Automation vs. assistance
So far you have probably used AI as an assistant: you open a chat, ask, review and copy the result somewhere. Automation is different. A workflow runs by itself when something happens, such as a new email, a form submission or a time of day, and carries out a series of steps without you starting it.
The best automations combine both worlds: ordinary app-to-app automation (move data, send notifications) plus AI steps that read, classify, summarize or draft text.
What makes a task a good candidate?
Look for tasks that are:
- Repetitive: they happen the same way, many times a week
- Rule-based or pattern-based: you could explain the steps to a new assistant
- Digital: the information already lives in apps such as email, forms, spreadsheets or CRMs
- Tolerant of occasional errors, or easy to check before anything important happens
- Annoying: the tasks you procrastinate on are often the ones worth automating first
Tasks to avoid automating, at least at first
- High-stakes, irreversible actions: payments, deleting data, sending legal notices
- Tasks needing judgment and relationships: negotiations, sensitive customer issues, hiring decisions
- Rare tasks: if it happens twice a year, automation probably is not worth the setup and maintenance
- Broken processes: automating a messy process just makes the mess faster. Fix the process first.
The automation scorecard
Score candidate tasks from 1 to 5:
| Task | Frequency | Time each | Rules clear? | Risk if wrong (low = good) | Total |
|---|---|---|---|---|---|
| Log new web form leads into a spreadsheet and notify sales | 5 | 2 | 5 | 5 | 17 |
| Summarize and file meeting notes | 4 | 3 | 4 | 4 | 15 |
| Draft replies to common support questions (for review) | 5 | 3 | 3 | 3 | 14 |
| Approve supplier invoices | 3 | 3 | 2 | 1 | 9 |
Start with the highest scores. Notice the invoice approval scores low because of risk, even though it is frequent.
Use AI to find candidates
Here's a description of my typical work week, including the apps
I use: [DESCRIBE TASKS AND APPS].
Identify the 5 tasks most suitable for automation. For each,
describe: the trigger (what starts it), the steps, where AI could
help (reading, sorting, summarizing or drafting), the main risk, and
whether a human should approve before the final step.
A possible suggestion: “Trigger: a new message arrives in your ‘quotes’ inbox. Steps: AI extracts the customer name, product and quantity; the details are added as a row in your quotes spreadsheet; a draft reply is created for you to review. Risk: misread quantities, so keep the draft-for-review step.”
Estimate the payoff
A simple formula: time saved per week × weeks per year − setup and maintenance time. A workflow that saves 20 minutes a day saves well over 70 hours a year. That justifies an afternoon of setup and an occasional check-up.
Your first automation should be boring
Pick something simple, low-risk and frequent, like logging form submissions or sending yourself a daily summary. Early success builds confidence and teaches you the tools before you tackle anything complex.
Practice task
List at least eight recurring tasks from your work week and the apps involved. Score each using the automation scorecard and run the AI prompt from this lesson for a second opinion. Pick your top two candidates and, for each, write down the trigger, the steps, where AI helps, and where a human should approve.