The Day AI Shut Down My Own Company: Two Cautionary Tales From a Phoenix Agency

Editor's note: portions of this piece were provided as commentary to Neil Schneider for Go Local Arizona's feature on artificial intelligence.

So I'm going to tell you the story of what happened recently with two experiences that I can confidently say are on the timeline of The Terminator coming to exterminate all of us.

What We Actually Build With AI

Quite often, my agency is recognized as one of the leaders in the contractor market AI tool space. What does that mean? It means we build tools and visualization platforms for contractors so they can better close deals.

For example, they can take a picture of a backyard from a web app that we've built, input exactly what they want in that backyard, and get a client ready PDF back within about 60 seconds.

I'm sure you're saying, "Yeah, well, we can do that with ChatGPT."

No.

No, you can't.

We've built the tool so that if you want to remove just that one bush that's been there for twenty years, you can remove just that one bush. Regenerative hallucinations are not something you're going to avoid with off the shelf AI tools. We build custom tools.

So wherever your mind just went about how you can do the same thing, no, you can't.

Without trying to sound arrogant, we've had a couple of technological firsts with the use of AI here in Arizona.

One of those was our employee training program.

We decided that a small company like mine, with only five employees, doesn't have time to spend months training new people. Most companies with five or fewer employees feel exactly the same way.

So we built a tool.

Her name was Stephanie.

Stephanie is a voice enabled trainer that takes all of our internal knowledge and presents it in a conversational question and answer format. New employees sit down, click a button, and begin asking Stephanie questions about their job from a spreadsheet we built in the form of a Google Form.

They converse with Stephanie about everything from standard operating procedures, clocking in, clocking out, emergency exits, security protocols, computer logins, and how we do things as a company. They then enter those answers into the Google Form, which is automatically sent to me or one of their managers.

That's how we evaluate training.

Instead of taking months to get someone productive in an industry that requires real expertise, we've shrunk that time to effectiveness down to weeks.

We've also deployed the first painting website where a customer can text a picture of their home through the website, choose paint colors, and receive a PDF quote from a Sherwin-Williams certified painter.

Along with the landscaping visualization tool we built earlier, we've also built a pool visualizer and several other contractor focused visualization tools.

But it's not all sunshine in the kingdom.

There are two recent missteps that I feel very strongly illustrate why companies are beginning to rethink how aggressively they replace people with AI.

The Report That Shut Down My Company

A lot of you know we're also an SEO company. We specialize in bringing contractor websites to the front page of Google, into Gemini conversations, into Google Maps rankings, and we manage Google Ads, Local Services Ads, and Google's entire advertising ecosystem.

Recently we brought on a mortgage client that was slightly outside the industries where we normally work.

They asked, "Can you show me examples of the SEO work you've done?"

I said, "Absolutely."

I turned to our Director of Digital and yelled across the room, "Can you build me a spreadsheet real quick with all of the first page rankings we have for our clients?"

She said, "Sure."

The meeting continued. We closed the deal, walked the client downstairs, and everything seemed great.

When I got back to my desk, I opened the spreadsheet.

I absolutely lost my mind.

The spreadsheet only showed about eighty first page rankings.

We had more than forty SEO clients.

To me, that meant one of two things.

Either we were running a scam, or something had gone catastrophically wrong.

I immediately pulled social media, development, and advertising to a complete stop so we could figure out what was happening.

About an hour into this, I was completely perplexed because I knew the report couldn't possibly be accurate.

Our own website alone has around eighty first page rankings.

So I started digging through the data.

My first bad management decision was going straight to the source.

I asked, "Did you even pull the report like I asked? How can this possibly be true?"

She replied, "I used AI inside the reporting platform to generate it."

Now remember, I only have five employees, but each one runs an entire department.

Had this been a company with hundreds or thousands of employees, this little reporting error could have shut down departments all day over something that wasn't even real.

After manually reviewing everything, we found that we actually had well into the five hundreds of first page rankings.

More than ten rankings per client on average.

Because we trusted AI to summarize the data, I shut down my own company for hours, caused unnecessary stress, and placed blame where it never belonged.

I suspect that exact scenario is happening all over the country inside serious businesses every single day.

Somebody Else's Cheaply Built Robot

The second incident happened that same day.

