234. If Writing is Thinking…

…then what happens if AI is doing the writing and reading?

Something I worry about with generative AI in business and commercial use: almost no one fully reads anything in those environments.

Now imagine when even the author hasn't read what was written... yikes. How does AI writing and reading impact this reality?

I used to write long memos—significant ones—maybe once a year. I'd send them to thousands. That scale alone signals, "someone else will read it." I hoped direct reports and close colleagues would read them. I could count on 2 or 3 people to definitely read them.

Bill would read. Steve would read—but only if we discussed it in person, because that's how he worked.

Just an old memo featured in Hardcore Software.

I knew this, so I always made a slide version. I'd use it in dozens of team meetings. But even then, for months after sending a memo, I'd be referring members of the team back to what was in it. Could I have done better, of course. I did the best I could at the time. I figured once a year people could read 20-30 pages for their job.

People want context. They want the big ideas. But getting an organization—of any size—to actually read is almost impossible.

The only reliable thing people read? Org memos. And even then, if one (as I often did) didn’t include an org chart picture—rather than just words—people would skim or skip and wait for (hopefully) a tree graph in the email.

And these were from the “big boss,” sending out “big strategy.” So if you think folks in big orgs are reading 40-page PRDs, budget plans, new product proposals, or deal docs deeply and regularly… you're probably kidding yourself. I know how the Amazon process has evolved from friends there. It too is breaking down which is a bummer as I am a huge fan of that.

Now enter AI. What happens when it's doing the writing—and not even the author has deep knowledge of what was written?

That’s like a compiled or multiple author memo no one ever actually read end-to-end.

And if people are asking AI to summarize—but the summary is lossy or invents data—what then?

I say all this as part of the "TV" and later "MTV" generation. Back then, we were told that fast-paced, cut-cut-cut media made us incapable of absorbing anything. Meh...ok boomer, I know you can't follow the plot of "24" but that's your problem not mine.

So maybe this is just old man yelling at cloud. But for me? My entire career has been defined by the reality that people in business don’t really read.

And this isn’t just a tech or big-company problem.

Take science: the reproducibility crisis is, in substantial part, because almost no one—not even reviewers—carefully reads full research papers. Same for grant proposals. They look quickly for pet issues (like statistics, sample size or technique, or if their own work was referenced). They skip over what is outside their domain. They miss explicitly fraudulent work which takes effort to detect (maybe AI reading helps with that?)

Take Wall Street. Every day analysts output 30 page write-ups on companies with detailed financial modeling. Almost no one is checking all those. People consume them for B/S/H and mostly for narrative confirmation one way or another. Tens of billions of dollars change hands on these that few read and even the authors don't always know the entire thing deeply.

Just thinking…