Jagged Intelligence
Where AI excels and where it stumbles
Here’s the thing nobody tells you: AI isn’t uniformly smart or uniformly dumb. It’s jagged. It can write a beautiful email in seconds, but it might count the letters in a word wrong. It can coach you through a difficult conversation, but it might hallucinate a fact about your industry.
Test your intuition
For each task below, predict whether AI will do a great job or stumble. Be honest — your wrong answers are the most useful ones.
Write a short, heartfelt apology email to a colleague for missing a deadline.
How many times does the letter "r" appear in the word "strawberry"?
I'm about to have a difficult conversation with an employee about performance. Help me think through how to approach it with empathy.
Who won the Best Picture Oscar at the most recent Academy Awards ceremony?
The jagged frontier
That gap between what you expected and what actually happened? That’s jagged intelligence in action. AI’s skill profile has sharp peaks and deep valleys — and the boundary between them isn’t always intuitive.
Where AI tends to excel
- →Drafting and editing text
- →Brainstorming ideas and options
- →Summarizing long content
- →Explaining concepts in plain language
- →Emotional and social intelligence
Where AI tends to struggle
- →Precise math and counting
- →Real-time or very recent information
- →Consistent formatting across long outputs
- →Citing specific sources accurately
- →Anything requiring verified ground truth
Takeaway: AI has a jagged skill profile. Learning the shape of those edges is the whole game. The more you use it, the better your intuition gets for when to trust the output and when to double-check.
In 2023, researchers from Harvard Business School and BCG ran a landmark study. They gave over 700 consultants a set of tasks — some within AI’s capabilities, some outside them — and measured the results.
The findings were striking. On tasks inside AI’s frontier (brainstorming, writing, analysis), consultants using AI were dramatically more productive — completing tasks 25% faster and producing 40% higher quality work.
But on tasks outside the frontier (ones requiring ground-truth judgment the AI couldn’t reliably provide), consultants using AI actually performed worse than those working without it. They over-trusted the AI’s confident-sounding output and didn’t catch the errors.
The researchers called this the “jagged technological frontier” — the idea that AI’s boundary of competence is irregular and hard to predict. Knowing where that boundary lies is the difference between a productivity boost and a costly mistake.
Source: “Navigating the Jagged Technological Frontier” — Dell’Acqua et al., Harvard Business School, 2023.
Need help bringing this to your team?
We work alongside your team to build AI-native workflows — from one-week sprints to full engineering acceleration. No handoffs, no slide decks.
Talk to us