Behind the AI Breakthroughs: A Journey of Failure, Growth and Transformation

HOST: Welcome to Digital Journeys, the podcast from Cherry Bekaert that explores how technology is transforming business optimization. Each episode, we unpack the latest trends, strategies, and innovations driving efficiency and growth in today's fast-paced digital landscape.

HOST: Join us as we navigate the path to smarter, more agile business practices. This is Jerry Herden with Cherry Bekaert.

HOST: Joining me today on Digital Journeys is Greg Kaupp, who is also with Cherry Bekaert. He joined us from ArcherPoint not too long ago and has started a fantastic series on LinkedIn called AI Over 40.

HOST: He is getting a lot of impressions and engagement on this series, so we thought we would conduct a podcast on it. Greg, I would like for you to introduce yourself and tell the listeners a little bit about you.

GREG KAUPP: Thank you, Jerry. I celebrated my one-year anniversary at Cherry Bekaert just last week; October 9th marked one year.

GREG KAUPP: It has been great being here. I am a partner at Cherry Bekaert and lead the Microsoft ERP practice. We focus on Dynamics 365 Business Central for the mid-market.

GREG KAUPP: As you mentioned, I started a series called AI Over 40. I was feeling frustrated because I knew AI was transformational, but I did not feel like I was being transformed.

GREG KAUPP: I decided I could not be the only one feeling this. I was in a meeting with business leaders from different companies and asked how AI had changed their specific jobs.

GREG KAUPP: Not a single person in the room could tell me anything about how it changed their job. The best responses were about summarizing email, filling out RFPs, or helping create presentations.

GREG KAUPP: That is a far cry from the transformation we hear about where AI is going to replace us all. I realized we needed to start with personal transformation before we could talk about how AI would transform our work, our lives, or our coworkers.

HOST: The "over 40" part is interesting. I actually think this series could be for any age because there is so much knowledge in it. What was the hook for over 40?

GREG KAUPP: My son is a junior at the University of Minnesota studying data science and computer science. As a Gen Xer, I grew up playing video games, but I would not call myself a digital native the same way I would call my children digital natives.

GREG KAUPP: There is a fundamental difference in how we approach technology. For me, trying to do anything on TikTok, Instagram, or Facebook feels like an unnatural act.

GREG KAUPP: The "over 40" title acknowledges the difference in how we approach this transformative technology compared to college-age kids. My daughter joked that I should have titled the series for the "well over 50."

GREG KAUPP: We are at the tail end of Gen X, and I think we are all facing similar challenges as we try to figure out how to embrace this transformative technology. I was looking for a hook that would get people to look twice.

HOST: It was a good hook. You have written several articles and posted them on LinkedIn. In week one, you mentioned you have been chasing the AI dragon for years. What do you mean by this?

GREG KAUPP: Tesla has been around for a while, and seeing what AI did for self-driving cars made me realize it must be transformative for the type of work we do. I decided to go back for my MBA in 2019 to get a better understanding of what AI was doing.

GREG KAUPP: I went to Kellogg in Chicago right before COVID and graduated in 2021. At that point, we were still in the realm of machine learning, which was not yet translatable to what we are seeing today.

GREG KAUPP: I came away somewhat disappointed. Then ChatGPT came out in 2023. Initially, it seemed like the only thing people were doing was making things sound like Dr. Seuss.

GREG KAUPP: Between the serious hallucinations and the struggle to find real use cases, I decided it was not quite ready. Fast forward to this year at Cherry Bekaert, where one of my big challenges was forecasting.

GREG KAUPP: That challenge forced me back into evaluating how the models had progressed from late 2023 to early 2024. The progress was astounding. Now, I cannot imagine doing what I do without using AI.

HOST: I agree with you, Greg. I remember when I first started using ChatGPT, I could not even remember the name. I looked at it like Google on steroids, which was the wrong way to look at it.

HOST: It reminds me of the 1990s when the internet was becoming popular and people were trying to figure out how to use it. Does this feel like the same thing to you?

GREG KAUPP: I definitely agree, and it is that experience on steroids. The pace of how these models are evolving and progressing is astonishing.

GREG KAUPP: In the early days of the internet, it took a while for the infrastructure to reach a place where we could realize its promise. Now we are seeing the great AI buildout regarding data centers and electricity production.

GREG KAUPP: The amount of money spent on the AI buildout this coming fiscal year is more than what was spent on an adjusted basis for the entire U.S. highway network. The pace of evolution is faster than anything we have seen in our lifetime.

HOST: What is AI not?

