You're Probably Automating the Wrong Things
The mistake most business owners make with AI isn't moving too slow. It's automating things that feel annoying instead of things that actually eat time.
You build a ChatGPT prompt to write Instagram captions, saving yourself 20 minutes a week. Meanwhile, you're spending 4 hours every Friday manually compiling a report that could be fully automated. The captions are more visible. The report is more valuable.
An AI audit fixes this. It's not complicated. One focused day of tracking, then an hour of analysis. The output is a prioritised list of exactly what to automate and in what order.
Here's how to do it.
Step 1: Track Everything for One Day
Pick a normal workday, not your best day, not a disaster day. A regular Tuesday.
Every time you switch tasks, write down:
- What you just did
- How long it took (rough estimate is fine)
- Whether it was something you've done before in a similar form
Don't overthink the categories yet. Just capture. Use whatever's lowest friction: a notes app, a voice memo, a spreadsheet. The tool doesn't matter. Consistency does.
At the end of the day, you should have a rough log that looks something like this:
| Task | Time | Repeatable? |
|---|---|---|
| Responded to 3 client emails | 45 min | Yes |
| Updated weekly report | 90 min | Yes |
| Reviewed contractor invoice | 15 min | Sometimes |
| Wrote this week's newsletter | 2 hrs | Yes |
| Dealt with a weird billing issue | 30 min | No |
You're looking for the "Yes" column. Those are your targets.
Step 2: Identify What's Repeatable
Go through your log and mark everything that:
- Follows a pattern or template
- Requires gathering info from the same sources each time
- Involves writing something you've written a version of before
- Is essentially data transformation (input A goes in, output B comes out)
These are automatable in principle. The question is just which tools fit.
Don't mark things that require real judgment, relationship context, or creative decisions that change significantly each time. AI can assist with those, but full automation isn't the goal there.
Step 3: Match Tasks to Tools
Not all AI tools are the same. Here's a rough map:
Writing and drafting: Claude, ChatGPT, Gemini Good for: emails, reports, proposals, social posts, SOPs, meeting agendas. Feed it a template and your notes; it produces a structured draft.
Transcription and summarisation: Whisper, Otter.ai, Fireflies Good for: turning meeting recordings, voice memos, or calls into notes and action items. Saves enormous amounts of time if you're in a lot of meetings.
Research and information gathering: Perplexity, Claude with search Good for: competitor research, summarising long documents, pulling key facts from PDFs or web pages. Not a replacement for deep analysis, but speeds up the first pass significantly.
Data and reporting: ChatGPT Advanced Data Analysis, n8n + AI, custom scripts Good for: anything where you're manually moving numbers around. If you're copy-pasting data from one place to another, there's probably a better way.
Workflow automation: Zapier, Make, n8n Good for: connecting tools together, triggering actions based on events, routing information. Often the glue between AI tools and your existing stack.
Go back to your repeatable task list and write the tool next to each one.
Step 4: Calculate the Time Savings
Pick your top 3 highest-time repeatable tasks. For each one:
- Estimate current time per occurrence
- Estimate how often it happens per week/month
- Estimate realistic time with AI assistance (not zero, AI outputs need review)
- Calculate the gap
Example:
- Weekly report: currently 90 min/week. With AI drafting from a template: 20 min to set up the prompt + 10 min to review output = 30 min/week. Savings: 60 min/week.
- Client email responses: currently 45 min/day. With AI drafts to edit: 15 min/day. Savings: 30 min/day, ~2.5 hrs/week.
- Newsletter: currently 2 hrs/week. With AI-assisted drafting from voice notes: 45 min/week. Savings: 75 min/week.
Total weekly savings in this example: roughly 5 hours.
Now put a dollar value on your time. If your billable rate is $150/hour, that's $750/week in reclaimed capacity. Per year: ~$39,000 in time value, not counting what you do with those 5 hours.
Most people underestimate this. Run the number. It usually changes how seriously you take the setup work.
Step 5: Prioritise and Start Small
Here's the order of operations that actually works:
First: Pick the single highest-time, highest-frequency task. Set up one AI workflow for it. Use it for two weeks. Get the kinks out.
Second: Once that's running smoothly, add the second one.
Don't: Build five automations at once. They'll all be half-functional, you'll lose track of which one does what, and when something breaks you won't know where to look.
The trap is the setup. Most AI workflows take longer to configure than one-off manual work. That's fine. You're amortising that setup cost over months or years of use. But it means you need to finish the setup before you see the return.
Pick one. Finish it. Then pick the next one.
What to Do With the Hours You Get Back
This is the part people don't plan for, and then the time just dissolves into busywork.
Before you automate, decide: what is this time for? More client work? Content creation? Strategic thinking? Rest?
Write it down. The hours you reclaim from automation are only valuable if you direct them somewhere. Otherwise your brain (especially an ADHD brain) will fill them with the next shiny distraction.
The AI audit tells you where the time is going. You decide where it goes next.