Quick Overview
This guide will help you identify which tasks in your business are worth automating with AI.
In 30 minutes, you will review the work you repeat each week, such as replying to leads, preparing client calls, creating content, updating your CRM, summarizing customer feedback, and producing reports. You will then break those workflows into smaller tasks and score them using six practical criteria.
By the end, you will have a shortlist of suitable automation opportunities, a clear recommendation for how much control AI should have, and one workflow to test first.
The aim is to identify a small, useful workflow that saves time, is easy to review, and does not create more work than it removes.
Who This Is Useful For
This guide is for solopreneurs, consultants, founders, and small business owners who know AI could save them time but have no interest in spending a week mapping processes or learning a complicated automation platform. You need a quick way to look at the work already on your plate, find the strongest automation candidate, and leave with one workflow you can begin testing today.
What You’ll Learn
How to find automation opportunities inside the work you already do each week
How to separate tasks that need AI from tasks that only need a simple rule or integration
How to score a task based on frequency, time cost, consistency, risk, and review effort
How to decide whether AI should assist, draft, recommend, or act automatically
How to choose one small workflow you can test without rebuilding your business
The Setup
You do not need an automation platform for the audit itself. Open a blank spreadsheet, Notion page, or document and create these columns:
| Task | Frequency | Time per run | F | A | S | T | E | R | Total | Recommended lane |You can also complete the audit with ChatGPT, Claude, or Gemini by using the audit prompt shared later in this guide. The AI can help organize your answers, but the quality of the audit depends on how honestly you describe the work.
Set a 30-minute timer and divide it like this:
5 minutes to list recurring work
5 minutes to break workflows into tasks
10 minutes to score the tasks
5 minutes to choose the right automation lane
5 minutes to define your first test
For your first audit, stay inside one area of the business. Pick marketing, sales, client delivery, operations, finance administration, or customer support. Auditing the entire company at once makes every task look equally urgent.
The 30-Minute AI Automation Audit
Step 1: List the work that keeps repeating
The first question is simple: what did you do more than once during the past two weeks?
Look through your calendar, sent emails, task manager, browser history, and recent documents. Memory tends to surface the most annoying task, while your actual work history reveals the tasks that quietly consume the most time.
Write down workflows, not ambitions. “Grow my audience” is an objective. “Turn each podcast episode into a newsletter and three social posts” is a workflow.
Useful places to look include:
Leads arriving through a form, email, or direct message
Meetings that require the same preparation or follow-up
Client onboarding and recurring client updates
Research, drafting, editing, and content distribution
Reports assembled from the same sources each week
Information copied between apps
Support questions that repeat
Documents created from a standard format
Aim for 8 to 12 workflows. Do not evaluate them yet.
Best practices
Use the last two weeks as evidence. A theoretical task can wait for a later audit.
Include small tasks. Ten minutes repeated every day can matter more than a two-hour task performed once a quarter.
Write the trigger and output where possible: “When X happens, I produce Y.”
Tools or resources
A spreadsheet is enough. RescueTime, browser history, calendar records, and task-manager activity can help when your week feels too fragmented to recall accurately.
Step 2: Break each workflow into smaller tasks
AI usually automates parts of a workflow better than it automates the whole workflow.
Take “create the weekly newsletter.” That may include choosing a topic, collecting sources, reading the sources, identifying the central argument, outlining, drafting, fact-checking, editing, formatting, adding links, scheduling, and reviewing performance.
Those tasks have very different requirements. Drafting from approved notes is a strong AI-assistance candidate. Choosing the editorial position may require your judgment. Scheduling the finished newsletter is ordinary automation. Fact-checking still needs careful verification.
For each workflow, write the sequence as individual actions. Start each action with a verb.
Tools or resources
Whimsical, Miro, and FigJam can map workflows visually. A numbered list is faster for this 30-minute exercise.
Step 3: Score each task using the FASTER test
Now score each task from 1 to 5 across six criteria. The acronym is FASTER:
Use a spreadsheet formula to add the six scores. The audit prompt later in the guide can also calculate and rank them.
Step 4: Put each task in the right automation lane
A task can be a strong candidate and still require the wrong type of solution. Put every shortlisted task into one of four lanes.
Tools or resources
ChatGPT, Claude, and Gemini work well for Lane 2. Zapier, Make, and n8n can support Lanes 3 and 4 by connecting apps, triggers, AI steps, and approval points.
