My top 3 AI tools of 2025
End of year means time to reflect. Here are the three tools that genuinely changed how I work in 2025. The ranking wasn’t even close:
- Cursor
- Claude
- NotebookLM
Let me walk you through each one.
3rd place – NotebookLM
Most people know NotebookLM for its podcast or presentation generation. I use it primarily for two things: rapid research and processing large volumes of data.
The rapid research part fits into what I call my “Power Hour” – a way to build almost anything in 60 minutes using AI, whether that’s a board report, a client presentation, or a working app. The process has three steps – Research, Design, Build – and NotebookLM handles step one.
Here’s how it works in practice. Say you want to generate images for your website using Nano Banana. Instead of testing one prompt after another, you run a research query: “Find me the best methods for writing prompts for AI image generation in Nano Banana as of late 2025.” Within minutes, you have dozens of sources in one place, and a single follow-up prompt synthesizes everything into a structured overview. From there, you feed it into your assistant and start creating.
The second use case is straightforward: I need to quickly extract information from dozens of documents, or pull a specific technique from an hour-long video. Rachel Wolan from Webflow, for example, shared how she uses AI agents to manage her calendar and email. I paste the video link into NotebookLM and ask it to describe what’s covered. Within thirty seconds I know whether it’s worth watching in full – or whether I can just work with the key ideas directly.
I also use the podcast feature regularly. After each module of Future AI Leader – my intensive program for professionals who want to genuinely master AI – I process participant inputs and generate a motivational recap that I share as both audio and video. When I finish a major presentation, I turn it into a podcast and listen to it at the gym. Hearing “strangers” discuss your own content always sparks new ideas. I also try more unconventional scenarios – exporting a task list and turning it into a podcast, or feeding in transcripts from the past few weeks of team meetings. It’s surprisingly revealing to hear how AI interprets your own priorities.
2nd place – Claude
Every AI tool has its own skill set, features – and personality. Over the course of the year, it became clear that Gemini has improved significantly, while ChatGPT is starting to fall behind. That’s exactly why being fluent in multiple tools matters.
For me, Claude from Anthropic leads by a wide margin.
There are several reasons, but the most important one is this: it’s exceptionally intelligent, it understands how I think, and we share a common way of working through problems. On top of that, it offers the most advanced data analysis capabilities and two features that have become indispensable for serious AI work: strong coding ability for non-programmers like me, and deep integrations with other apps and services.
Two features I rely on most:
Projects function like a dedicated AI assistant for a specific project or recurring task – an email generator, a prompt builder, a research assistant. You store your instructions and relevant documents there, pin it in your browser, and it’s always at hand.
Artifacts are a working canvas alongside the chat where you can build documents, interactive reports, diagrams, or presentations – and also custom micro-apps with built-in AI functionality. These artifacts can be shared directly with your team or clients, which I use constantly.
A few weeks ago, Claude Opus 4.5 arrived – the most intelligent model I’ve ever worked with. I’ve developed a mild dependency on it, which is starting to show up in my monthly bill.
My subscription runs about $90 per month, with model usage in other applications adding roughly $300 per month on top of that. That might sound steep. But Claude handles work that would otherwise require one or two additional team members.
1st place – Cursor
This one wasn’t a decision – it was obvious.
Cursor is the application where non-programmers like me suddenly start “programming.” You give the AI a goal and it handles the execution. Andrej Karpathy, one of OpenAI’s co-founders, calls this vibe coding.
And it’s not theory. Websites, marketing materials, graphics, surveys – I built all of it in Cursor this year. Two full projects in a single weekend.
Here’s what I actually do with it:
File and data work – Need to split data into dozens of individual files? Merge data from five different spreadsheets? Convert documents into a different format? I tell Cursor and it gets done.
Agents for recurring tasks – I have an agent that, after every Future AI Leader workshop, processes the transcript, creates a summary, drafts YouTube copy, and pulls out the key takeaways. I have dozens of these agents running. I don’t need to remember any of them – I just describe what I need and Cursor figures it out.
Rapid prototypes and presentations – Before a client meeting, I have an interactive presentation built instead of a static PowerPoint deck. The reactions are consistently enthusiastic.
Why is this genuinely different? Chatbots can write code, but they don’t sit on your machine – they can’t work directly with your local files. And when they do produce a script, you still have to figure out where to run it. In Cursor, everything lives in one place.
Last week, mid-presentation, I had a live idea: a system that transcribes key points from a talk in real time. Cursor designed the solution, connected to Deepgram for real-time voice transcription, built a clean display page for the output, and drafted the vendor email – all in a single session. Work that used to take weeks now takes a few prompts.
Cursor isn’t the only tool in this space – Claude Code, GitHub Copilot, and others work similarly. The right fit depends on your workflow, so it’s worth testing a few. One fair warning: these tools were originally built for developers, and they look the part at first. Change the font size, adjust the color scheme, and it becomes an entirely different experience.
This is the future of knowledge work. And that future is already here. The best part? With the right tool, you can step into it in a single afternoon.
FD
