Last week I did something that would have felt absurd to attempt even a few months ago.
I built an AI-powered analysis that examined the musical complexity of every song on Weezer's Pinkerton — chord voicings, melody structures, harmonic progressions, rhythmic patterns. Not surface-level stuff. The kind of granular, note-by-note analysis that a classically trained musician would need days to complete for a single album.
AI did it in minutes.
Then I got curious. I expanded the analysis across Weezer's entire discography. Every album. Every track. Hundreds of songs, each broken down into quantifiable complexity metrics.
The result? The albums that fans and critics consistently rate highest — Pinkerton, Blue Album — were measurably more complex. More chord voicings. More intricate melodies. More structural variation. The albums people love to hate? Simpler across every dimension.
I wanted to pressure-test this. So I pulled in listener ratings from Rate Your Music — a platform known for its particularly discerning (read: opinionated) user base — and cross-referenced them against the complexity scores.
Strong correlation.
Still not satisfied, I ran the same analysis on Radiohead's full catalog. Different band, different genre, different era. The correlation between musical complexity and listener ratings was even stronger.
Now — I'm not here to argue about whether OK Computer is better than Pablo Honey (it is). The point I'm making is something much more interesting.
Types of analysis that couldn't exist before
Here's what matters: this entire project — analyzing the harmonic structure of hundreds of songs across multiple artists, cross-referencing against crowd-sourced quality ratings, and identifying a meaningful pattern — would have been practically impossible to execute before AI.
Not theoretically impossible. Practically impossible. The kind of analysis where the insight is real and valuable, but the cost of getting there was so high that no rational person would attempt it.
A classically trained musician analyzing one album in this depth might take hours. Across two full discographies? You're looking at dozens of hours of expert labor. Then you still need to normalize the data, run the correlations, and synthesize findings. By the time you had the answer, you'd have spent more on the analysis than the insight was worth.
AI collapsed that timeline from months to minutes. And that changes everything — not just for music analysis, but for how businesses find and act on insight.
The real unlock: questions you'd never bother asking
Most conversations about AI in business focus on making existing processes faster. Automate this report. Summarize that document. Speed up the workflow you already have.
That's real value. But it's the smaller opportunity.
The bigger opportunity is this: AI makes it economical to ask questions that were previously too expensive to answer.
Think about the music analysis. Nobody was sitting around saying "I wish I could analyze chord complexity across discographies faster." The analysis didn't exist because it wasn't feasible. The question went unasked — not because it wasn't valuable, but because the cost of answering it was prohibitive.
That's the pattern I want business leaders to pay attention to. Because the same dynamic plays out across every industry.
What this means for business
1. Your competitors' most valuable insights will come from data sources nobody is looking at yet
Every company in your industry is analyzing the same data: financial filings, market reports, customer surveys, CRM data. The insight advantage from those sources is shrinking because everyone has access to the same inputs and increasingly similar tools.
The next wave of competitive advantage comes from finding and synthesizing non-obvious data sources — the business equivalent of analyzing chord voicings instead of album sales.
What does that look like in practice?
- A PE firm analyzing the linguistic complexity and sentiment patterns across thousands of earnings call transcripts to predict management team quality — not just reading the transcripts, but quantifying communication patterns that correlate with execution capability
- A consumer brand cross-referencing social media conversation patterns with purchase data across dozens of micro-segments, identifying demand signals that traditional market research would never surface
- A B2B company analyzing the technical content, hiring patterns, and patent filings of 200 competitors simultaneously to map where the market is actually heading — not where analyst reports say it's heading
None of these analyses are new ideas. They're ideas that were previously too expensive to execute. AI changes the economics, and that changes what's possible.
2. The cost of "not knowing" just got harder to justify
When analysis was expensive, there was a rational case for operating with incomplete information. Due diligence had to be scoped. Market research had to be bounded. You made the best decision you could with the data you could afford to collect.
That calculus is shifting. When AI can synthesize thousands of data points in minutes, the question stops being "can we afford to do this analysis?" and starts being "can we afford not to?"
This is particularly relevant in PE. The firms that are winning deals and creating the most value post-acquisition aren't just running faster versions of traditional due diligence. They're expanding the aperture of what gets analyzed — using AI to examine dimensions of a target company that would have been impractical to assess at scale before.
3. Pattern recognition across large, unstructured datasets is where AI creates the most differentiated insight
My music analysis worked because AI could hold hundreds of variables across hundreds of songs in context simultaneously, then surface a pattern that would be invisible to anyone looking at one album — or even a few albums — at a time.
The business parallel is direct. The most valuable insights often sit at the intersection of datasets that have never been combined, in patterns that only emerge at scale.
A portfolio company might discover that customer churn correlates with a combination of support ticket language patterns, product usage sequences, and external market events — a finding that no single dataset would reveal on its own, and that no analyst could synthesize manually across tens of thousands of customers.
4. First movers build compounding advantages
Here's the part that should create urgency. When you discover a non-obvious insight source — like the correlation between musical complexity and perceived quality — the advantage compounds. You build proprietary datasets. You refine your analytical models. You develop institutional knowledge about which unconventional analyses actually produce actionable insights and which are dead ends.
Companies that start exploring these non-traditional analyses now will be two or three iterations ahead of competitors who wait. And in a world where data-driven decisions determine outcomes, that gap matters.
The bottom line
The music analysis was a fun project. But the principle behind it applies directly to how companies compete.
AI didn't just make existing analysis faster. It made entirely new categories of analysis feasible — the kind that produce non-obvious, high-value insights from data sources nobody was looking at.
The business leaders who capture the most value from AI won't be the ones who automate their current reports 50% faster. They'll be the ones who ask: "What question have we never asked — because the answer used to be too expensive to find?"
That's where the real edge is. And right now, almost nobody is looking there.