365 Days of Diving into AI Headfirst: A Reflection
In February 2024, I made a few life-changing decisions, one of which was to dive head-first into the world of AI and figure out what the next iteration of work technology would look like. After a year of doing this, I am starting to see patterns emerge on what makes customers buy, what makes HR and work tech companies successful, and changes in my investment mindset.
Among all of these learnings (and let me know if you are interested in learning more about them; I’m on the fence as to how interested anyone truly is about the rabbit holes I tend diving down), one of the biggest “Aha!” moments I’ve had in the last year is on my relationship with AI.
The beauty of content creation is that I get to go back and read my perspective from other points in time. Let’s just say that when 2025 Lydia reads perspectives from 2024 Lydia, the first thought that comes to mind was “Aww. That’s cute” (read: Wow! You were naive). So, after a year of diving headfirst into the AI for HR rabbit hole, I will share some of my, perhaps more mundane, learnings and appreciations.
1. “What’s that?” became my most asked question
I was pretty quick to pick up and understand HR tech back in the early 2010s when I was in consulting. After that, it took me a couple of years to go from zero to building an industry best practice People Analytics organization. So, when I first came into the AI space, I thought “Well, how complicated can this possibly be?” Yea, it was (and still is) complicated.
After spending most of my days diving into AI’s history, latest developments, releases, critics’ comments, etc. I realized sometime in mid- to late-2024 that the biggest difference between AI and previous iterations of technology in HR is that before, HR picked up the tech after the tech community and enterprise communities have already figured it out and stabilized most of the ecosystem. Meanwhile, with AI, the academic, tech, enterprise, HR, and consumer markets are all figuring it out simultaneously. And “messy” doesn’t even begin to describe the chaos and pace of change that is happening.
In realizing that, I have also come to a few opinions/conclusions:
I get why AI is so confusing to HR buyers and users. I spend most of my days looking at this stuff, and I’m even feeling whiplash from how fast things are happening. I can only imagine what this feels like to an HR professional who is working 10-12hrs a day and then trying to find time on top of that to figure out this AI thing
It’d be irresponsible for anyone to claim to be an expert of AI in HR at this stage. Some of us spend more time and effort in the field than others, but at the rate of current developments, the word “expert” in it’s traditional sense should be used and received with abundant caution
I will not be able to know everything, need every AI tool, or have to use most of the AI tools out there. This is as much a figure-it-out-as-you-go process for me, as it is for most out there
2. Habits and favorites will develop over time
Just as there are Chrome and Safari people, there are camps for favorite/most used AI tools. After a year of signing up for AI tool accounts, trying a bunch of things out, and exchanging feedback and experiences with other users, I can honestly say that there is no right or wrong answer when it comes to what AI tools you use/prefer to use.
As with most things, my “what do you use?” conversations will always boil down to a matter of preference. Unless you are doing substantial and large volume work with AI tools, the difference between LLMs is negligible and most of the time your choice of tools will come down to its UI and UX (i.e., did the provider really think about how they want users to engage with the platform and design it with that intention in mind?). Some of the best tools I’ve come across have such clunky UIs that I never want to touch them again, yet some of the simplest tools with better UIs are the ones I choose to go back to over and over again.
So, instead of figuring out if you are “on trend” with your AI tech stack, spend the time and effort trying things out before you commit and think about how the tool can/will fit into your personal and professional realms.
3. AI saves time, but doesn’t get rid of the task
This is actually the bullet that made me want to write this whole article. I spent most of my day yesterday writing content copies for a few projects I’m working on. If this was pre-2022, I would have probably only gotten through one set of the content and hired a copywriter to do the other few given the turnaround time.
So, between hiring someone to do a small task and using AI to do the work, I saved some money by doing this myself with AI. It also allowed me to get more done in a shorter time. I think it’s a pretty typical value prop for the use of AI in everyday work.
However, the one thing I did realize at the end of the day was that while AI saved me time and money, it didn’t get rid of this thing I was supposed to do.
I don’t do enough copy content to warrant setting up and training AI agents to do the job. So I wasn’t about to realistically spend a few days to get an agent up and running to save a few hours in copy writing (that’s an example of bad ROI if there ever was one). Like most people, I needed some extra help with a side-of-desk task that had to be done. So, I used AI like an assistant, but I was doing most of the “navigating”.
This brings me to the finer point of ROI on AI that I don’t hear most people talking about: as the technology stands today, unless there is a highly repeatable and large volume task that you can move to AI entirely, it’s tough to showcase a dramatic difference in the before/after of implementing AI. You’re working better and faster, but the human still needs to be at the wheel…just in a different capacity.
4. It needed a new set of communication and thinking skills
Speaking of different capacities, this brings me to my next learning. I don’t know if this is just me, or if everyone else is experiencing it and we’re just not talking about it, but talking to a machine is a whole different skillset that has to be built over time.
When I first started, I would work with ChatGPT (my tool of choice) the same way I would Google search something. I would ask a general question and expect some responses that would help guide my next steps. I realized mid-2024 that wasn’t the smartest way of doing things because the answers I got were super vague, and quite honestly not worth the $20/month I was paying into it.
Then I started figuring out how to talk to AI to get what I needed. The best way I can describe this is that it’s like starting with a highly intelligent analyst every time. They know a lot about everything, but nothing about what I was trying to do. So, I have to understand how to give them the background, engage them in the conversation so the flow of thought can go down the path I need, etc. Also, I had to remember my “please” and “thank you”s to motivate them and keep them going.
So, all said and done, I’ve gotten much better at giving context, background and explicit instructions over the past year. I’ve also gotten a lot better at describing something even if I don’t have the right words for it.
(Shoutout to Nathalie Salles-Olivier for making me talk to her agents as if they were real people. It was weird to start, but it actually helped me understand how to best work with LLM interfaces)
5. Knowledge is $$
Ok, I’m going over my word count a bit here, so I’ll make the last one short and sweet: in exploring the market for AI tech, I have also come to appreciate the complexity of the tech ecosystem. In the era of spreadsheets, you had to buy the software from one vendor who developed the tool, stored your data, and did most things in-house. In the era of ERPs, there were a few third-party processors you had to be aware of. In the era of AI, almost no one owns their ecosystem end to end. Winning the tech game today is more like “who is doing a better job of puzzling all the upstream pieces together and bringing it to the downstream customers?”, than “who is building the best tech?”.
So, with that in mind, we are in an era where knowledge holds more power than ever before, and knowledge arbitrage is what’s going to drive actual monetary and market arbitrage opportunities.
Something for you to noodle on until next time 😊