretation through agencies, research firms, employees, and survey data. All of it turning sharp insights into safe, statistical averages.

That premise is over.

Customers are people first, data second.

Done right, AI lets you have your best customers on call 24/7, ready to pressure-test your ideas, reveal the hidden “why” behind their purchases, and make Non-Obvious connections no focus group ever could.

You can:

• Pressure-test your next big move before you spend a dollar.

• Translate gut instinct into a repeatable system your whole team can use.

• Spot the non-obvious growth levers your competitors will never see coming.

• Avoid category destroyers, like price changes or campaigns that alienate your Supers.

Get the inputs right, and AI becomes a category-creation machine. Get them wrong, and you’ve just built a faster, more confident way to be average.

This is market intimacy, scaled.

AI can simulate the same intimacy any founder, niche player, or Costco-level culture has with its Supers, if you build it from real Superconsumer data.

Unstructured data matters a lot here too.

Imagine if everyone in the C-suite, everyone in sales, support, and all customer-facing functions dumped their notes into AI. “I heard a fascinating thing today from one of our top customers and wanted to make sure we all benefit from the insight…” is a legendary way to start an AI prompt.

Technologies like AI-powered “conversational intelligence” that capture (and learn from) each and every customer interaction create a flywheel of learning what works, what doesn’t, and what customers have to say.

You want to capture as much data as possible.

Simple things like entering transcripts from Zoom sessions with customers and dumping them into your AI yield compounding results over time.

Said a different way.

With every customer interaction you don’t capture in AI, you get more stupid.

Agents Beat AI Every Time

We’ve noticed a few things on LinkedIn. We wonder if you have seen the same?

• Lots of folks are adding “AI” to their LinkedIn profiles

• Nothing on their LinkedIn has anything to do with AI two years prior

AI should not be an adjective. AI should be a noun. More specifically, it should be a proper noun, also known as an agent.

Execs want AI strategies.

Boards want AI decks.

Advisors want to sound smart about AI.

The problem is LLMs are incredible at giving you “the obvious answer.” LLMs are probabilistically choosing the right next words based on the broadest audience possible.

That’s a big problem for Pirates who reject the premise.

The obvious answer is the enemy of category creation. It’s copycat thinking at scale. “Drop the price.” “Make the logo bigger.” “Do what everyone else is already doing.”

Monkey See, Monkey Do is the unfortunate outcome of AI without proper training and context.

But check out what Pirate Jay said about the Pirate Eddie Bot:

Now, check out what Pirate Loris said:

Is the Pirate Eddie Bot helping people get these outcomes because the Delphi.ai model is superior to OpenAI’s model?

No.

The Pirate Eddie Bot drives outcomes because it is an Agent trained to be an AI librarian with a massive context window of our extensive library of content. It has a clear role (AI librarian) with a clear use case (jamming partner in Category Design).

Said a different way: Pirate Eddie AI Bot is not open-ended like ChatGPT or Claude. It’s purpose-built. Focused.

AI Agents are specialists.

They’re scoped to deliver specific outcomes and nothing else. Think of them like your own Navy SEAL unit. Small, trained for a mission, and ruthless about execution.

One Agent might be trained on your weirdest customer data to spot non-obvious growth levers. Another might be a financial “Size of Prize” Agent that keeps everyone’s eyes locked on the BHAG instead of shiny-object nonsense. Another might be looking for how to convert customer problems and complaints into new products and services.

(Remember: Problems are what drive demand for solutions.)

The power of Agents isn’t that they’re smarter than generic AI. They’re not. It’s that they’re structured—and designed to solve the actual problem, not just entertain you with plausible sentences. They take your organizational Supers’ instincts, codify them, and then clone them so the whole team can operate at a higher level. Instead of bottlenecking brilliance in one person’s head, Agents democratize it across the org.

Agents don’t try to master everything.

They just master the things that matter.

In 2024, we tried to build our own AI librarian through a Caltech dude we paid $20K. It didn’t work. In the first half of 2025, we launched our first AI Agent—the Pirate Eddie Bot through Delphi.

Big win.

Because we deployed AI tech (Delphi) to build niche down (tight-use-case) functionality.

In the second half of 2025, we will be launching a series of AI Agents to drive outcomes.

Our next book, Lightning Strike Marketing, is coming out soon. We’re launching a team of Lightning Strike Agents simultaneously. Because a book is Intellectual Capital that can be converted from passive words on a page to dynamic agents that help readers (actually) do the stuff, while they are learning. Because doing is the best way to learn and drive outcomes at the same time.

The Lightning Strike Agent is a collection of just under 10 agents, all designed to help you strike.

(Agentic learning is the future of business books and education books.)

Category Pirates is an AI-first business. We are going all in on Agents.

And you should, too.

Side note: If we’d been trapped in status quo thinking about Category Pirates, we’d have stayed (just) a business writing band. But you see, we work hard to drink our own bourbon. So we expanded from writing to digital teaching to AI to agents. In less than two years, we’ve gone from a writing-first business to an AI-first business. Because we’re obsessed with the problem our Supers grapple with every day. How to design and dominate the markets of the future.

We work at being open to new and different ways to make a difference, using state-of-the-art tech.

This is foundational to category design thinking.

We’re not anchored to the past. We’re creating (our own) different future. In public.

And that is exactly what we want you to do.

How To Build Your AI Superconsumer Agent

If you want AI to think like a Super, you have to train it like one.

Specifically, you must create Superconsumer Agents (plural).

The companies that figure this out first will compound their advantage. Because the more you use your AI Superconsumer Agents, the sharper they get. The sharper it gets, the more you use it. And the more you use it, the faster you outpace everyone still feeding AI in the middle of the bell curve.

There are 18 steps to train your AI Superconsumer Agents, and we’ve grouped them into three phases:

• Phase 1 (Steps 1-4): Train Your AI Superconsumer Agents To Be Wizards Of Weird Data & Stories

• Phase 2 (Steps 5-11): Build A Size Of Prize Agent To Size New Category Potential

• Phase 3 (Steps 12-18): Unleash Multiple AI Superconsumer Agents As Cross-Functional Strength Coaches

Phase 1 helps you think like a Superconsumer, even if you aren’t one.

Phase 2 is where AI Superconsumer Agents truly come to life, as AI can size up the prize with fewer preconceived notions and biases that humans have.

Phase 3 is where the organization becomes exponentially powerful, with marketing, sales, finance, R&D/product, operations, and HR all having their own AI Superconsumer Agents.

Phase 1: Train Your AI Superconsumer Agent to Be a Wizard of Weird Data and Stories

The goal of this phase is to understand the mindset of your Supers so you can use it to train your AI Superconsumer Agent.

Here’s how: