I Tracked Every AI Tool I Used for 30 Days — The Results Shocked Me
I decided to do an experiment: for a whole month, I logged every AI tool I used, how long I used it, and what for. I expected maybe a nice efficiency boost, but the data told a wild story of huge time-savings, surprising costs, and new habits I never saw coming.
My AI Toolkit
Over 30 days I tried everything: ChatGPT (OpenAI’s GPT-4), Google Bard, GitHub Copilot (for coding), Microsoft 365 Copilot (in Word/Excel), Midjourney and DALL·E (for images), Notion AI (for notes), Jasper AI, and a few others. I often combined them: e.g. drafting an email in ChatGPT, then refining in Grammarly, or generating a blog image with Midjourney. Here’s a quick rundown of each tool’s role, with one example input→output:
- ChatGPT (OpenAI, GPT-4/GPT-4.5): My all-purpose helper. For example, I asked it, “Write a polite reminder email about next week’s meeting.” It produced a coherent email draft in seconds, saving me ~10 minutes versus writing from scratch. I used it almost daily for tasks like blogging, brainstorming, and even light coding help.
- Google Bard (PaLM 2): A quick answer generator. I’d use it for fact-checking or when I needed a second opinion. For example, “List recent stats on remote work trends.” Bard spit out bullet points from recent articles. It was free to use (no subscription), but I noticed it often took a few passes to refine its output.
- GitHub Copilot: My coding copilot. When I needed a code snippet or to refactor logic, I’d write a comment or a function signature and Copilot would autocompile code. For example, after writing
// parse JSON file, Copilot suggested a ready-to-use parser. It felt like having a junior dev assistant, saving maybe ~30–50% of my usual coding time. - MS 365 Copilot: Integrated into Word/Excel. I tried this for tasks like drafting a report summary or cleaning spreadsheet data. For instance, I uploaded a budget Excel file and asked Copilot to highlight anomalies. It did, which was impressive. Note: Copilot Chat is included at no extra cost with eligible Office subscriptions.
- Midjourney: AI image art. Whenever I needed a unique image (e.g. a blog header), I described a scene and Midjourney generated beautiful visuals. For example, “A futuristic cityscape at sunset” yielded a stunning result. I had a paid plan (~$30/mo) for unlimited fast generations.
- DALL·E 3: Another image tool (via OpenAI). I used this for quick illustrations. E.g. “Draw a cat wearing glasses reading a book” produced a fun cartoon cat. It cost ~¢2 per image on API, so I used it sparingly.
- Notion AI: Built into Notion notes. It helped auto-fill meeting notes, summarize pages, etc. For example, I pasted a meeting transcript and asked Notion AI “Summarize key action items”, and it did so nicely. Notion AI is included in certain paid plans.
- Jasper AI: A marketing copywriter. I tried it for a tweet or an ad draft. Its strength is writing with a specific tone. For example, I gave it a blog outline and asked for an “energetic summary”, and it delivered marketing-sounding text. Jasper Pro runs ~$59/month, so I only used it when I needed a polished spin.
Each tool had a learning curve: ChatGPT took a day or two to get prompts right. GitHub Copilot also had a brief setup in my IDE. Others (like Bard or Notion AI) were quite intuitive. In total I logged ~25 distinct sessions across these tools each week, averaging a few minutes to 30 minutes per session.
Get Weekly Reporters’s stories in your inbox
Join Medium for free to get updates from this writer.
30-Day Quick Stats
I kept a simple daily log. Here’s a condensed summary (each day, which tools I used and for roughly how long):
- Day 1: ChatGPT (20m, email draft), Midjourney (15m, blog image)
- Day 2: ChatGPT (30m, blog outline), Copilot (30m, code review), Notion AI (15m, meeting notes)
- Day 3: Bard (10m, quick facts), ChatGPT (25m, report outline), DALL·E (10m, illustration)
- … (Similar mix each day, totaling about 2–3 hours of AI use per weekday, less on weekends.)
- Day 30: ChatGPT (25m, report drafting), MS Copilot (15m, slide generation).
Across 30 days, I logged roughly 15 hours of AI usage. That included every session, big or small. On average, sessions were quite short: maybe 15–20 minutes each for writing tasks, 10–15 for images, and about 30–60 minutes for coding sessions.
Patterns and Visuals
The data above revealed some clear patterns:
- ChatGPT dominated my usage. I leaned on it every single day for writing and ideas. It was my go-to brainstorm partner.
- Coding was faster. GitHub Copilot and MS Copilot cut my coding and data tasks roughly in half (for example, debug a script or format a report). I estimate I saved ~10–15 hours over the month just there.
- Images were occasional. I didn’t need images daily, but when I did (social posts or blog headers), Midjourney’s creative control and DALL·E’s convenience made it quick.
- Privacy mattered. In sensitive tasks (like drafting private emails or handling client data), I consciously avoided Bard or unprotected tools. Notion AI and Office Copilot felt safest due to their privacy design.
- Learning curve was short. Most tools took 1–2 uses to feel natural. The biggest delay was picking up Copilot syntax in code or the right descriptive prompts for images.
- ROI was huge. Conservative estimate: I spent ~15 hours using AI. How much time did it save? I’d guess 3–5 times that. Writing that weekly report took 10min with AI vs. an hour by myself
