8 Practical ways I use AI as a Software Engineer (2025)
I’ve been using AI daily in my work and personal projects as a software engineer. Here's how I'm using AI to build faster and more than without AI so you can, too.
AI has become a practical tool that helps me start, unblock, and finish projects. If you’re a developer who is curious about AI but not sure where to start, this is for you. These are the concrete ways I use AI to get more done, not as a magic wand, but as a powerful assistant. Here are some of the strategies I've found most effective.
1. Let the AI Write Its Own Prompts (Meta-Prompting)
This is one of the most powerful, yet less obvious, techniques. When an AI's output isn't quite right, it’s tempting to endlessly tweak your prompt. Instead, I tell the AI to fix the prompt for me. I provide my original prompt, explain the flaws in the output it generated, and ask it to create a better prompt that will avoid those issues. This is the fastest way I’ve found to get a high-quality prompt that I can reuse. You let the model's own logic do the heavy lifting of figuring out how to be instructed.
2. Defeat "Blank Page" Procrastination
We all have tasks that we put off. The activation energy required to start just feels too high. When I feel that resistance, I turn to Gemini and describe the task. I don’t ask it to do the work, but to break it down for me. "Give me a step-by-step plan to accomplish X."
This simple conversation helps in two ways:
The act of explaining the problem helps clarify it in my own mind.
I get a concrete, numbered list of actions that I can follow to get started and build momentum.
3. Use It as an Ideation Partner
Sometimes the hardest part of a project is shaping the initial idea. I use a combination of tools for this. I'll collect my thoughts, research, and planning documents in NotebookLM. I might use Gemini Deep Research to explore a topic more thoroughly. Once I have a source collection, I can ask questions against all of that information. This helps me synthesize the different pieces and understand all the angles of the project. It's like having a conversation with your own research. The AI might offer suggestions or connections I hadn't seen, and my expertise is crucial for sorting the valuable insights from the noise. The process helps me formulate my ideas and see the bigger picture more clearly.
4. Get Up to Speed on New Technology, Fast
Technology moves quickly, and ramping up on a new stack, library, or architecture can be time-consuming. I use AI chat to accelerate this process. I can ask specific questions like, "What is the most current, idiomatic way to handle state management in a React project with these specific constraints?" or "Explain the key differences between these two database technologies for this use case." It gives me a synthesized starting point that is often faster and more direct than sifting through documentation and tutorials.
5. Accelerate Data Analysis in Colab
Many software engineering tasks involve some form of data analysis. If you've ever had to dig into a dataset to understand a trend or debug an issue, you know how much time can be spent looking up the right pandas functions. I've found the Gemini integration in Colab to be a huge time-saver here. I can just describe the analysis or transformation I need in plain English, and it will generate the Python code for that cell. It’s perfect for getting the syntax right without breaking my flow to search through documentation.
6. Create Single-Page Web Tools in Minutes
Sometimes you need a small, specific tool—not a script, but a simple web page you can access from anywhere. It could be a little calculator for a project-specific formula or a simple data visualizer. These are often things I wouldn't have bothered to build because the overhead of setting up a web project felt like too much for a small utility. Using AI in a tool like Canvas, I can describe the single-page application I want, and it generates the complete HTML, CSS, and JavaScript. This makes it incredibly fast to create and deploy useful web-based utilities that I can use and share.
7. Create Better Documents, Faster
Writing documentation, design docs, or team updates is a critical part of engineering. I use Gemini to help me generate these documents more efficiently. The key is to provide it with a style guide. I have a personal style guide that defines my voice, and my team has templates for technical documents. By feeding these guides into the prompt, I can ensure the output is not only well-written but also formatted correctly and consistent with our team's standards.
8. The Obvious One: AI Code Assistants
This is the most common use case, but it’s still worth mentioning. AI code assistants are incredibly good. If you've tried one and weren't impressed, I encourage you to try another. They are not all the same. For me, these tools have been the key to unlocking personal projects. The ability to quickly do things like:
Generate boilerplate
Write tests
Refactor code
has made it so much easier to build websites and applications in my spare time, enabling a level of personal productivity I didn't have before. My personal projects are linked on my website, mbdr.ai. Check it out to see what these AI coding tools helped create.
The Tools I Use
Just to be specific, here are the primary tools I use for the tasks mentioned above:
Gemini, NotebookLM, and Gemini Deep Research: For ideation, planning, research, and document drafting.
Gemini in Colab: For accelerating data analysis.
Gemini Canvas: For generating single-page web applications.
Code Assistants (like Cursor, Lovable, and the Gemini CLI): For everyday coding, refactoring, and testing.
What Will You Try?
AI is not about replacing the engineer's judgment but augmenting it. It’s a tool that can help you overcome inertia, learn faster, and ultimately, build more. The best way to understand its potential is to simply start using it for small, practical tasks.
So, what will you try first?
Disclaimer: I work at Google, but the opinions stated here are my own. I personally pay for my subscriptions to Gemini Pro and other AI tools on my personal accounts.