← Blog

// AI automation for business

AI Automation for Your Business — Where to Actually Start

August 11, 2026

When someone asks "should my business be using AI?", I ask back: "What's the recurring task you dread most every week?" That's the starting point.

This is the hub post of the business automation series. Each topic gets its own deep-dive; this post sets the overall order.

How to pick what to automate

Not every repetitive task is a good first target. My three criteria:

  1. Does it recur on a schedule — daily or weekly work
  2. Can the judgment be written down — "when X, do Y" expressible in plain language
  3. Is a mistake recoverable — a human can review before it goes out

Hit all three and it's a candidate. The usual suspects:

  • Review replies — answering Google reviews one by one. (Automating Google review replies covers the how.)
  • Content drafts — first drafts of social posts and blog articles
  • First-touch inquiry responses — drafted answers to the questions you get every week
  • Report assembly — collecting the same numbers into the same summary every Monday

The rollout order — small and reviewable

Failed automations share one trait: "everything at once." The order I recommend:

  1. Pick one task. The most frequent, most dreaded one.
  2. Start with draft generation. AI writes, a human approves and sends. Fully automatic publishing comes after trust is earned.
  3. Run it for a week, then adjust. Collect the awkward outputs and fold them into the instructions.
  4. Then expand to the next task.

What it costs

Agencies quote thousands for "AI adoption consulting." The actual structure is usually:

  • AI subscription: $20–200/mo
  • The automation script: $0 if you build it yourself (and you can have AI build it)
  • Optional glue tools: tens of dollars a month

The core expense is one AI subscription. If a quote says otherwise, ask what the difference pays for. Sometimes it's legitimate (complex system integration) — often it isn't.

Managing expectations

Being honest:

  • What works: big cuts in repetitive-task time, faster response times, higher content output
  • What doesn't: judgment calls on exceptions, fully unattended "it just handles everything" operation
  • The surprise benefit: building the automation forces you to write down how the work is done — long-term, that documentation is the bigger asset

New experiments and their results land in the AI This Week series and the newsletter.

Posts in this series