The AI tools landscape changes rapidly. What works today might be obsolete tomorrow. I’m testing new tools as they emerge or we identify a task where an AI tool might help.

I tend to focus on finding tools that genuinely accelerate a specific workflow rather than trying every new AI app that launches, which is a quick way to experience AI overload.

The AI tools I use are ones that solve actual problems in my personal workflow, not those that solve imaginary problems or duplicate existing capabilities of the tools we use or some good old critical thinking.

The goal isn’t to use AI everywhere. It’s to use AI where it genuinely makes things better, faster, or possible that weren’t before. That’s a much higher bar than most people set, but it’s the only way to build a sustainable, productive relationship with these tools.

Principles for assessing AI tools

Follow the “CEO” (Check, Edit, Own) principle. Check, Edit, Own every AI output. Tools that make this difficult are ineffective and carry more personal risk when trusting their outpit.

Context is key. Raw AI outputs aren’t enough. You need to process the outputs alongside the relevant context to maximise value.

Voice input transforms productivity. Dictation tools multiplies your ability to input AI tasks. Try speaking to an AI rather than typing a prompt. This has the added bonus of engaging your brain to clearly communicate what you’re trying to acheive.

Create your own AI workflows. Custom shortcuts and micro-apps that connect AI tools to your person workflow matter more than any individual AI tool’s capabilities.

Speed versus depth is a trade-off. Quick tasks don’t need heavyweight solutions. Recogniising this helps you choose the right tool for the job.

First drafts matter. Even poor first drafts provide value by overcoming blank page syndrome and creating something to improve upon.

Iteration beats overprompting. Shorter, sharper prompts that generatre shorter outputs, then re-prompting to create an output loop, generally create more valuable outputs than single, larger prompt processes.

Not every feature is useful. An AI tool’s capabilities might be impressive in demos but aren’t always practical in real workflows. You have to test each feature personally to see which ones have the most value.

Use AI to help you use AI. For more complicated tasks, you can even use AI to improve how you work with AI tools.

My AI toolkit

After testing a range of AI tools, here are the ones that work for me and help improve my daily work.

Claude has become my primary tool for substantial work. It excels at deep thinking, ideation, and problem-solving in ways that feel genuinely collaborative, iterating on ideas to get them where I need the to be. Claude is great for document analysis, learning from uploaded files, and all forms of writing and editing.

Granola‘s Meeting transcripts and voice memos create a powerful knowledge capture system for meetings. Transcripts alone aren’t enough, they need processing alongside agendas and project plans to maximise their impact, whihc Granola’s template handle well.

Note: More tools to be added.