There’s a lot of noise from pretty much everywhere about how AI is making work more efficient. In many aspects, this is very true. I’ve experienced it with programming—used the right way, AI can reduce development time by orders of magnitude. That’s not an exaggeration: done properly, a development process that would normally take months can be completed in days.

There’s still debugging, corrections, and refinement involved, of course, but the total time from concept to product with AI assistance can be reduced to a fraction of what it would be without AI. If I get to it, I’ll write more about that in a future blog, it is completely changing the development process.

But…

When computers first started entering the work environment, people said we wouldn’t need paper anymore. Today, we print so much that the idea sounds like a bad joke. When the first company I co-founded, VocalTec, introduced the first ever VoIP product, Internet Phone, and later added video conferencing to it, people said there would be no need to travel anymore. I spent last night on a red-eye flight, and business travel is booming. 

AI is great. It’s so great that it now takes a person five minutes to produce a 17-page document and proudly send it to ten people on their team, each of whom then spends two hours reading it and commenting on it. The person who wrote it did not go through the process of not just producing the document, but actually thinking while writing it—reviewing ideas as they formed, letting them brew overnight, rewriting sections that weren’t the best fit, reducing redundant information, or adding new ideas that came up along the way.

Instead, that thinking is pushed downstream: ten other people now repeat that process independently, wasting roughly ten times the effort and time, sending that info back to the person for more time spent sifting through the feedback.

Even in software development, if AI is used the wrong way, you can see a similar effect. When a programmer writes code while checking each element as they develop it, they eliminate a chain reaction of wasted time. If a bug is left in the code, it will be found by QA—wasting not only development time, but also QA time. QA now has to reproduce the issue, report it, document the reproduction steps, and possibly communicate back and forth with the developer for clarification. The developer then needs to return to the code, reproduce the issue, figure out what went wrong, fix it, and send it back to QA. If QA misses the bug as well, there’s yet another layer of wasted time when customers run into it. At that point, the waste multiplies—and that’s before even considering reputation and client experience. Using AI—especially when used incorrectly and without a mandatory validation process—can create a massive waste of time and resources trickling bugs, wrong features etc. up and down the programmer-QA-client-support organization. 

So don’t produce 17-page documents just to show how productive you are. Think about the chain reaction you’re creating. It might impress people the first or second time—but eventually it will click: that person is wasting our time with piles of AI-produced documents.