You don’t have to use generative AI. It’s possible to avoid it and continue doing whatever you’ve been doing, the way you’ve been doing it. I don’t believe that sentence will be true in twelve months. Not because you’ll have to use it—although in some cases it may be unavoidable—but because you’ll want to use it. I thought about how my work will change next year.
#1. I’ll start most efforts by asking “can AI help with this?”
Do I need to understand a new market or product area? Analyze a pile of data? Schedule a complex series of meetings? Quickly generate a sample app for a customer demo? Review a blog post a teammate wrote? In most cases, AI can give me an assist. I want to change my mental model to first figure out if there’s a smarter (Ai-assisted) way to do something.
That said, it’s about “can AI help me” versus “can AI do all my work.” I don’t want to end up in this situation.
#2. I’m going to do much better research.
Whether planning a strategy or a vacation, there’s a lot of time spent researching. That’s ok, as you often uncover intriguing new tangents while exploring the internet.
AI can still improve the process. A lot. I find myself using the Gemini app, Google AI Studio, and NotebookLM to understand complex ideas. Gemini Deep Research is almost unbelievable. Give it a prompt, it scours the web for dozens or hundreds of sources, and then compiles a report.
What an amazing way to start or validate research efforts. Have an existing pile of content—might be annual reports, whitepapers, design docs, or academic material—that you need to make sense of? NotebookLM is pretty amazing, and should change how all of us ask questions of research material.
#3. I will learn new things faster.
Many of us have jobs where we need to quickly get up to speed on a topic. I want help in context, so that I stay in a flowstate.
Back to NotebookLM, I might use this to get easier-to-digest audio overviews of complex new ideas.
And then with coding assistance tools, I also am getting more and more comfortable staying in my IDE to get help on things I don’t yet know. Here, my Gemini Code Assist extension is helping me learn how to fix my poorly-secured Java code.
Finally, I’m quite intrigued by how the new Gemini 2.0 Multimodal Live API will help me in the moment. By sharing my screen with the model, I can get realtime help into whatever I’m struggling with. Wow.
#4. I’ll less time debating and more time coding.
My day job is to lead a sizable team at Google Cloud and help everyone do their best work. I still like to code, though!
it’s already happening, but next year I expect to code more than in years past. Why? Because AI is making easier and more fun. Whether using an IDE assistant, or a completely different type of IDE like Cursor, it’s never been simpler to build legit software. We all can go from idea to reality so quickly now.
Stop endlessly debating ideas, and just test them out quickly! Using lowcode platforms or AI assisted coding tools, you can get working prototypes in no time.
#5. I will ask better questions.
I’ve slowly learned that the best leaders simply ask better questions. AI can help us a few ways here. First, there are “thinking” models that show you a chain of thought that might inspire your own questions.
LLMs are awesome at giving answers, but they’re also pretty great at crafting questions. Look at this. I uploaded a set of (fake) product bugs and asked the Gemini model to help me come up with clarifying questions to ask the engineers. Good list!
And how about this. Google Cloud BigQuery has an excellent feature called Data Insights which generates a bunch of candidate questions for a given dataset (here, the Google Cloud Release Notes). What a great way to get some smart, starter questions to consider!
#6. I want to identify where the manual struggle is actually the point.
I don’t want AI to do everything for me. There are cases where the human struggle is where the enjoyment comes from. Learning how to do something. Fumbling with techniques. Building up knowledge or strength. I don’t want a shortcut. I want deep learning.
I’m going to keep doing my daily reading list by hand. No automation allowed, as it forces me to really get a deeper grasp on what’s going on in our industry. I’m not using AI to write newsletters, as I want to keep working on the writing craft myself.
This mass integration of AI into services and experiences is great. It also forces us to stop and decide where we intentionally want to avoid it!
#7. I should create certain types of content much faster.
There’s no excuse to labor over document templates or images in presentations anymore. No more scouring the web for the perfect picture.
I use Gemini in Google Slides all the time now. This is the way I add visuals to presentations and it saves me hours of time.
Generate code, docs, and images, sure. We’ve seen that, but the image generation tech is getting tremendous.
But videos too? I’m only starting to consider how to use remarkable technology like Veo 2. I’m using it now, and it’s blowing my mind. It’ll likely impact what I produce next year.
#8. I’m going to free up some valuable time.
That’s what most of this is all about. I don’t want to do less work; I want to do better work. Even with all this AI and automation, I expect I’ll be working the same number of hours next year. But I’ll be happier with how I’m spending those hours: learning, talking to humans, investing in others. Less time writing boilerplate code, breaking flow state to get answers, or even executing mindlessly repetitive tasks in the browser.
I don’t work for AI; AI works for me. And in 2025, I’m expecting to make it work hard!
Richard Seroter's Architecture Musings 
Leave a reply to kimchanwoo Cancel reply