With a little help from my bots

Conventional wisdom suggests that using AI tools makes us more productive. Think about it: you've got this bot that can do stuff for you, whether that's writing code, conducting research or editing this commentary. In theory, that should save time and reduce the administrative busy work most people don't enjoy. That sounds like a good thing. But according to a recent Harvard Business Review article, AI doesn’t actually reduce work — it intensifies it.

"One of the promises of AI is that it can reduce workloads so employees can focus more on higher-value and more engaging tasks. But according to new research, AI tools don’t reduce work, they consistently intensify it," the authors of the HBR article wrote. 

While the HBR article points out the long-term ramifications of intensifying our workloads, I’ve found the impacts to be more immediate. Consider my experience the other day. I had to create an invoice to get reimbursed for airfare. I uploaded my logo, the receipt and the name and address of the company paying me. I then instructed ChatGPT to write an invoice using these three elements and generate a PDF. As Bruce Springsteen once wrote, "ought to be easy, ought to be simple enough," but it wasn't.

In spite of clear instructions, and the fact it had created a PDF for me a few days before, ChatGPT couldn't create a downloadable PDF and claimed it didn't know how. I tinkered with it for 20 minutes. In that time I could have easily opened Google Docs and done it myself, but I was determined to make it work. I finally dropped the same instructions into Claude and it produced a professional-looking invoice on the first try. Yeah, just a little bit of work intensification to do a simple task.

But what about programmers?

The strongest AI use case so far may be programming, where a code generator like Claude Code or Cursor can create a program for you from a plain-language prompt. You may have heard about a viral blog post this week from a programmer named Matt Shumer who claimed that using AI programming tools, applications practically write themselves. Gary Marcus, a guy like me who tends to have a skeptical view when it comes to the limits of today's tooling, wasn't convinced it was quite that easy. 

I tend to lean more toward Marcus' skepticism than Shumer's hype.

If you read my Adventures in AI section in this newsletter and on LinkedIn (and if you don't, you should), you know that AI is more complex than 'it just works.' It sometimes feels magical, sometimes obstinate and frustrating. Case in point: while fact-checking this very piece, ChatGPT confidently told me the Bruce Springsteen lyric I quoted earlier was from Hungry Heart. It is actually from Tunnel of Love. Sure, you could let it go and produce stuff, but that is probably going to come back to bite you sooner rather than later.

Does it help me move faster in some ways? It absolutely does, but that doesn't mean that you can set it and forget it. If you use it every day and you're being honest, you know that. As Jim Rowan, US head of AI at Deloitte, told me this week, using research as an example, it helps you get there, but it doesn't produce a finished product on the first try. You still need human intervention. "It helps you do the research. It’s not the research product,” he said. 

I've found that AI is a force multiplier for me, especially as a solo publisher. I can do research faster. I can pressure test ideas. I can get advice on word choice, sentence construction, accuracy, clarity and flow. I would be lying if I said it hasn't changed the way I work for the better.

But that doesn't mean I think we are all about to be replaced by our AI overlords. Tech people love to overstate what tech can do, or how quickly we can go from point A to point B in terms of usage and capability. This is unquestionably a big leap forward in terms of how we get work done, but ultimately I'm getting by with a little help from my bots. They don't do everything.

~Ron