GREG KAUPP: I initially treated AI like Google. I would ask a question, get a poor response, and give up. I realized I was thinking of it as a search engine.

GREG KAUPP: It was not until I stuck with it that I realized I need to engage these models in an ongoing conversation. The more context I provide, the better the responses get.

GREG KAUPP: I spoke to another business leader who said AI was terrible because it could not solve a simple math problem. I asked him what role he asked it to play. He just expected it to act like a calculator.

GREG KAUPP: The key insight is that AI is like improv; it will play whatever role you ask. If you do not give it a role, you get a random one. If you ask it to solve a math problem like a mathematician, you will get the right answer.

GREG KAUPP: It is all about an interaction and the "yes, and" idea from improv. Context is everything. It is not all-knowing, but if I ask it to play a certain role, it will mimic that context in a way that looks like understanding.

GREG KAUPP: If I ask it to be a comedian, even though AI does not understand humor, it will sound funny. It is about predicting the next word to convey context and meaning, even if the model itself does not understand what that context represents.

HOST: For me, it is a conversation. I ask questions, disagree with it, and give my own opinion. It is interesting how it will come back with a different perspective or ask if I have thought about another angle.

GREG KAUPP: My wife accused me of having more meaningful conversations with AI than with her. While AI can be very affirming, we have to realize that sycophantic AI does not always serve us well.

GREG KAUPP: The latest models have begun to address hallucinations and sycophantic responses. I have moved from models that were purely affirming to newer versions that are more critical and challenge me to think more deeply.

HOST: You talked about different platforms in your week two blog, including Claude, ChatGPT, Gemini, and Microsoft Copilot. I had never heard of Claude until you mentioned it, but now it is my favorite. Can you talk about the strengths and weaknesses of each?

GREG KAUPP: When I started this journey for better forecasting, my son suggested I experiment with other models beyond Microsoft Copilot. That was eye-opening.

GREG KAUPP: At the time, Claude was much better at helping me reason through problems. ChatGPT was better for extracting and transforming data from Excel files.

GREG KAUPP: Gemini continues to be my go-to for anything I would have previously searched for on Google. Their expertise in indexing the internet provides better responses for searching knowledge bases.

GREG KAUPP: I ran an AI Olympics where I asked the models to produce an output and then grade each other. All the models graded Claude the best. Microsoft Copilot came in a distant third or fourth.

GREG KAUPP: Microsoft Copilot is built on ChatGPT, but the fine-tuning for the Microsoft Office suite can risk "catastrophic forgetting." It does not function as well as a general-purpose Large Language Model like Claude or Gemini.

GREG KAUPP: Gemini shines with its context window. While Claude and ChatGPT have limits around 200,000 tokens, Gemini has a million. I recently bumped against those limits with Claude, which was frustrating.

HOST: I read that article about "token hell." We should touch on that in a future podcast. In your week three blog, you talked about search engines and improv. Do you still use Google for anything?

GREG KAUPP: Honestly, I do not use Google search for anything. This is a fundamental restructuring of the internet economy.

GREG KAUPP: I find I do not need the ad clutter. I go into Gemini, ask my question, and get a response. I also value having my history saved as a second brain.

HOST: You have written 13 of these blogs. Are you planning on a full 52-week series?

GREG KAUPP: I committed to a 52-week series, but it is getting harder. The challenge is saying something that is not already being said everywhere else.

GREG KAUPP: Each week I tackle a new challenge for myself. The series forces me to learn, try something new, and share my personal evolution.

HOST: You are doing a fantastic job, and people are loving these. Since we are both from Cherry Bekaert, what should someone do if they want to get started on their AI journey?

GREG KAUPP: I encourage people to just have a conversation. We have not done our clients a service by talking about agentic AI when we are barely literate in the basics.

GREG KAUPP: I advocate for literacy before agency. We should also prioritize personal transformation before organizational transformation. It is hard to transform an organization if AI hasn't transformed your own job yet.

GREG KAUPP: Data governance is also critical. If you turn on AI without security labeling and governance, you risk exposing confidential information.

GREG KAUPP: We have a team at Cherry Bekaert that focuses on helping people make thoughtful decisions about data governance and labeling. This ensures you do not inadvertently open up sensitive documents to the world.

HOST: Those are valid points. Greg, I really appreciate your time today. I would like to get with you again in the future to talk about your other blogs.

GREG KAUPP: Thank you, Jerry. It has been a pleasure.

HOST: For our listeners, you can find Greg Kaupp on LinkedIn. His last name is spelled K-A-U-P-P. Thanks for joining us today.