top of page

My (Hilarious) Journey in Building AI Bots & Automation

Lessons Learned and Insights for Business Growth


🔑 Key Takeaways:


  1. Embrace Imperfection: Building AI bots is a learning process filled with unexpected challenges and humorous missteps.

  2. Start Small, Dream Big: Begin with manageable projects but keep your ambitious goals in sight.

  3. Simplify to Succeed: Complex workflows can often be streamlined by focusing on simplicity and clarity.

  4. Human Oversight is Crucial: AI is a powerful tool, but it requires strategic guidance and human intuition to truly thrive.

  5. Continuous Learning and Adaptation: Each setback is an opportunity to refine your approach and improve your processes.



You’d think someone obsessed with operational efficiency would crush building an AI bot, right?


Spoiler alert: life had other plans. 🚗💥 (Side note: My mom’s fine after the car accident chaos.)


Here’s how it all started.


I decided to take a week “OFF” (which, as a business owner, really just means tackling the pile of things you’ve been ignoring) to focus on two projects:


1️⃣ Build a lead-gen AI bot to help business owners pinpoint operational inefficiencies they can act on immediately.


2️⃣ Automate a back-end workflow to reduce the manual steps leading to a smoother, less labor-intensive process.


Simple enough, right? Spoiler: nope!




 

PART 1


Step 1: Dream Big, Start Small (Or So I Thought)


I set out to create a tool that wasn’t just useful but also fun—a bot that could take business metrics and spit out actionable insights like, “Hey, your gross margin is flirting with disaster,” or, “Congratulations, you’ve achieved operational nirvana!” (Yes, I do know a few owners in that last category!)


My goal was simple: help business owners pinpoint the key drivers of their success so they can focus on what truly matters. As a client said: "The roadmap you provided allows me to focus on what’s truly important to achieve our goals."


This became the foundation of everything I wanted the AI bot to deliver—clarity, direction, and actionable insights.


The first thing I knew… I had absolutely no idea how to build a bot. I’m not a coder—far from it. But I learn fast, and I really wanted to give this a try. I’ve always been drawn to problem-solving. So, naturally, I dove in headfirst, armed with determination, Google, YouTube, and a sprinkle of naivety.



Step 2: Life Said, “Hold My Latte”


Before I could dive in, life showed up. My mom was hit in a car accident, and suddenly my week off became a whirlwind of helping her navigate insurance, repairs, and all the chaos that comes with it. By the time I sat down to work on my AI bot, more than half the week was gone. Classic.

This is why we build automations, isn’t it? So when life throws us curveballs, the business doesn’t fall apart.



Step 3: The Website Rabbit Hole


When I finally got back to the bot, I realized my website needed an overhaul to make any of this work. Yay for unexpected detours! Here’s where I refined my limited knowledge to create better landing pages, hook up CRMs, and—this is the highlight—add a popup. 🎉


I know, for marketers, this is “welcome to 2010.” For me, it was a major victory. Now I had a system to capture emails because, let’s be real, lead-gen starts with emails.


My goal? Overdeliver on value so people wouldn’t begrudgingly give their email—they’d WANT to.




Step 4: Fun with Calculators (and Color Coding!)


The next step was adding calculators to help business owners visualize their margins, recurring revenue, and many more important items. Of course, I couldn’t stop there. I color-coded them:


  • Green = You’re crushing it.

  • Yellow = Room for improvement.

  • Red = Yikes.


Nine calculators later, I was ecstatic. Was it perfect? No. But seeing a business owner’s eyes light up when they realize, “Oh, that’s where my margin is slipping!” makes it all worth it!



Step 5: Barometer Magic


I also added the Business Barometer, a tool that tells you exactly where to focus. One client took this and turned a six-figure business into a seven-figure powerhouse. Seeing that transformation is why I do this.


The Takeaway: Building what I wanted wasn’t the smooth, quick process I imagined. But it reminded me of something Brian from Elevated Advisors (a client) said: "We are now a well-oiled machine that runs effectively with the right people in the right seats knowing what they need to do to be successful."


Want to see how it turned out? Head over HERE and tell me what you think. Is it helpful? Is it fun? Or does it scream “rookie coder vibes”?


Stay tuned—I’ll keep the storytelling coming as we dive into larger AI automation attempts….


 

Part 2:  Tackling A Workflow Head-On


With the first AI project wrapped up, I was feeling pretty good—the operational efficiency tool was live, helping business owners focus on the key drivers that matter most before year-end.


