Insights · cargomastersgroup.ro
Which tasks to automate first: a scoring method for small teams
In short: Automate first the tasks that happen often, follow clear rules, take real time on each run and cause problems when done late or wrong. Score each candidate from 1 to 5 on frequency, time per run, rule clarity, error cost and data readiness, subtract a risk score, and start with the highest total that one person can own.
Workflow automation means having software carry out a repeatable sequence of steps, such as reading an incoming request, filing a document, sending a reminder or compiling a report, with people stepping in only where judgment is needed. For a small team the hard part is rarely the technology. It is choosing where to start.
Pick the wrong task and you spend weeks automating something that happens twice a month. Pick the right one and the team notices the difference in the first week. The method below takes about an hour once you have a list.
Step 1: list the busy work
For one week, have everyone note each repetitive task, who does it and roughly how long it takes. Do not filter yet. In a growing business the list usually includes items like these:
- Answering the same pre-sales questions by email or website chat.
- Asking clients for missing documents, then chasing them.
- Copying details from forms or PDFs into a spreadsheet or CRM.
- Sending appointment, renewal or payment reminders.
- Creating proposals, engagement letters or certificates from a template.
- Compiling a weekly status or sales report from several sources.
- Setting up a new client or employee: accounts, forms, welcome emails.
Step 2: score each task
Rate every task from 1 (low) to 5 (high) on five value criteria, then rate its risk on the same scale.
| Criterion | Question to ask | 1 means | 5 means |
|---|---|---|---|
| Frequency | How often does it happen? | A few times a month | Many times a day |
| Time per run | How long does it take each time? | Under two minutes | Over 20 minutes |
| Rule clarity | Could you write the steps on one page? | Depends on judgment | Same steps every time |
| Error cost | What happens when it is late or wrong? | Minor inconvenience | Lost client, missed payment or compliance problem |
| Data readiness | Is the input digital and consistent? | Handwritten or scattered | Structured form or standard document |
| Risk (subtract) | What could an automated mistake cause? | Easily reversed | Affects money, contracts or sensitive data |
Priority score = Frequency + Time per run + Rule clarity + Error cost + Data readiness − Risk. The maximum is 24. Above 16 is a strong first candidate; below 10 can wait.
Step 3: a worked example
For example, a twelve-person commercial cleaning company might score three tasks like this:
| Task | Freq. | Time | Rules | Data | Risk | Score |
|---|---|---|---|---|---|---|
| Answering quote requests from the website | 5 | 3 | 4 | 4 | 2 | 18 |
| Chasing signed service agreements | 3 | 2 | 5 | 4 | 2 | 16 |
| Weekly hours report for payroll | 2 | 4 | 4 | 3 | 4 | 14 |
Quote requests come out on top. They arrive daily, follow a pattern (location, frequency, floor area, start date), and a slow reply loses the job to a faster competitor. An AI assistant on the website can collect those details, answer common questions and route a complete request to the right person, who still sets the price. The payroll report scores lower despite its high error cost, because a mistake affects people's pay; it is a good second-phase project with a human check built in.
Step 4: check the practical constraints
A high score is necessary but not enough. Before committing, confirm each of these:
- One person owns the process and can say what a correct result looks like.
- The tools involved, such as email, forms, spreadsheets, accounting or CRM software, can be connected directly or through exports.
- There is a clear point where a person reviews or approves the output.
- You know what data the automation will touch and who is allowed to see it.
- You can measure before and after, even roughly: hours spent, reply time, number of chasers sent.
Step 5: start small and measure
- Map the current steps on one page, including the common exceptions.
- Automate the most repetitive part first rather than the whole process.
- Run it alongside the manual method for a week or two and compare results.
- Fix the edge cases, then switch over while keeping a manual fallback.
- Review after a month: time saved, errors, client reactions, and what the team still does by hand.
Honest trade-offs
Automation has running costs of its own. Someone has to update the rules when your process changes, and an AI assistant needs boundaries: what it may answer, when it must hand over to a person, and what it must never promise. Document tools such as OCR read clean, typed documents well but struggle with poor scans and handwriting, so plan a review step where errors would matter. Finally, some tasks are better simplified or dropped than automated. If a report nobody reads takes three hours a week, the cheapest fix is to stop producing it.
If your top-scoring task involves enquiries, documents, follow-ups or reporting, see how we scope AI automations for small teams. If the work starts on your website, the website cost guide helps frame the budget, and you can describe your workflow in a request.
Questions and answers
How many tasks should we automate at once?
Usually one. Starting with a single high-scoring task lets the team learn how automation behaves, builds trust in the results and gives you a clean before-and-after comparison before you commit to the next project.
Does every automation need AI?
No. Many workflows, such as reminders, filing and scheduled reports, run on fixed rules. AI is useful where input varies, for example reading free-text enquiries or extracting data from documents.