When Your Mechanical Engineer's Claude Bill Beats the Entire Software Team

A tweet made the rounds recently that raised some eyebrows:

"Our mechanical engineer's Claude Code bill was higher than all of the software team combined"
— @seungho__yang

At first glance, this seems backwards. Software engineers write code all day. A mechanical engineer? Maybe some Python scripts for analysis, but surely not enough to rack up the highest AI bill in the company.

But after spending a minute thinking about it, this actually makes perfect sense. And understanding why reveals something important about how AI coding tools deliver value.

The Leverage Paradox

Here's what most people miss: the engineers who get the most value from AI coding assistants aren't necessarily the ones who are already great at coding.

Your senior software engineers have years of muscle memory. They know the syntax, the patterns, the idioms. They've written the same boilerplate a thousand times. For them, Claude Code is a nice productivity boost—maybe 20-30% faster on routine tasks.

But for a mechanical engineer who codes occasionally? AI transforms what's possible.

Instead of spending three days learning a new library to parse CAD file formats, they describe what they need and Claude writes it. Instead of googling "python pandas groupby multiple columns" for the fifteenth time, they just ask. Instead of abandoning an automation idea because the implementation would take too long, they actually build it.

The leverage isn't 20%. It's 10x.

Let's Do The Math

Say your mechanical engineer makes $150,000 per year. That's roughly $75/hour loaded cost.

If Claude Code costs them $500/month (which would be heavy usage), that's about $6,000 per year.

For that $6,000 investment, let's conservatively estimate they save 10 hours per month on coding tasks they would have struggled through or outsourced to the software team:

And that's the conservative math. The real value is often in the projects that wouldn't have happened at all. The simulation automation that would have been "too much work." The data pipeline that would have stayed a manual Excel process. The custom tool that would have required filing a ticket with engineering.

Why Software Teams Use Less

Meanwhile, your software engineers might actually be under-using AI tools. A few reasons:

• Pride: "I don't need AI to write a for loop"
• Trust issues: They've seen AI generate buggy code and decided it's not worth reviewing
• Workflow friction: Their existing setup is optimized, and adding a new tool feels slow
• Code review concerns: "What if my PR looks AI-generated?"

None of these are wrong, exactly. But they can lead to systematic under-adoption even when the tools would genuinely help.

What This Means For Your Team

If you're tracking AI spend across your organization, don't assume high usage is bad or low usage is good. The distribution might be telling you something important:

High spend from non-engineers often means they're unlocking capabilities they didn't have before. This is usually good—as long as the outputs are quality.

Low spend from engineers might mean they're not getting value from the tools, or it might mean they're being artificially conservative. Worth a conversation either way.

Uneven distribution isn't automatically a problem. Different roles, different leverage points. The mechanical engineer writing 10x more code than before has different needs than the senior dev who just wants faster autocomplete.

The Uncomfortable Question

The tweet that sparked this post got 366,000 views because it pattern-matches to "AI costs are out of control." But the more interesting question isn't "why is the mechanical engineer spending so much?"

It's "what is that mechanical engineer shipping that they couldn't have shipped before?"

Because if the answer is "custom tooling that used to require engineering tickets" or "automated analysis that used to be manual" or "prototypes that used to stay on the whiteboard"—then the spend isn't a cost center. It's an investment with visible returns.

And if your software team is spending less despite having the most obvious use case? Maybe the real question is what's holding them back.

If you're curious about your own team's AI spend patterns, MarginDash can help you see what's actually happening across providers and team members.