Intro to sears ai failure
I love AI.
I think it’s incredibly beneficial, and I use it constantly.
But anyone who has spent enough time actually pushing AI eventually learns the same thing:
It has limits.
I don’t view AI as a replacement for people. I view it more like a paintbrush in the hand of a painter.
The paintbrush can be extraordinary.
But it’s still only as effective as the person holding it and the process surrounding it.
I learned that in a pretty funny way while developing the theory behind my doctoral research.
At some point, I turned it into a game:
How far could I push AI with the entire framework before I reached the point where it could no longer reliably reason across that level of abstraction as a unified system and I had to start breaking it into smaller pieces and spoon-feeding it the structure?
And I got there more than once.
The problem wasn’t necessarily the individual concepts. It was the architecture.
I was trying to model interconnected gravitational structures, institutional behavior and game-theory implications of governmental systems potentially destabilizing under the cumulative weight of their own actions.
I’d hand AI the entire framework and effectively say:
Here. Look at this from 30,000 feet. What happens?
And sometimes it couldn’t.
Presented as one unified system, the framework exceeded the level of abstraction the model could reliably reason across at the time.
So I learned to decompose it or, less academically, spoon-feed it.
Break the structure apart.
Explain Layer A.
Make sure it understood A.
Introduce B.
Connect A to B.
Add C.
Build another relationship.
Then another.
Then put the entire puzzle back together.
Suddenly:
“Ohhhhhhh.”
Okay, maybe the AI didn’t literally say ohhhhhhh.
But that was basically the experience.
Once it understood the individual pieces and their relationships, it could analyze the larger structure, challenge assumptions, run through game-theory scenarios and help me interrogate the framework.
Then the models improved.
After a major update, concepts I’d previously had to painstakingly deconstruct became considerably easier for AI to process.
But even then, there were limits.
And that’s when something became very clear to me:
AI capability and AI implementation are two entirely different things.
Which brings us, somehow, to Sears.
My brother had a Sears Home Services appointment.
Nobody showed.
Then came automated messages suggesting an appointment that hadn’t yet passed had already passed.
Then came the AI telephone system.
Then came the bizarre scheduling loop where the appointment window actually had passed, the AI offered to keep the appointment “as is,” rescheduling couldn’t produce another available appointment, and reaching an actual person became its own adventure.
So we started researching.
And apparently my brother wasn’t alone.
That turned an annoying missed appointment into a much more interesting business question:
What happens when a company automates customer service before properly designing what happens when the automation fails?
Because this isn’t an anti-AI article.
I’m probably the wrong person to write an anti-AI article.
It’s about understanding what AI actually is.
AI can be extraordinarily powerful.
But if the technician doesn’t show up, the customer doesn’t need an AI capable of repeating three predetermined options.
The customer needs a system capable of recognizing:
Something went wrong. The normal process no longer applies. Escalate.
And sometimes that escalation needs to end with a human being.
AI doesn’t magically repair a broken process.
Sometimes it just automates the broken process.
Which brings me back to my paintbrush.
AI is the brush.
The business is still the hand holding it.
And if the hand holding the brush is already shaking, giving it a more powerful brush doesn’t necessarily fix the painting.
It might just let the hand shake faster.
And Sears?
I’ve apparently been watching them shake for a while.
Back in 2013, I tweeted at Sears while Man of Steel was everywhere, joking that I could see Sears all over Superman but they couldn’t even keep their own tower in Chicago.
Sears responded: “You haven’t seen the last of us! Muah-ha-ha…”
Thirteen years later, apparently they were right.
F’nAround introduces:
Sears Survived Bankruptcy. Can It Survive Its AI?
A F’nAround Look at What Happens When Automation Stops Helping Customers