I used tools like Napkin.ai for visuals and a clever (if I do say so myself) use of custom code and color-coded calculators made it a fun and rewarding build. It was exciting to see something I created helping others gain clarity on their businesses.


So, naturally, I thought, "Round two should be a breeze, right?"


ROTFL!


This time, the mission was to tackle a larger workflow head-on—the kind that has too many manual steps, too many sticky notes, and not enough actual automation.


The Idea: Use an AI-powered tool that could take context, rules, and several inputs and transform them into two smooth, actionable outputs.


Think fewer headaches, more time saved, and scalable processes. The exact thing all business leaders want. And I wanted to know—could someone with no skills make it happen?


As they say, "Every plan is perfect until you hit a bump in the road."


Spoiler alert: I hit a few.


Step 1: Building a Database (Content) File


To even start tackling this project, I knew I needed something solid to work with. Enter the "content file." This wasn’t just about jotting down ideas or creating a checklist—it was about taking a year's worth of sales conversations, testimonials, and client feedback and turning them into a focused, strategic document that would allow me to speak in the language of my clients.


That’s what marketers tell us, right? We need to describe the challenges we solve using our client's language.


Why? Because building a workflow bot that can adapt to business needs requires understanding not just the processes and products but also the nuances of the work itself. What makes clients tick? What do they need most? What bottlenecks keep showing up? Themes. Context.


By feeding this into the AI, I could ensure the bot didn’t just process data but understood the rules, the goals, and the tone. The content file became the brain of the operation—guiding everything from decision-making to how the bot would interpret messy inputs.


The Result: A foundation that didn’t just solve problems but did so intelligently. It also helped me refine messaging for the workflow itself, ensuring every step aligned with what users actually needed.


Special Note: I did this process as with all the great things that people have said about us that we should have been using all along as 'social proof' and haven’t been. That resulted in a new website page called “results” … check it out HERE. I think my design skills are getting better! There are 92 individual testimonials on that webpage selected from the 335+ that we have! AI helped me categorize them all too!


Step 2: Choose the Tool


With the content file in place, I was ready to start building. I have big plans for this bot—simplifying key workflows, streamlining data transfers, and cutting down on manual effort.


But, as always, things didn’t go quite according to plan (more on that soon).


As a teaser, I took the time to look into which platform I wanted to play on. I opted for RelevanceAI.com. It has a free version, looks user-friendly to my untrained eye, and they have a knowledge base in case of trouble. I’ve chatted with people who find it easy to use.


Step 3: Garbage In, Garbage Out


You know what they say, right: Garbage in means you’ll get garbage out. So the next step really focused on thinking time.


Questions to Ask:

  • What data goes into the process?

  • What are the inputs?

  • What are the key metrics?


I wanted the metrics color-coded again for ease of readability, so what are the levels for each of these colors.


Then I created a Google Sheet for the first of the two outputs with conditional formatting for the colors. I took the time to create a knowledge base for the LLM/Bot to use that covered all of that. Which took a lot longer than expected.


I assumed the time spent upfront would be helpful because it should make the steps easier to create when the time came to do the actual build.


I am convinced that this was helpful. HOWEVER, in hindsight, I think there were pieces missing with my lack of experience.


This gap is closing!





 

Part 3: A Lesson in Simplicity


So here’s the deal: With my first bot experiment under my belt, I confidently marched into Bot #2, thinking, “This will be a piece of cake.”


Spoiler: It was not cake. It was more like soufflé – complicated, delicate, and prone to collapsing under pressure.


The Plan? Automate the processing of 28 key business metrics from four PDFs, populate a beautifully color-coded Google Sheet, and bask in the glow of operational efficiency.


Simple, right? Not so much.


Step 1: Tools or Agents?


First, I had to figure out if I needed an AI tool or an AI agent. A tool is straightforward: input X, get output Y. An agent, on the other hand, is more interpretive, like the philosopher of AI. After much deliberation (and reading up on the difference), I decided on a tool.


Why? Tools are efficient—so it seemed like a natural fit.


Win #1: I successfully uploaded the 4 PDFs. Yay me! But that’s when things started to unravel.


Step 2: From PDFs to Array


The PDFs needed to be scanned, interpreted, and turned into an array of data points.


Sounds smooth, right? Wrong. Turns out, my prep work for the bot was… lacking on a handful of the metrics.


Pro Tip: If the bot isn’t finding what you want, it’s probably because you didn’t teach it how.


But hey, the bot managed to extract most of the data. I was still feeling optimistic.


