Why AI Isn't Making You Faster

There's probably a folder at your job that you hate opening up. And you know the one, right? Nobody is quite sure which version of anything in there is the most current one, and nobody has touched the naming since whoever set it up left the company.

That folder is the reason AI isn't doing much for you.

And I know exactly how that sounds. You've got a tool that can read a thousand documents in a minute, and here I am telling you the problem is probably your filing system.

That's exactly what I'm telling you.

I'm not guessing about this either. I spent 20 years going from analyst to CFO, and I've been on both sides of this one. I've been the guy with the mess, and I've been the guy who was brought in to clean up somebody else's mess.

So let me start with the mess.

The binder I never used

I was the kid in high school with a binder he never actually used. No dividers, nothing at all. Everything just went straight into my backpack. And yet I still managed to graduate as valedictorian, and that taught me exactly all the wrong lessons.

This brand of organized chaos held up right until my first job out of college, when somebody needed a document that I had while I was out of the office and they couldn't find it.

So what do they do? They came over to my cubicle. They went through the stack of papers sitting on my desk, they went through my cabinet, and when I got back to the office, they were not happy.

You've got to get your stuff organized. We can't find anything in here. Nobody knows where this document is.

So I went in and I found it in about ten seconds, because I knew where it was.

And that was the actual problem. My so-called system worked well enough for exactly one person. This guy.

We already ran this experiment

We've run this experiment before, and I had a front row seat during the first go.

What I'm talking about is back in the booming BI, or business intelligence, days. It does kind of feel like it was a whole era, right? I watched company after company decide that they wanted, no, they needed dashboards. Big, beautiful executive dashboards.

But what almost none of them wanted was to pay for the plumbing underneath.

What's different this time with AI is the price of entry. BI was expensive when you did it right. You needed a team of sorts, and most places were never going to fund that, so most companies ended up in the land of underwhelm.

AI is cheap-ish, and it's easy to deploy, and it doesn't stop at intelligence. It also automates and it streamlines.

So this time everybody gets to run the experiment, and everybody is running it all at the same time. Scroll for about ten seconds and you're going to see lots of AI peacocking. It's disgusting.

The base of the fire

And here's where the mad dash truly gets mad.

Every workflow that you want to automate eventually comes back down to brass tacks. It comes down to the environment where the work you do every single day exists. The files, the naming convention, where things actually get saved.

That's your base.

Ignoring your base and pointing AI at all the fancy work is like just spraying the top of a fire with some water. The flames may go down, which looks like progress, but it keeps smoldering underneath. So now you're back at it again tomorrow.

Same tool, opposite result

You can watch this play out department by department.

In teams where the work is already kind of structured, developers and engineers being the most likely case, they're going to get real benefit from AI almost immediately, because everything is documented and named and roughly where it belongs. Usually. If it's a good job.

But where the work is ambiguous, and the files live wherever somebody happened to drop them, the same tool is going to produce a mess with every bit as much confidence as it produced results for the organized group.

Same tool, opposite result. And the only difference is your base.

Which is also where the opportunity is, by the way. Ambiguity is the thing that drags AI down, and almost nobody is looking at that as something to fix.

Punching without footwork

So we've already established that I am the ambiguous case, what with my organized chaos sensibilities.

I have hopes and dreams and ambitions scattered across folders on my own machine, with no consistent rule for where any of it needs to land. It just depends on the week, and my ambition, and whatever happened to be in the works at the time. What's on my drive is basically the wake of ambition strewn all over the place.

So if I go looking for something I wrote months ago, or even days ago, and I get back the wrong version of it or nothing at all, that leaves me flailing once again, trying to get organized in some futile spiral.

But despite all that, I'd say I'm actually fairly productive. A heavyweight even, at least when it comes to output. It's just done in a very unorthodox way, with a lot of footgun potential.

It's almost like a boxer who punches really hard but has no footwork and no head movement.

You can think of the footwork as your clean workflows, and the head movement as your clean infrastructure. Combine those two things with power and you're going to take less damage, and you're going to land more of what you actually throw.

What it looks like when I got it right

Now let me show you what it looks like when I actually got this right, because I do have at least one example of that. At least one.

