The Week I Realized AI Had Made My Inbox Worse ~via Jay Thornton
This post from Jay Thornton, AskLedo, REALLY RESONATES. AI isn’t creating the problem… it’s amplifying it. Technology can scale output at an incredible pace, but it can’t replace judgment, restraint, or human oversight. I’ve been saying for a while now that the biggest AI challenge won’t be implementation… it’ll be remediation. Great piece, Jay.
P.S. Don’t miss this at the end of Jay’s post… The fuller, more buttoned-up version of the argument is on my blog, if you want it all in one place: AI Was Supposed to Cut the Noise. In Most Companies, It Became the Noise. -Ted
I want to tell you about the moment this whole idea stopped being a theory for me and became something I could feel.
It was a Tuesday, nothing special about it. I'd cleared maybe thirty emails before prepping my lunch and I was wiped out, which made no sense, because I hadn't actually decided anything or moved a single thing forward. So I scrolled back up and reread them. Almost every one was long. Not thoughtful-long. Padded. A one-line question buried under an intro, a couple of bullets, and a little summary at the bottom recapping the thing I had just read one paragraph earlier. One was a note from an AI meeting tool I’d connected weeks before, writing me a full paragraph of context about a “lead” who was actually my business partner.
None of it was technically wrong. All of it was noise. And here is the part that got me.
That has stuck with me since. We were sold the idea that this stuff would take weight off. For a lot of people, me included, it quietly piled more on. The tools got very good, very fast, at producing. Almost none of them got any better at deciding what was worth producing in the first place.
I have started calling it AI bloat, because that is what it is. Inboxes and CRMs and shared drives, all slowly filling with output that nobody actually shaped. Everyone got handed an assistant. Almost nobody got a system. Give a few hundred people a powerful generator and no real rules about when to use it, and you do not get clarity out the other end. You get volume. Confident, well-formatted volume.
Here is what actually keeps me up, though, and it is bigger than a heavy inbox. That same unshaped output does not just sit in your email being annoying. It travels. A summary invents a detail nobody said. It lands in your CRM. Your CRM rolls up into a pipeline report, the report feeds a forecast, and the forecast is the thing you use to decide where the money goes next quarter. One made-up sentence at the top of that does not stay one sentence. It gets averaged and charted, and by the time you see it, it is wearing a suit and calling itself data.
That unsettles me more than I expected it to. Hallucinations do not warn you. They show up fluent and calm, in the same voice as everything true sitting next to them. And the whole time it is happening, your dashboards look fantastic. Activity up. Output up. Everybody busy. Meanwhile the numbers you actually trust are drifting away from anything real, and you usually do not find out until a forecast misses by a mile and you go digging and hit three layers of generated summary before you reach the mistake.
We have watched this movie before, if we are honest with ourselves. The internet spent two decades paying out for noise. Whoever shouted loudest won, and we all quietly accepted a permanent low hum of overwhelm as the rent for being online. Unstructured AI took that same broken incentive and strapped a rocket to it. The old web could only make noise as fast as a person could type. A model does it in a second, on any topic you want, and it never once stops to ask whether the thing needed to exist at all.
So I quit believing the answer was a smarter model. The models are already incredible, and they will keep getting better with zero help from me. The thing almost nobody is selling is restraint. That is what I got a little obsessed with. The edge from here will not be intelligence. It will be restraint. Good technology should leave you feeling more capable, not more frantic, and when I looked around I could not name many tools that were clearing my head instead of stuffing it.
That obsession is basically what Ledo is now. We call the work Calm Intelligence, and I want to be plain about it. This is not a feature we screwed on afterward. It is a discipline, and it is a hard one to hold. The rule we keep is easy to say. Every time you touch Ledo, you should walk away more clear-headed than you came in. Sounds soft until you try to build to it. Mostly it means we kill things. A notification that is not worth interrupting you for does not ship. A summary that adds words but no understanding does not ship. One clear signal beats ten you have to scroll through, even when ten would demo better and make the next investor call a little easier. The philosophy is not the hard part. The hard part is what you choose to leave out at two in the afternoon, when the feature sounds reasonable and building it would at least feel productive.
The other thing we did was refuse to let Ledo talk out of thin air. Bloat comes from generating against sources nobody checked, so we built the reverse. Ledo only learns from things that actually worked, real outcomes, stripped of any client identity, and it will not treat a pattern as real until more than one source backs it up. Every piece of that carries a confidence score, so the system knows how sure it is instead of faking certainty. Before it says a word to you, it goes and finds what has genuinely worked in a spot like yours. It is a deliberately boring way to build. The boring is the whole point.
I do not think the companies drowning in this made some dumb mistake by adopting AI. They adopted the capability and skipped the discipline, and that is just a faster way to make noise. I would rather build the calm one, even though it is slower, less flashy, and means saying no to features that would look incredible in a demo.
If any of this lands for you, come see what we are making. It is at askledo.com, and it is the most honest answer I have to that tired Tuesday that kicked the whole thing off. The fuller, more buttoned-up version of the argument is on my blog, if you want it all in one place: AI Was Supposed to Cut the Noise. In Most Companies, It Became the Noise.
Talk soon, Jay