Step 3: Populate the Google Sheet


Here’s where optimism took a back seat. I needed to:

  1. Move the data to the right fields in a Google Sheet.

  2. Create a new tab dynamically for each upload.

  3. Ensure the tabs had the correct formatting.


This required APIs, JSON coding (which I had no idea about until now), and a whole lot of patience. I leaned heavily on the LLM for help.


Lessons Learned:

  • Dynamic Fields Don’t Work in Test Mode: It took me an hour to figure out I had to replace my dynamic fields with static ones when testing in the sandbox. Think Homer Simpson - DOH! Frustrating? Yes. Funny in hindsight? Also yes.

  • Free Plans Have Limits: RelevanceAI’s free plan gives you 100 credits a day. My workflow required 120 credits per run. I upgraded real fast after I learned that it wasn’t 1 credit per step but 4 credits.


By the time I thought I had all the pieces working, the bot said it had finished successfully. I checked my Google Sheet… no new tab.


The Result: No data. Just the mocking emptiness of a blank sheet.




 

Part 4: When Plans Fail… Improvise


Picking up from where we left off ... My bot had checked all the boxes but failed to deliver the goods. No new tab. No data. Just my growing frustration and a Google Sheet that refused to cooperate.


Where It Went Wrong:

  • PDF Uploads: The bot couldn’t retain information when restarted. Every time I walked away, I had to re-upload the PDFs. Ultimately what I wanted but at the testing phase…annoying, but manageable.

  • Google Sheet Integration: The API worked… except when it didn’t. Creating a new tab dynamically was hit-or-miss.

  • Free Plan Woes: I burned through my 100 daily credits trying to troubleshoot. My assumption that one step equals one credit? Laughable. Each step used four credits. Goodbye, free plan.


The Pivot:

I realized I needed a simpler solution. Instead of creating new tabs dynamically, I decided to:

  1. Manually upload the PDFs (a small sacrifice).

  2. Create a fresh Google Sheet for each new run, removing complexity.

  3. Adjust the API to populate data without requiring new tabs.


Tools That Saved the Day:

  • RelevanceAI: Despite the hiccups, it handled the core data extraction like a champ.

  • Google App Script: Once I figured it out, it was magical (and infuriating).

  • LLM Assistance: For someone who didn’t know about coding, this was my lifeline.


The Result:

After six to eight hours of trial and error, I managed to get a functional process—though not as elegant as I’d hoped. The data moved, the colors appeared, and the dream of operational efficiency survived another day.


But I know I can do better!


So the story doesn’t end here.


Spoiler - I pivoted again to a more elegant solution...that worked :)



 

 Part 5: AI & Automation—Powerful for Scaling & Not Taking Over (Yet)



AI and automation have changed how businesses scale, improve efficiency, and streamline decision-making. Despite the hype, AI isn’t replacing business leaders, strategic thinking, or the complexity of human intuition—at least, not yet.


From my own deep dive as a complete newbie into building AI-powered bots for automation, I’ve seen firsthand where AI accelerates efficiency—and where it falls apart without human oversight.


AI as an Accelerator, Not a Standalone Solution


Over the past few months, I’ve been refining automation workflows to streamline business valuation, operational insights, and process efficiency.


AI was supposed to make things faster, smarter, and scalable. But here’s what really happened:


  • The First Bot Worked Well: I fed it structured data, and it followed the format accurately.

  • The Second and Third Bot Needed Adjustments: I had to guide it, tweak responses, and ensure it was generating the right recommendations.

  • The Fourth Bot Derailed for a Moment: The outputs were off, the logic fell apart, and I had to step in manually to course-correct. All looks great now—expensive to run and not currently scalable—but works great.


This experience was a wake-up call—AI doesn’t automatically understand context, and it doesn’t self-correct when it misinterprets data.


Hallucinations, as we all call it.


The Mathler Story: Why AI Can’t “Think”


A perfect example of AI's limitations comes from a simple game I play with my daughter weekly—Mathler.


Mathler is a numbers game where you’re given an answer and have to figure out what equation gets you there. It requires a mix of logic, deduction, and pattern recognition.


One day, I decided to test AI’s ability to solve Mathler puzzles. I gave the bot the answer and the available numbers, expecting it to reason through the possibilities like I would. Instead, it failed completely.


It couldn’t adjust its approach based on previous guesses, didn’t recognize the constraints of the game, and had no ability to develop a strategy. It was simply brute-forcing calculations without context or intuition.


