5 Ways Real People Are Using AI to Save 5+ Hours a Week

Type: Use Cases / Practical Value

Forget the demos. Forget the press releases. Here's what AI actually looks like in practice — from five people who've quietly rebuilt parts of their workweek around it.

I want to be honest upfront: this isn't a list of miracles. Not everything they tried worked. A few things were actively annoying. But all five of these people have found something that stuck — and more importantly, something transferable.

1. Maya, Secondary School Teacher — Saves ~6 hours a week on lesson prep

Maya teaches history to 13-16 year olds and used to spend Sunday afternoons building differentiated materials for different learning levels. Now she uses Claude to generate three versions of the same explanation — one simplified, one standard, one extended — for any concept she's teaching that week.

She's clear that she still edits everything. 'AI writes at a Year 9 level naturally, which is fine for some students but not all. I always adjust.' But the starting point saves her what she estimates is three to four hours per week. Add in the discussion questions and quiz drafts it generates and she's closer to six.

What didn't work: She tried using it to mark student essays. It wasn't consistent enough, and she didn't trust it to be fair.

Tool she uses: Claude. Free tier, browser-based.

2. James, Freelance Graphic Designer — Saves ~4 hours a week on client communication

James's problem wasn't his design work. It was the emails. Scope clarifications, revision requests, project updates, difficult conversations about budget. He found these stressful and time-consuming to write.

Now he pastes the context into an AI, describes what he wants to say and the tone he's going for, and gets a draft. He edits it, sends it. Occasionally he uses it almost unchanged.

'It's not about the time, really,' he told me. 'It's about the mental energy. I used to dread those emails. Now I just... deal with them.'

What didn't work: He tried using AI for creative concepting and found it too generic. His instinct is right — for visual creative work, it's a weak assistant at best.

Tool he uses: ChatGPT Plus. $20/month.

3. Priya, Small Business Owner (Online Retail) — Saves ~5 hours a week on content

Priya runs a small business selling handmade ceramics. She needs regular content: product descriptions, Instagram captions, email newsletters, the occasional blog post. She used to write it all herself, which she hated.

Now she gives AI a product, a photo description, and her brand voice notes, and gets five caption options to choose from. She picks one, tweaks it, posts it. Newsletter drafts take her 20 minutes instead of two hours.

'I write better than it does,' she says. 'But I don't have time to write everything well. This gets me to 80%, which I can get to 95% in ten minutes.'

What didn't work: Early on she let captions go out without editing. They were fine but they didn't sound like her. The editing step is non-negotiable.

Tool she uses: ChatGPT, free tier. Considering upgrading.

4. Tom, Software Developer — Saves ~7 hours a week on boilerplate and debugging

Tom is a mid-level developer and was initially sceptical. He's now one of the most enthusiastic people I've spoken to about AI tools.

He uses GitHub Copilot for in-editor suggestions and Claude for longer debugging sessions. 'I paste in an error, paste in the relevant code, and say what I was trying to do. Nine times out of ten it spots the issue faster than I would.'

The bigger gain has been boilerplate: configuration files, test cases, data transformation scripts. Things he knows how to do but finds tedious. 'I just describe what I want and review the output. It's not perfect but it's a strong first draft.'

What didn't work: He does not trust it with security-critical code. 'It'll write you something that looks right but has a subtle vulnerability. You need to know what you're looking for.'

Tools he uses: GitHub Copilot ($10/month) and Claude Pro.

5. Amara, Nurse — Saves ~3 hours a week on documentation

Amara's use case is the most careful on this list, and she insisted I include the caveats.

She uses AI to help structure her written documentation notes after shifts — not to generate clinical content, but to turn fragmented bullet points into coherent prose. She reviews everything, always. Nothing goes into a patient record without her full read.

'Nurses spend an enormous amount of time on documentation. Anything that helps me do it faster without cutting corners is worth exploring.' Her trust level is high but her verification habit is absolute.

What didn't work: She tried using it to help answer clinical questions. 'It sounds confident about things it shouldn't be confident about. For anything clinical, I use actual medical resources.'

Tool she uses: Claude, browser-based. On personal device only, not work systems.

How to Find Your Own Use Case

The pattern across all five: they found one task that was repetitive, time-consuming, and didn't require their unique judgment — and they applied AI there first.

Start with your most tedious recurring task. The one you put off. The one that's important but draining. Give AI a try on that one thing. Don't try to overhaul your whole workflow. One task, genuinely tried, is worth more than five shallow experiments.

Next week: a strong opinion that I think the whole AI jobs debate is getting wrong — and what that means for what you should actually be doing right now.