Home AI Edge Pipeline Discovery Closing Playbook About
AI on the Ground

The Last Person You Trust With This

By Jon Ekanger · July 23, 2026 · 5 min read

Want the summary?
The Last Person You Trust With This

It’s 6:45 AM. The bar is empty, still smelling of last night’s disinfectant. You’re not here to set up. You’re here because it’s the only quiet hour you’ll get all week to look at the stack of 93 applications for your new bartender. You open your laptop. You click the button that says “AI Screen & Rank.” In thirty seconds, it’s down to eight. You feel a guilty wave of relief. You just got your morning back.

Two weeks later, you’re standing beside your best server, watching your new “top candidate” struggle to make basic small talk with a regular. The resume said “five years hospitality experience.” The AI scored them a 94% match. They can make the drinks. They can’t read the room. The relief is gone. Now you’re just tired.

This isn’t a story about AI being bad. It’s about us using it for the one job it is catastrophically unqualified to do: the first human impression.

The promise is seductive. Dump the resumes in, get a ranked list. It filters for keywords, years of experience, job hopping. It writes polite, legally-safe rejection emails to the other 85 people. It saves you hours of scanning. And on paper, it works. The candidates it surfaces look right. They have the right words in the right places.

AI is brilliant at finding candidates who are good at applying for jobs. It’s blind to the ones who would be good at the job.

The failure is invisible, and it happens in the filter. The machine is sorting for a pattern of past success, not for potential for future fit. In a small business, the second one is the only thing that matters.

What the Filter Throws Away

Let’s talk about the 85 “rejects.” In that pile is the resume of the woman who managed a busy coffee shop for three years but calls it “customer happiness and flow management,” not “shift supervision.” The AI, looking for “shift lead” or “supervisor,” flags it as weak.

In that pile is the guy with a two-year gap because he helped his family’s restaurant through a bankruptcy. The AI, trained to see gaps as red flags, downgrades him.

In that pile is the applicant whose previous job title was “Utility Operations Specialist” which, you’d only know if you called them, meant they were the go-to problem-solver for a local event venue. The AI has no idea.

The machine is optimizing for a clean, predictable narrative. Your business runs on messy, unpredictable reality. You don’t need a perfect pattern. You need a person who can handle when the pattern explodes at 7 PM on a Friday.

The Cheap Copy of a Human Gut

So we use the AI to get to the shortlist faster. Then we trust our gut in the interview, right? Maybe. But something subtle happens first.

Your brain, seeing that “94% match” score, enters the interview with a bias. This person is vetted. The hard part is done. You start looking for confirmation, not for truth. You might overlook the slight stiffness, the rehearsed answers, because the data said they were perfect. You’ve outsourced your skepticism.

The owner two blocks over, the one who’s quietly winning? He’s not reading 93 resumes either. He’s doing something simpler and smarter. He used the AI for the part that sucks but doesn’t require judgment.

The Quiet Win: Letting AI Do the Dogwork, Not the Decision

Here’s what he did. He wrote the job post himself, in his own voice. (“Need a bartender who smiles before their feet hurt.”) He let the applications roll in.

Then, he used a simple, cheap AI tool to do one thing: summarize each resume into three bullet points.

* Bullet 1: Their last two jobs, in plain English.

* Bullet 2: One thing they seem proud of (pulled from their summary).

* Bullet 3: Any weird gap or jump, flagged neutrally.

He didn’t ask it to rank. He didn’t ask it to judge. He asked it to translate. In 20 minutes, he had 93 resumes turned into 93 concise, readable paragraphs. He skimmed them all with his own eyes, his own gut. He flagged twelve that felt interesting. Some had perfect experience. One had a weird career jump from teacher to barback that made him curious.

The AI did the tedious reading. He did the human seeing. It cost him a coffee. It saved him hours. And it never decided who was worth his time.

The Action You Can Take on Monday

Stop using AI to say “yes” or “no.” Start using it to say “here’s what this is, in plain English.”

The next time you’re hiring, take every application and run it through a basic chat tool with this prompt: “Summarize this resume for me. Give me: 1. Their last two job titles and companies in simple terms. 2. One thing they highlight about themselves. 3. Note any employment gaps or major job changes without judgment.”

Then, you read. You look for the story the machine can’t write: the problem-solver, the weird career pivot that shows grit, the person who might not know your POS system but knows how to calm down an angry guest.

AI’s real job in your hiring isn’t to be the gatekeeper. It’s to be the translator, getting you past the paperwork fog so you can stand at the door and trust your own instincts. The last person you should trust with your first impression is a machine that’s never met a customer, never worked a double, and never had a gut feeling in its life.

Post-Ready

The LinkedIn Version

Copy, tweak the first line, and post. Built to stop the scroll.

LinkedIn Post
The worst hire I ever made… the resume looked perfect. Every box ticked. The AI screening tool said "strong match." I didn't even have to write the rejection letters. The problem was, I didn't have to write the interview questions either. I outsourced the first, most human filter to a machine that's trained to find patterns, not people. It found a pattern. It missed the person. Now, every time an owner tells me they're using AI to screen resumes, I think about that hire. The cost of the mistake, the time lost, the team morale. We're handing the front door of our businesses to a polite, efficient gatekeeper that has zero intuition. It sorts for keywords, not for grit. For past job titles, not for potential. It filters out the weird, interesting resume that doesn't fit the template… the exact person who might actually save your shop. The AI isn't the problem. Using it to do the part that requires a gut check is. Are you letting a pattern-matching machine decide who gets a chance to walk in your door?

1020 characters · Optimized for the LinkedIn feed (hook + line breaks + CTA)

Keep Reading

The Revenue Endeavor

One sharp idea on B2B selling every week, prospecting, discovery, closing, and the operator mindset. Plus the post-ready version for your feed.