I’m going to say something that might make some people in EdTech uncomfortable. Your expensive, shiny new AI grading tool? It’s probably garbage.
There, I said it. I’ve spent the better part of two decades building and investing in tech companies. I’ve seen fads come and go. And right now, the way most companies are peddling AI for grading essays and creative work feels like a solution in search of a problem—a very expensive, ineffective solution.
Just last year, a founder pitched me on his “revolutionary” AI grading platform. He had all the buzzwords: neural networks, large language models, sophisticated algorithms. He claimed it could grade an English essay with 99% accuracy compared to a human teacher. I was skeptical, but I’m always open to being proven wrong. So, I gave him a challenge. I took a dozen essays from a high school English class—essays I had personally reviewed—and ran them through his system. The results were, to put it mildly, a disaster.
An essay that was a solid A, full of original thought and a unique voice, got a C-. Why? It used complex sentence structures the AI wasn’t trained on. Another essay, a B- piece that was basically a regurgitation of SparkNotes, got an A+. The AI loved its predictable structure and keyword stuffing. The system didn’t grade for insight; it graded for conformity. It was a pattern-matching machine, and a bad one at that.
This isn’t a one-off experience. I’ve tested over a dozen of these tools. The story is always the same. They miss nuance. They penalize creativity. They turn learning into a game of figuring out how to please the algorithm.
The Core of the Problem: AI's Blind Spot
So why are these tools so bad at a task that seems, on the surface, perfect for a machine? It’s because they fundamentally misunderstand what education is about.
Writing an essay isn’t just about stringing together grammatically correct sentences. It’s about developing a voice. It’s about learning to think critically, to argue a point, to tell a story. True learning is messy, unpredictable, and deeply human. And that’s exactly what current AI models are terrible at understanding.
These systems are trained on massive datasets of existing text. They learn to identify common patterns, grammatical rules, and popular arguments. What they can’t do is recognize a truly original idea. They can’t appreciate a clever turn of phrase that breaks the rules for stylistic effect. They can’t distinguish between a student who is genuinely grappling with a complex topic and one who is just good at mimicking the style of the source material.
In short, they optimize for the average. And in doing so, they punish the very students we should be encouraging: the ones who think differently, who take risks, who have a unique perspective on the world.
A Better Way: The Co-Pilot Framework
After my frustrating journey through the world of AI grading, I started to think about how we could use this technology in a way that actually helps students, instead of just making life easier for administrators. I call it the Co-Pilot Framework. The philosophy is simple: AI should be a tool for teachers, not a replacement for them. It should handle the grunt work, so teachers can focus on what they do best: mentoring, guiding, and inspiring their students.
Here’s how it works:
Step 1: The First Pass AI. Use a simple AI tool to do a first pass on student work. But here’s the key: you’re not using it to assign a grade. You’re using it to check for the basics: grammar, spelling, plagiarism, and basic structure. Think of it as a super-powered spell checker. This part is easily automated and saves a ton of time.
Step 2: The Teacher's Deep Dive. With the basics out of the way, the teacher can now focus on the important stuff: the quality of the arguments, the originality of the ideas, the student’s unique voice. They’re not bogged down in correcting comma splices. They’re having a high-level conversation with the student through their feedback.
Step 3: AI-Powered Feedback Generation. Here’s where it gets interesting. Instead of writing the same comments over and over again, the teacher can use an AI to generate personalized feedback based on their specific notes. For example, if a student is struggling with thesis statements, the teacher can simply tag the relevant section and have the AI generate a detailed explanation with examples. This is a huge time-saver, and it provides students with more targeted support.
Step 4: The Final Grade (Is Human). The final grade is always, always assigned by the teacher. It’s based on their holistic understanding of the student’s work, not on some arbitrary score generated by an algorithm.
Putting It Into Practice
I know what you’re thinking. This sounds great in theory, but what does it actually look like in a real classroom? Let me give you an example.
I worked with a history teacher who was spending hours every weekend grading essays on the American Revolution. He was burned out and his feedback was becoming generic. We set him up with the Co-Pilot Framework. He used a simple plagiarism checker for the first pass. Then, he read through the essays, focusing on the historical arguments. He used a text-expansion tool to insert his pre-written (but still personalized) comments on common issues. The result? He cut his grading time in half. His feedback was more detailed and helpful. And his students started to engage more deeply with the material.
This isn’t about some fancy, expensive software. You can implement this framework with a combination of free and low-cost tools. A Google Docs script, a basic plagiarism checker, and a text-expansion app are all you need to get started.
Stop Chasing the Wrong Metrics
I’ve made over 200 angel investments in my career, including in some of the biggest names in AI like Anthropic, OpenAI, and Scale AI. I believe in the power of this technology to change the world. But I also believe that we have a responsibility to use it wisely.
In education, that means focusing on tools that empower teachers, not replace them. It means recognizing that learning is a human process, and that there are no shortcuts to developing a curious, creative, and critical mind.
So, the next time a founder tries to sell you on a “revolutionary” AI grading tool, ask them one simple question: does it help teachers do their job better, or does it try to do their job for them? The answer to that question will tell you everything you need to know.
Let’s stop wasting money on tools that don’t work and start investing in our teachers. They’re the real key to unlocking our students’ potential. And no algorithm can ever replace that.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.