AI tools at work: what’s saving time and what’s just adding steps
Everyone has an AI tool recommendation right now. Most of them are genuinely useful. Some of them have quietly added 20 minutes to your day while making you feel productive.
I’ve been using AI tools across three different work contexts — a corporate job, an accounting practice, and a mortgage broking business — so I’ve had a decent chance to see what actually sticks and what doesn’t.
The ones that genuinely save time
Writing first drafts is the obvious one, and it’s real. The blank page problem is gone. Whether it’s a client email, a policy summary, or a blog post (yes, including this one’s first pass), having something to react to is faster than starting from nothing. The key is that you still have to edit. If you’re publishing AI output without reading it properly, the quality shows.
Summarising long documents is underrated. I used to dread reading through lengthy contracts, policy updates, or ATO rulings. Now I paste them in, ask for the key points, and read the full thing only if something flags. It’s cut my reading time on dense documents by more than half.
Transcription and meeting notes have also been a genuine win. Tools that sit in a meeting, transcribe it, and give you a summary with action items are doing something that used to take 30 minutes of re-listening. That time is just back now.
The ones that added steps without admitting it
Prompt engineering rabbit holes. I’ve spent 15 minutes crafting a prompt to get output that would have taken me 10 minutes to write myself. There’s a skill to prompting well, but there’s also a point where you’re just procrastinating with extra steps.
AI image generation for anything professional. The output still looks slightly wrong in ways that are hard to name, and the iteration time to get something usable is longer than sourcing a decent stock image or briefing a designer.
Chatting with AI instead of deciding. This is the subtle one. Using a language model to “think through” a decision that you’ve already made, or that you just need to make, is a form of productive-feeling delay. The tool reflects your uncertainty back at you with more words. It doesn’t resolve it.
What I’ve actually kept
My current setup has Claude for structured tasks — anything that needs logic, drafting, or document work. Gemini handles my personal knowledge base and long-form note processing. Local models on my Mac for quick things I don’t want going to a server.
The common thread in what I’ve kept is that each tool has a defined job. The ones I dropped were the ones I was using for everything, which meant they were optimised for nothing.
The honest take
AI tools are genuinely useful. But the productivity gains aren’t automatic — they come from figuring out the 20% of tasks where the tool is clearly better than you doing it manually, and ignoring the hype around the other 80%.
If you’re still trying everything that gets posted on LinkedIn, that’s the phase. Most people go through it. At some point you stop collecting tools and start using fewer of them properly.