This was a powerful reminder: AI can process data, but it doesn’t think. It can’t pause, reconsider, or recognize subtle patterns the way a human brain does. This same flaw shows up in business automation. AI can execute tasks, but without human oversight, it will blindly follow patterns—even if they’re wrong.


Where AI Works Well


AI is an incredible tool when applied strategically. Here are the areas where it’s been a massive time-saver in my business:

  • Speeding Up Document Creation: Instead of writing from scratch, AI can generate structured drafts, cutting time in half.

  • Summarizing and Extracting Key Insights: AI helps process transcripts from meetings, pulling out the most relevant data.

  • Refining Messaging & Formatting: I use AI to tighten up writing, ensure clarity, and enhance readability.


For example, I’ve used AI-driven automation to generate business valuation reports, pulling from structured inputs and predefined benchmarks. While the first draft sometimes needs adjustments, it dramatically reduces manual workload in report generation.


Where AI Falls Short

  • AI Misunderstands Complex Workflows: The bot I built in RelevanceAI burned through hundreds of credits running the various actions. I thought one action = one credit, but I quickly learned that multiple background steps meant costs escalated fast.

  • AI Can’t Handle Nuance Without Human Input: When I tried to let AI handle SOP creation, it worked fine for the first draft but quickly lost context and needed human intervention for the final product.

  • AI Can Be Expensive if Not Optimized: I migrated later bots off an AI model to Google App Scripts because it was costing too much for tasks that didn’t require machine learning.


This is why automation without strategy is just expensive inefficiency. AI is an amplifier of expertise, not a replacement for decision-making.


The Future: AI as a Partner, Not a Boss

One of the biggest misconceptions is that AI is a magical solution that will replace decision-makers. The reality?


AI can’t trust a gut feeling. It won’t spot hidden connections. It won’t step back and ask if it's even looking at the right picture.


Automation can be amazing—until it isn’t. Take my experience with Bot #2, for example. I lost two hours trying to figure out why in the Google sandbox I could get it to create a single dynamic tab in Google Sheet. The answer is it can’t do that inside the sandbox as I’m not giving it the right input. AI didn’t flag the issue. It didn’t suggest a fix. It just kept running the troubleshooting instructions.


Sometimes, human oversight is still essential.


AI is a powerful assistant, but it requires a business leader to guide it, refine its outputs, and optimize its processes. The companies that win with AI won’t be the ones that try to replace thinking with automation—they’ll be the ones that integrate AI to accelerate, not replace, their decision-making.


What This Means for Growing Businesses

  • Use AI to enhance efficiency, not to replace human insight.

  • Invest in automation where it makes sense—but monitor costs carefully.

  • Refine workflows continuously, balancing AI with strategic oversight.


AI isn’t taking over the world (yet). Businesses that learn how to use it strategically will have a serious competitive advantage.


My 2 Cents: Embrace AI with Strategic Oversight

My journey into AI bot building was filled with laughter, frustration, and valuable lessons. From personal life interruptions to technical mishaps, each experience taught me the importance of simplicity, strategic planning, and human oversight in automation. These insights are now the backbone of the Fractional COO services we offer at Scaling Management Consulting Group.


Why Choose Scaling Management Consulting Group?

  • Right Person for Your Company: We meticulously match our Fractional COOs and Integrators to your unique business needs, ensuring a perfect fit for seamless and effective partnerships.

  • Tailored Solutions: Customized strategies addressing your specific business challenges.

  • Expert Leadership: Access seasoned Fractional COOs without the overhead of a full-time hire.

  • Proven Results: Streamlined operations, increased profitability, and exit-readiness achieved within six to nine months.

  • Continuous Support: Our wrap-around services ensure ongoing support and collaboration, fostering long-term success.


Ready to Transform Your Business?

Embrace the power of AI and automation with the right strategic oversight. Book a call with me and  learn how our Fractional COO services can help your business thrive, streamline operations, and achieve sustainable and profitable growth. 


Let’s work together to turn your business into a profitable, self-sustaining operation.





 

 

Explore More Resources to Grow Your Business


Looking for more insights to scale your business? Check out these resources:


Stay Connected

Follow me on LinkedIn, subscribe to our newsletter  or our YouTube channel for tips, expert advice, and updates on how to grow and scale your business efficiently.


Let’s Chat

Ready to take the next step? Schedule a call and see how we can help you achieve your business goals.

Comments


Commenting has been turned off.
Scaling Management Consulting Group, Inc.
support@scalinggrp.com
bottom of page