Most founders already suspect they should automate something. The harder call is which processes to automate first. Pick well and the first build pays for itself within weeks and gives the team confidence. Pick badly and you spend money speeding up a task that happens twice a month, or you lock a broken process into software so it fails faster. This guide gives you a scoring method you can run in an afternoon, a short list of good first candidates, and the work we would keep with people.
Start with the work, not the tool
The usual mistake is to begin with a product. Someone sees a demo of an automation platform or an AI assistant and starts looking for things to plug into it. You end up with clever automations nobody needed.
Start instead by listing the repetitive work your team does in a normal week. Ask each person to keep a rough note for five working days: what they did more than once, roughly how long it took, and which tools they had to copy information between. You will get an untidy list. That is fine. Untidy and honest beats tidy and imagined.
Group the notes into processes. “Copying form enquiries into the CRM”, “sending appointment reminders” and “making the weekly sales sheet” are processes. “Admin” is not.
Score each candidate
For every process on the list, answer five questions with a score from 1 to 3.
- How often does it happen? Daily scores 3, weekly scores 2, monthly or less scores 1.
- How long does each run take? Over 15 minutes scores 3, five to 15 minutes scores 2, under five scores 1.
- Are the rules clear? If you could write the steps on one page and a new hire would follow them without asking, score 3. If it depends on “it depends”, score 1.
- What does a mistake cost? Here, a lower risk earns a higher score for a first project. A missed internal report scores 3. A wrong invoice to a customer scores 2. A wrong salary payment scores 1.
- Is the process stable? If it has not changed in six months, score 3. If you are still arguing about how it should work, score 1.
Add the scores. Anything at 12 or above is a strong first candidate. Between 9 and 11, it may be worth doing later. Below 9, leave it for now.
Say a Kochi interiors studio lists “send site visit reminders to clients”. It happens daily (3), takes about ten minutes each time (2), follows a fixed rule of one reminder the day before and one on the morning (3), costs little if a reminder is late (3) and has not changed in a year (3). That scores 14. Their “quote approval” process, which depends on the owner’s view of each client, scores 8 and stays manual.
Which processes to automate first: common good candidates
Every business is different, but the same kinds of work tend to score well for small and growing teams.
- Lead capture: moving website form, WhatsApp or ad leads into one place, tagged with the source, and alerting the right person.
- Reminders and confirmations for appointments, deliveries or payments that follow a fixed schedule.
- Handoffs between tools, such as creating a project in your task tool when a deal is marked won in the CRM.
- Recurring reports that someone builds by hand from the same sources every week.
- Document creation from data you already hold, such as generating an invoice or a contract draft from a signed quote.
These share a pattern: high volume, clear rules, data already in a system, and a mistake that is easy to spot and fix. They are also the processes where time saved shows up quickly, which helps when you need the team to trust the next project.
Processes to leave alone for now
Some work should stay with people, at least until it settles down.
- Anything still changing. If the team is redesigning the onboarding flow, automating the current version means paying twice.
- Decisions that depend on judgement, such as discount approvals, hiring choices or how to handle an angry customer. You can automate the paperwork around them, but keep the decision itself human.
- Rare tasks. A process that runs four times a year will rarely repay the build and the upkeep.
- Moments that build a relationship. A founder’s personal message to a new client after the first project is worth the five minutes.
- Broken processes. If the manual version produces errors, an automated version will produce them faster. Fix the steps first.
A simple decision rule
If you only remember one thing from this article, make it this: automate a process when it is frequent, rule-based and stable, and when the person doing it today agrees it is tedious. That last condition matters more than it sounds. The person who does the work knows the exceptions, the workarounds and the customers who always need special handling. If they are not part of the decision, the automation will miss what they know.
Before you build anything, map the process step by step so you can see where it starts, where it ends and who touches it. Our separate guide to process mapping for automation walks through how.
What to do in the first month
- Week one: collect the five-day work notes from the team and group them into processes.
- Week two: score each process and pick one or two candidates above 12.
- Week three: map the chosen process, agree the rules in writing and name an owner.
- Week four: build a small first version, run it alongside the manual process for a few days and compare the results.
Keep the first build small. One process, end to end, done properly, teaches you more than five half-finished automations. Measure the time spent before and after so you have a number to point to when you decide on the next one.
If you want an outside view, an automation audit usually takes about two weeks and ends with a ranked list of what to build, what to fix first and what to leave alone. For broader questions about which systems your business should run on, our technology advice work covers tool choices and part-time CTO support.
Not sure where your list points? Book a free 30-minute call and we can look at it together.
Frequently asked questions
How much time should an automation save to be worth building?
There is no fixed number, but compare the hours saved over a year with the cost of building it plus a few hours a month of upkeep. If a process takes ten minutes a day, that is roughly 40 hours a year for one person. If the build and upkeep cost less than those hours are worth to you, and the rules are stable, it is a reasonable candidate.
Should I automate customer communication?
Automate the predictable parts, such as confirmations, reminders and status updates, and keep a person on anything that needs judgement, like complaints or negotiations. Make sure customers can reach a human easily, and check that your messaging follows the channel’s rules and your local data protection law before sending anything at scale.
Can I automate a process that only one person understands?
You can, but write it down first. If the steps live only in one person’s head, an automation built from a short conversation will miss the exceptions they handle without thinking. Sit with them, map each step and each exception, and have them check the written version before anyone builds anything.