Just a tool

Sorry to interrupt your programme to share this excellent post by Jeff Johnson: The obligatory A.I. post. Not only do I basically agree with everything Johnson writes — I could quote every sentence and paragraph — but it’s very well written too.

Have you read it yet? Great, let’s continue.

I’ve been thinking a lot about A.I. too, obviously, using it a lot at work for various (and obvious) reasons that I won’t get into here. What particularly resonates with me in Johnson’s post is the fact that, indeed, we don’t really see great use cases for A.I., at least not in the usual and expected sense of accomplishment, success, and transformative benefits for everyone involved.

Johnson writes:

…as a consumer of software I’ve seen no indication that A.I. has benefited the industry. So many software developers claim that A.I. makes them so much more productive, but I honestly haven't noticed or enjoyed the results of this alleged productivity.

What we see contains a lot of examples, demos, and extensive details of workflow configurations. In these superficial use cases — let’s not kid ourselves, these are mostly shared on LinkedIn — we are told what can now be done with various agents, with Claude Code, or with whatever is being used and/or promoted by users.

What we rarely see are the outcomes of said workflows and automations: what are the objectives beyond “it works”? What is really achieved? What are the benefits of this so-called efficiency gain? Does it make better products for your customers? Does it improve the lives of employees? Does it generate more income for your company?

When it comes to A.I. and its implementation, we hear a lot of fanfare, but we don’t really ever get to listen to the music. Or maybe the fanfare is so loud that we’ve all become deaf or aren’t able to tell what music is anymore? Perhaps I'm missing something because all I see are the implementation steps of A.I. (and massive layoff justifications), claims of immediate benefits, but I never hear about long-term objectives that are actually reached.

Another thought I’ve had — recently triggered by Linus Torvalds’ email on the matter — revolves around the fact that A.I. is presented as “a tool, just like other tools we use.” For me, A.I. can indeed be considered as a tool, just like the car could be considered as a faster horse.

In the twentieth century alone, the car reshaped how cities were built, generated tremendous amounts of air pollution, contributed to numerous geopolitical crises and wars over petroleum, and degraded soil permeability like no horse — regardless of its speed and power — ever could. (At least the car didn’t need to steal and digest all the world’s knowledge to get created in the first place.)

A.I. is expected to have a similar worldwide impact: we’re told repeatedly that it will change everything, and this ongoing A.I. revolution will take merely years, not decades. So can it still be described as “just a tool”?

I guess from an individual point of view, like a developer working on an app, A.I. is indeed just another tool. But this feels like a rather limited view of how the world works, doesn’t it? To me, it sounds like someone parking their car on the pavement and justifying it by saying “It will only take a minute!” If every driver in the city thinks the same, there won’t be any pavement left to walk on soon.