Everyone talks about how to use AI, so let’s talk about how not to use AI. I couldn’t co-sign an apartment lease without losing my mind. Here’s what Equity Residential got very wrong — and what every leader deploying AI needs to hear.
My daughter needed a co-signer for her first apartment.
I figured it would take 20 minutes.
It was easier buying a house.
I’m not exaggerating. The mortgage process — with the underwriting, the appraisal, the title search, the closing documents — was less frustrating than co-signing a lease at Alexan Harrison, an Equity Residential property in New Jersey.
Here’s what happened.
Meet Ella
The moment I submitted my information, I got an email from Ella. Ella is the AI Resident Assistant for Equity Residential. Ella is automated. Ella told me so at the bottom of every single email.
“Replies may be human and/or AI generated.”
Okay. Fine. I build AI tools. I understand automated responses. I was willing to work with Ella.
The problem wasn’t that Ella was AI. The problem was that Ella was broken — and nobody was watching.
20 Emails. Zero Progress.
Ella sent emails. A lot of them. Each one with a different subject line, so they never grouped in my inbox. I didn’t have a thread. I had a inbox full of individual messages from an AI that had no memory of what it had already asked me.
Each email asked for something new. Or asked for the same thing again. Or told me the application was under review. Or told me it needed more information. Or told me something had been received.
I had 20 emails from an AI assistant and still had no idea what the status of the application was.
They never ran my credit. Not once through the entire process. For a co-signer application. The one thing that actually determines whether a co-signer qualifies — never happened.
Instead, Ella kept asking for documents. Then more documents. Then a letter of explanation. Then tax returns.
Tax returns. For a co-signer on a one-bedroom apartment.
The Phone Call That Fixed Everything in 90 Seconds
After weeks of this, I called.
A human answered. I explained what had been happening. She paused and said:
“Oh — just ignore those emails. You’re fine.”
That was it. Done. Application approved.
The AI had spent weeks creating friction, confusion, and frustration. A human solved it in 90 seconds.
If your AI creates a problem that only a human can fix, you don’t have an AI solution. You have an AI problem with a human patch.
What Equity Residential Got Wrong
I want to be fair here. AI in property management makes sense. Leasing offices field hundreds of routine inquiries. Automation handles volume. I get it.
But Ella wasn’t handling volume. Ella was handling my application — and she was doing it badly, repeatedly, without escalation, without memory, and without a safety net.
Here’s where it broke down:
😤No human fallback trigger. When an application stalls — same person, multiple document requests, no credit pull after two weeks — that should automatically route to a human. It didn’t. Ella just kept sending emails.
😤No conversation threading. Twenty emails with twenty different subject lines is not a communication system. It’s noise. A customer shouldn’t have to build their own thread to understand where they stand.
😤No progress visibility. At no point did I have a clear answer to the simplest question: where does my application stand right now? That’s a basic UX problem that AI should solve, not create.
😤Requests without logic. Asking a co-signer for tax returns before running their credit isn’t a process. It’s a broken workflow that nobody reviewed before it went live.
The Lesson
My daughter got the apartment. Eventually.
Ella is still out there, sending emails with different subject lines, asking co-signers for tax returns, and creating unneeded friction.
AI should make that easier. Not harder.
If you’re a leader deploying AI in your customer experience, please — make sure someone is actually watching what it does when it goes wrong. Because it will go wrong. And your customers will find out before you do.
