Automate first the task that is repetitive, rule-based, high volume and lives in text or documents, and that scores high on time saved and low on risk. Run a one-week time audit, score each task from 1 to 5 on volume, time and risk, and start with the top-left of the grid: big savings, low risk.
- Start with boring work, not the exciting idea.
- Four signs: repetitive, rule-based, high volume, text-based.
- Score volume, time and risk from 1 to 5.
- Your first project lives in the high-savings, low-risk corner.
Why most AI projects start in the wrong place
The most common mistake is starting with the most impressive idea, like a chatbot on the website, instead of the work that quietly eats hours every week. Impressive projects tend to be complex, customer-facing and risky. Boring projects tend to be simple, internal and quick to prove. Prove one boring win first, and the exciting ideas get much easier to justify.
The four signs a task is ready for AI
Step 1: Run a one-week time audit
For one normal week, write down every task that takes more than fifteen minutes. Note who does it, roughly how many hours it takes across the week, and what tends to go wrong. You can’t fix what you haven’t measured, and the list almost always surprises people.
Step 2: Score each task on volume, time and risk
Score every shortlisted task from 1 to 5 on three things. Volume: how often it happens. Time: how long it takes each time. Risk: how bad it is if a mistake slips through. Volume and time together tell you the savings. Risk tells you how much checking it needs.
Big savings, low risk. Your first project.
e.g. order-status emails
Big savings, higher risk. Second wave.
e.g. invoice entry
Small savings, low risk. Pick up any time.
e.g. file naming
Small savings, high risk. Not worth it yet.
e.g. contract wording
Step 3: Write one goal sentence
Before touching any tool, write one sentence: In thirty days, AI will do [task], for [people], so that [result], with [person] checking [this part]. For example: “In thirty days, AI will draft replies to order-status emails for our front desk, so customers hear back within an hour, with Maria checking every reply before it’s sent.” One task, one team, one result, one named person.
Most owners want to add a second workflow at this point. Don’t. One workflow done well teaches you more than five half-finished experiments, and the second one is much easier once the first is working.
Step 4: Launch in review mode
Start with AI preparing everything and a person approving everything. Only loosen the checks when most recent results needed no fixes. Read how human-in-the-loop review works, then measure the result against your audit.
A worked example: scoring five tasks
Here’s how the example business above might score its list. Volume and time are multiplied to show the savings; risk is noted separately.
Order-status emails and appointment booking tie on savings, but order-status replies carry less risk, so they go first. Invoice entry saves a lot too, but because money is involved it needs a person checking every invoice, so it comes second, once the team is comfortable working with AI.
Common mistakes to avoid
Gut feel overestimates rare, annoying tasks and underestimates daily ones. Use a real week of notes.
A task that eats the owner’s evenings may be worth more than the hours suggest.
High-risk tasks can still use AI. They just need a person approving every result.
Keep the one-sentence goal fixed for thirty days so you can judge the result fairly.
Our first step with every client is exactly this audit, because the right first task matters more than the right tool. A boring task that saves six hours a week with low risk beats an exciting chatbot every time. When we score tasks with owners, the winner is usually something they never thought of as “AI work”.
Where to go from here
The fastest way to learn this is to try it on one real task this week. If you want a guided start, take the free course. If you’d rather talk it through, book a call.
Questions, answered
How do I know which tasks to automate with AI?
Look for tasks that are repetitive, rule-based, high volume and text-based. Score them on volume, time and risk, and start with the one that saves the most time with the least risk.
What are the worst tasks to automate first?
Low-volume, high-risk work like contract wording, sensitive HR decisions or anything that moves money without checks. Save those for later, with strong human review.
How long should a time audit take?
One normal week. Note every task over fifteen minutes, who does it, the hours and what goes wrong. It takes a few minutes a day.
Should I automate customer-facing tasks first?
Usually not. Internal tasks are safer to learn on. If you do start customer-facing, keep a person approving every message at first.