Step 5: Choose one workflow and define the first version
Choose one task with a high score and a clear review process. Then make the first version smaller.
A useful first automation has:
One trigger
One defined source of input
One main transformation
One destination
One human checkpoint
For example:
Too broad: Automate my content marketing.
Testable: When I add an approved webinar transcript to a folder, create a newsletter outline using my template and save it as a draft for review.
Write your first version in this format:
When [trigger] happens, collect [input], use AI to [specific transformation], save or send the result to [destination], and require me to review [decision or output] before [final action].
Step 6: Run a manual test before connecting tools
Before you build anything in Zapier, Make, or n8n, simulate the workflow manually inside an AI chat.
Take three to five real inputs and run the exact transformation you want. Keep the prompt, reference files, output format, and review checklist consistent.
You are testing four things:
Does the input contain enough information?
Can AI produce the required output consistently?
Can you review the result quickly?
What exceptions appear repeatedly?
Suppose you want to automate lead qualification. Paste five enquiry forms into an AI chat one at a time and ask for the same structured brief. You may discover that prospects rarely include company size, budget, or urgency. The automation is not ready because the input is incomplete. Fixing the form may save more time than improving the prompt.
This is why the audit comes before tool setup. Workflow quality depends on the trigger, inputs, rules, review process, and exception handling. The connector is only the plumbing.
The Audit Prompt
Use this prompt in ChatGPT, Claude, or Gemini. It will guide the audit, calculate the scores, and recommend the right automation lane.
You are helping me audit my recurring work and choose one task to automate with AI.
My business or role:
[Describe what you do in 2 to 4 sentences]
The area I am auditing:
[Marketing / sales / client delivery / operations / finance admin / customer support]
My recurring workflows from the past two weeks:
[Paste 8 to 12 workflows. Include frequency and rough time per occurrence where possible.]
Work in five phases.
PHASE 1: DECOMPOSE
Break each workflow into individual tasks. Start each task with a verb.
Separate collecting information, interpreting it, deciding, drafting, reviewing, sending, publishing, and updating records.
PHASE 2: SCORE
Score each task from 1 to 5 using the FASTER test:
F = Frequency
1 yearly or rare, 2 monthly, 3 weekly, 4 several times a week, 5 daily or more
A = Aggravation
1 easy/enjoyable, 2 mildly inconvenient, 3 often postponed, 4 draining/disruptive, 5 causes delays or dropped work
S = Standardization
1 different every time, 2 loose similarities, 3 recognizable pattern, 4 mostly consistent, 5 clear template or rule
T = Monthly time cost
1 under 30 minutes, 2 30 to 90 minutes, 3 1.5 to 3 hours, 4 3 to 6 hours, 5 more than 6 hours
E = Error tolerance
1 serious harm possible, 2 hard to detect or reverse, 3 visible during careful review, 4 easy to catch and fix, 5 internal/reversible/low risk
R = Review effort
1 review takes as long as manual work, 2 specialist or multi-system review, 3 meaningful review but still saves time, 4 quick checklist review, 5 obvious within one or two minutes
Show the reason for every score. Do not invent missing details. Mark uncertain scores as assumptions.
PHASE 3: CLASSIFY
Place each task in one lane:
Keep manual
AI assistance
AI workflow with human review
Automatic execution
Flag tasks involving sensitive data, money, contracts, legal issues, publishing, customer communication, or important business records.
PHASE 4: PRIORITIZE
Rank the top five candidates. Consider the total FASTER score, risk flags, ease of testing, and whether the input and output are clearly defined.
Do not recommend a high-risk task merely because it has a high score.
PHASE 5: DESIGN THE FIRST TEST
Choose the strongest first candidate and define:
Trigger
Input
AI transformation
Output destination
Human review point
Failure or exception path
Current estimated time per run
Target review time per run
Five-example manual test plan
Acceptance checklist
End with a one-sentence workflow in this format:
When [trigger] happens, collect [input], use AI to [transformation], save the result to [destination], and require human review before [final action].
Review any assumptions, lower scores that feel generous, and keep final authority over tasks that affect customers, money, sensitive information, or public claims.
How useful was this guide for you?
💌 We’d Love Your Feedback
If you need any guidance while implementing this, or if something isn't quite clear, feel free to ping the team. We're here to support you and clear things up.
Until next time,
Team PracticalyAI