Years ago I had a standing monthly deliverable. There was an analysis that had to get done, a standard email that had to go out, and a set of images that had to be packaged up in a very specific way, every single month. Same shape, same form.

And it got built the way these things usually get built, which is very organically. Somebody asks you for one more thing and you bolt it on, and you never once get the afternoon to stop and ask yourself, why am I doing it like this? Because the asks never stop and you're just trying to keep up.

By the time that thing reached its final form, it was a couple of hours to execute on, every single month.

So at that point I finally sat down and did the unsexy work. I structured the files properly, organized the images, and wrote scripts to move all of it across the systems that it had to touch, seamlessly.

After that, it was just a one button click. Under ten minutes, and most of that time was just me reviewing the output before it actually went out the door.

But here's the part I want you to pay attention to. The script was the easy part. Getting the inputs to sit still, in a place that I could count on, is what actually took the work.

Why you're capped at 2x

There's a book called 10x Is Easier Than 2x, by Dan Sullivan and Ben Hardy, and it really resonated with me.

The argument is that going ten times bigger forces you to throw out the constraints that got you here and reimagine what would have to be true in order to achieve a 10x breakthrough, where going twice as big just means that you're going to work harder to do the same thing.

Here's what I would add to that.

If you tell AI to go make some magic using the mess that you already have, 2x is the best that you're going to get, if you're lucky. Same work, just with a little bit more resource thrown at it, moving a little bit faster. That's the book's own definition of making a small move.

So until you fix the base, that mess is going to serve as a rev limiter on everything that you throw at it.

And here's the part that I find genuinely funny. AI is actually really good at fixing the base and issues like this. Digging through piles of files, reading what's actually in them, telling you what you've got and how it ought to be arranged. That capability is sitting mostly unused while everybody is out there chasing the big strategic connections.

So if you firm up the foundation, that right there is going to give you 10x uncorked. Then you point AI at the firmed-up foundation that you just got, and 10x on top of that.

It's multiplication. And that's the point that's worth going after.

Put your own mask on first

You know the safety briefing on airplanes, right? The one that says in the unlikely event of a loss in cabin pressure, you put your own mask on first. That's because you're absolutely no good to anybody if you're unconscious.

Same order of operations here. You're not going to go fix your company's entire data infrastructure while everything you personally work in is strewn across three files under different names that only you understand.

The one-hour move

So go pick one folder. The one that you genuinely don't like opening up, because you never know if that's where the thing is that you're looking for. Or am I the only one with one of those?

Anyway. Point an AI tool at that folder with very firm rules:

• Read only.
• Write nothing. Change nothing.
• Profile every file. Audit what's actually in them.
• Then tell me what you found, and how this should be structured if there's a better way to do it.

And back that folder up first. This could get really heavy real fast, and there are serious implications, so going in with a copy is a smart thing to do.

That's an hour of work, probably less.

And from there you can either fix that one folder as your proof, or you can take a few steps back and ask what you'd build if you were starting from nothing, and then go build the first piece of that.

"I don't have time for this"

Now, the objection that I always get is, I don't have time for this.

You don't have time not to do this. Everybody is moving too fast right now, and by skipping out on this you're putting that rev limiter on your own velocity. You're the one pinning yourself to 2x.

And this spills over, which is the second thing that you get out of this. Once your own house is in order and you've got workflows that actually work, you've got something real to rally people around.

Because everything you touch is representative of you. If all you're building is flash and pizzazz based on nothing, there's nothing underneath for you to stand on. And mediocre just becomes your ceiling.

And the people who do this work end up getting asked to do more of it. That's a real reward. A real opportunity. Just make sure that when you pick up somebody else's problem, you decide upfront whether you're keeping it or handing it back trained. Don't end up collecting monkeys.

You're not the only one who benefits here either. Clean it up once and everybody downstream of you is going to move faster as well. And this is the thing that organizations will happily pay for.

The punchline

AI is a heavyweight. Set it down on a cheap patio set and it's going to crush your furniture, and probably your spirits.

So pick the folder this week. Just choose one, the one that you avoid. Back it up, point something at it with read-only access, and let it tell you what's actually in there.

This is going to feel slow. But this is you methodically, meticulously removing your own rev limiter.

And that is the opportunity you should be chasing.