I'm still not entirely sure what the long term fallout is going to be.

We hired a company to do cold email outreach for us.

They used our information and an AI personality based on our brand to create outreach campaigns aimed at construction companies around the country.

Like every cold email campaign, some companies replied with "Stop" or "Unsubscribe."

That's normal.

Every legitimate outreach company should immediately place those businesses into a permanent do not contact list.

Well, about a dozen companies replied that they weren't interested.

Instead of unsubscribing them like it had for thousands of others, some glitch inside the AI responded to those business owners with this:

"Okay, then we will move on to someone who is serious about their business."

I couldn't believe what I was reading.

Our reputation.

Our stewardship of the Phoenix business community.

Everything we've spent years building.

It was being damaged by somebody else's cheaply built robot.

Needless to say, we forwarded every email, demanded a full refund, and got our money back.

Today we handle our outbound communication in house under Stephanie's supervision, the same AI that manages portions of our customer service and helps train our employees.

Some of you have already met Stephanie on the front page of our website.

Everybody Suddenly Thinks They're a Graphic Designer

Now let's move down to a much lower level of AI use.

This is actually the quote Neil was asking me about.

Everybody suddenly thinks they're a graphic designer because they have ChatGPT, Claude, Grok, Gemini, Perplexity, Copilot, or whatever new model comes out next week. They're not. The graphics all look the same because the models were trained on largely the same visual data.

Everybody suddenly thinks they're a graphic designer because they have ChatGPT, Claude, Grok, Gemini, Perplexity, Copilot, or whatever new model comes out next week.

They're not.

The graphics all look the same because the models were trained on largely the same visual data.

Yes, they're getting better.

The text is cleaner.

The edges are sharper.

The spelling mistakes are disappearing.

But if you actually want to use that logo on a truck, on embroidery, on signage, or in print, you're still going to have to take that JPEG or PNG into Adobe Illustrator and turn it into a proper vector file.

That's problem number one.

Problem number two is that everything looks exactly the same.

Just like you can instantly recognize writing that obviously came from ChatGPT, you can now recognize AI generated graphics from a mile away.

When people see your ad and immediately think, "Oh great, another company using AI," you've just replaced your brand value with generic automation.

That matters.

It changes how people perceive your business.

There's also another issue that nobody talks about.

You're probably showing your AI generated advertisement to someone who may have lost their own job because their employer replaced people with AI.

There is growing resentment toward artificial intelligence, and some of that resentment gets transferred directly onto your brand whether you intended it or not.

So if you're going to use AI to generate graphics, start small.

Use it for icons.

Use it for concept work.

Don't try to build your entire brand identity with it.

Don't generate your entire flyer.

Don't generate every banner.

Could that change in two months?

Absolutely.

These models improve almost weekly.

But today, it still looks cheap.

I know you're proud of the prompt you wrote.

I know you feel like you created something.

The majority of the people looking at your advertisement don't see it that way.

They know how easy it was.

They know a robot did most of the work.

When you begin calling those graphics your own, you're unintentionally lowering the perceived value of your own company because the public understands how little effort it took to create them.

You're standing on the shoulders of giants.

Like Prometheus bringing fire to humanity, the backlash may be your brand having its proverbial liver pecked out day after day through criticism you never saw coming.

Stay In Charge of the Process

For those of you running companies like mine, where you're trying to steer a ship while streamlining productivity, you absolutely need to stay in charge of the process before you allow a robot to make decisions on your behalf.

When I look at the recovery efforts we've had to make from the two situations I just described, all of the time we thought we had saved disappeared.

Every hour we gained through automation was immediately spent correcting automation.

Ultimately, AI is still an elaborate logic engine rolling the dice and hoping it gets the answer right.

And maybe, just maybe, if we stop letting AI make quite so many of our decisions, we can prolong humanity by a few precious decades...

Before the T-1000s finally invade our streets.

One last thing. None of this means we should retreat from the machines. Connecting with real people and real businesses is still what we need to be doing as a society. Every handshake, every referral, every conversation at a chamber event does something no model can replicate. If we keep outsourcing those connections, we drift farther away from each other and become more and more reliant on AI services to do the talking for us.

That is why I appreciate what Neil Schneider is building over at Go Local Arizona. Support the people in your own community first. The robots can wait.