Where to start with AI? Have AI score your processes on 4 questions

Yes. List the work your team does every day or every week (paste it from a spreadsheet or upload the Excel file) and ask an AI assistant such as ChatGPT, Claude or Gemini to score each item on four questions: how often it happens, how long it takes, whether the rules can be spelled out, and how bad a mistake would be. In a few minutes you'll have a ranked table, and the top scorer is the first process worth working on.

"Shouldn't we be using AI by now?" Over the past two years, it's the question we've heard from business owners more than any other. Many companies already use AI to write emails and tidy up data, and that's a good start. This article gives you prompts to copy, then who should check the scores, how your spreadsheet helps, and how to phase in the process you pick.

Ask AI directly: 3 prompts to copy

Copy these and swap in your own details (the ones below are examples).

Build the candidate list

Below is the work each department wrote down that they do every day or every week. Turn it into a table: process name, department, how often it happens, roughly how long it takes each time, and how many people handle it. Merge items that mean the same thing. Where frequency or time isn't clear, mark it "to confirm" instead of guessing.

Score it on four questions

Score each process from 1 to 5 on four questions: frequency (several times a day is 5, once a month is 1); time (half an hour or more, or a relay between several people, is 5); clear rules (a senior colleague can explain them clearly is 5, "it depends, it comes down to experience" is 1); risk, scored in reverse (mistakes are easy to spot and fix is 5, they could cause major losses or legal trouble is 1). Give a one-line reason for each score, add them up and sort from highest to lowest. Mark 16 or more "do first," 10–15 "tidy up first" and under 10 "keep manual." Give me a table I can paste back into Excel.

Pick the first project

For the top three processes, write two measurable success targets each (such as hours saved per day or fewer errors), and say whether each is better solved by connecting two systems so the data syncs, or needs AI to understand content (such as customer documents).

After AI scores your list, check these 3 things

1. Have the people who do the work look it over. The AI inferred frequency and time from your list, so ask each person to put in the real numbers.
2. Don't let the rules and risk scores run high. AI may treat "it depends" work as having clear rules. Where a mistake would affect payments, customers or compliance, keep the risk score conservative.
3. It only scores the list you give it. Ask front-line staff to add the time-consuming work nobody wrote down, then have it score again.

If your list mentions customer names, amounts or employee names, replace them with codes before pasting, or use your company's enterprise plan.

Make it easier: your spreadsheet and AI working together

1. Spend an afternoon listing candidate processes together

With department heads and front-line staff, list in a shared Google Sheets file or Excel workbook the work that happens every day or every week and takes a lot of time, without worrying yet about whether it can be automated: keying orders into the ERP, answering the same questions on LINE (the messaging app most Taiwanese customers use), month-end reconciliation, weekly reports. You'll usually have 15–30 candidates in no time, and the whole sheet can go straight to AI for scoring.

2. Paste the scores back and adjust them together

Use Data Validation to make the score columns a 1–5 drop-down list so colleagues can change them directly. Add up the scores with SUM, sort, and use conditional formatting to color the rows scoring 16 or more. High scorers happen often, have clear rules and manageable risk, so do them first. Keep anything under 10, such as occasional tasks or work that depends on experience, manual for now.

Tidy up the middle (10–15) first: write unclear rules on a rules sheet, list the common causes of discrepancies where there are many exceptions (reconciliation, for example), and agree on one format for customer codes and part numbers. This costs nothing, and the rules sheet can go along with every later request to AI.

Here's an illustrative set of scores:

An illustrative case: a 20-person trading company

The following illustrative scenario draws on several cases. A 20-person trading company wanted to adopt AI.

Asking AI first: sales, the sales assistants, accounting and the warehouse each wrote down their daily work, which gave 22 candidate processes. AI did the first round of scoring, then each person checked their own items:

• Top: entering customer orders into the ERP. Two sales assistants each spent about two hours a day keying in PDF orders from email. The rules were clear (a part-number cross-reference table already existed), and mistakes would be caught before shipping.
• Second: the weekly sales and inventory report. The sales manager spent half of every Monday exporting ERP data into a presentation.
• Middle: month-end reconciliation. More exceptions, so it went into the second wave, and accounting first listed the common causes of discrepancies.
• Lowest: quoting and negotiation. They rely on the sales team's experience and relationships, so they stayed manual.

Then connecting the systems: the first project was order entry. AI read the PDF orders straight from email, matched part numbers and drafted each order, and an assistant confirmed it before it was created (see reading orders and quotes with AI). The goal was to cut the assistants' data-entry time by at least half. After launch, they put the time saved into tracking delivery dates and reaching out to customers, and the owner started a second project.

When to consider the next step

• For your top-scoring process, you export, upload and paste the same files every day.
• You'd like AI to see the data in your ERP or on LINE without attaching it each time.
• As orders or messages grow, the same task needs several people in a relay.
• The rules are written down, and you'd like a record of every item handled.

The next step: phase in the top-scoring process

Once you've picked your first process, the next step is deciding how to handle it:

• It doesn't always need AI: some processes are solved by connecting two systems so the data syncs automatically, such as pushing e-commerce orders straight into the ERP. This is often cheaper and more reliable than AI (see connecting older systems without replacing them).
• Where the work requires understanding content, connect AI to your existing systems: customer documents and LINE questions come in on their own, AI checks them against your existing data and drafts the next step, and anything that doesn't match or that it isn't sure of goes to a person. Nothing is written to your systems until a person confirms, and every step is logged.
• In phases, side by side: the first phase usually goes live in about 4–6 weeks while your current way of working carries on, and you switch once the two match. Get results, then pick the next process.

Along the way, involve the people who do the work every day, and after launch keep watching the numbers, gathering feedback and adjusting the rules.

If you'd like help mapping your processes, start with a free AI Health Check, which gives you an automation blueprint. You can also look at our AI adoption and integration services and custom business systems, or book a process consultation.