An AI Saga: A Story No One Asked For But You’re Getting Anyway

Remember me? 

I’m still the wrong Doug – the one who writes marketing things and has near full mental meltdowns moments before companywide AI trainings where I ramble, often incoherently, and find myself wanting to crawl off and scream into the void. I’m still not the fun AI Doug. (Yet.) I’m just the guy who once read “For Whom the Bell Tolls” and thought, “Ernest… buddy…  you get me…” 

So here I am again. Days into this AI journey where Copilot is behind the wheel, and I’m the guy riding shotgun, messing with the radio, turning the heat up too high, throwing out half-baked directions, and saying things like, ‘Is your blinker on?” 

The Assignment: Build a Partner Promotion Strategy 

A bold endeavor that no one asked for, fueled by ambition, some hope, and my unrelenting confidence in machines.  

The Reality: Days of Trying Not to Throat Punch My Laptop 

Welcome to ‘An AI Saga: A Story No One Asked For But You’re Getting Anyway’, a process I document here not because it went well, but because it didn’t. And I think that’s the whole point when using AI.   

Turns out AI was not the one struggling. I was the one struggling. The guy trying to give directions while reading the wrong map upside down. 

Day 1: Hope, Prompts, and Pain 

Things started well. I had swagger. I was using all of the new tools – Researcher, Analyst, and Conversations – and all in unison to pull data from every contract, every business partner, every corner of the Internet, and learning just how much roofing and cloud computing apparently have in common… 

I built calendars. I layered sales data. I practically sang as I told Analyst to “rank stuff by importance,” and my robots happily complied… by putting Tremco Roofing in the top 5 for technology.  

Gorgeous roofs? Absolutely. But unless they’re adding Wi-Fi-enabled shingles, I think something got lost in translation. 

That’s when I learned the first AI truth. 

It’s not the model’s fault if I tell it nonsense and expect poetry. 

So, after my robot friends mixed up roofing with routers, it was time to try again. New prompts. New outputs. Same existential dread. 

Day 2: Fuzzy Matching and the Python Awakening 

At this point, I’d narrowed my scope to just technology and was running data through Analyst like it was a Vegas slot machine and I was desperate for a jackpot. But I kept getting back half-right answers and a lot of “value errors.” 

I tried matching vendor names across datasets using filters, but it turns out that “SHI” could also be matched to “eShipGlobal.” So I dusted off my new favorite phrase: fuzzy matching. I learned a little more, and prompted my way to the right percentage threshold and then… 

Magic. (Or so I thought in the moment.) 

Did it feel like cheating? A little. Did I care? Nope! 

Then came a moment that changed everything: Copilot generated a Python script to merge all my mismatched data and then walked me through how to run it in Excel like I was an honorary member of the Geek Squad. I’m a marketer. I’ve never merged anything more complex than exits on I95. I hear things like “pivot table” and instantly go cross-eyed. But here I was, typing code like I belonged on Stack Overflow.  

I won’t lie. I had a moment. I felt powerful. Like maybe I deserved to wear a hoodie, turn off the lights, and crack my knuckles before typing. Had I officially become a vibe coder? 

Day 3+: The Art of Asking Better Questions 

Turns out, AI gets better when you do. When I refined my prompts, gave better context, and stopped trying to make Copilot do everything for me, things smoothed out. A bit. 

But the deeper lesson came from Copilot pushing back. Asking questions like: 

• “Do you want this based on national or regional buying patterns?” 

• “Should the campaign reflect academic calendars, fiscal cycles, or both?” 

• “Is there a human in charge of this, and can I speak to them instead?”  

Suddenly, we were collaborating. And the more we talked, the more I stopped treating Copilot like a vending machine and started treating it like a very confused intern who, with proper guidance, could help me create a massively better outcome. Then I started treating copilot like a full staff of data scientists, researchers, and sales reps with years of industry experience. Finally, my thinking flipped. No longer was I looking at this from the data down but instead started looking at it from the member up. That simple mind switch, and one that Copilot led me to, is truly what changed everything. 

And that’s when it clicked: AI isn’t here to replace you. It’s here to annoy you into working smarter and thinking differently. 

Lessons Learned: Failing Forward Isn’t Just a Phrase 

AI is not here solely to make things easier. It’s here to expose every hole in our thinking, every sloppy spreadsheet, and every time you’ve pretended to understand vlookups and mail merges. 

But with each failure (and there were many), I learned something new. Not just about automation or syntax or matching algorithms, but about how I think. What I assume. And what I ignore until my beautiful robots force me to deal with it. 

If that sounds frustrating, it definitely can be. If it sounds valuable, it is beyond anything. And if you’re still reading this… congratulations: you’re now part of my coping mechanism. 

Here’s what I know today: 

If you stick with AI through the bumps and dead ends and incomplete data, you might just end up somewhere really, really awesome. 

And even when you see the mess you made along the way, you’ll notice that every bad prompt was a step toward something smarter.

Or you’ll just drink more wine with a nice cigar every night. Either way. Progress.