What does your team do by hand every week? A plain guide to automation readiness

What does your team do by hand every week? Pick one task and hold it in your head before you read on. Anything will do: the invoice run, the Monday report, the copying between two systems.

That question is where automation readiness starts. Before any tool or build, you need to know which tasks are worth automating, and whether they’re ready. Most teams know roughly where the hours go. The hard part is spotting the real opportunities and being honest about the process behind them.

In this guide I’ll show you how to find opportunities, score them with three tests, check whether they’re ready, tidy up what isn’t, and prepare the right information.

Start with one question: what does your team do by hand every week?

I like weekly tasks because they repeat often enough to matter, and people actually remember them.

Here’s the exercise. List five to ten tasks. For each one, write down:

• who does it
• how long it takes
• how often it happens
• which tools are involved

Ask the people who do the work, not just the owner. Owners often miss the small, repeated jobs.

Count it per person. One hour a week done by four people is four hours a week.

Where automation opportunities usually hide

If you’re wondering how to identify automation opportunities, these six are the usual suspects:

• Copying data between systems: the same details typed in twice.
• Invoicing and chasing payments: remembering who to remind, and when.
• Bookings and reminders: the back-and-forth over times.
• Reports: numbers pasted together every Monday or month-end.
• Onboarding: the same welcome steps and documents for every new client or staff member.
• Repeat emails: the same five replies, and follow-ups that someone has to remember.

Other signals:

• “I forgot to…” moments
• double-checking because mistakes are common
• work stuck waiting for one person
• spreadsheets that exist only to move data from A to B

If several of these sound familiar, automating repeatable work like this is usually where the hours are hiding.

Score each task: time, repeatability and pain

This is the same scoring the Automation Readiness questionnaire uses. Each task gets a score out of 100, which gives you a fair way to decide what to automate first.

Time (40 points): hours × people

Time is the clearest measure of what’s at stake. Multiply the weekly hours by the number of people who do the task. One hour for each of five people matters more than a two-hour job done by one person.

Repeatability (35 points): same steps, data that starts digital, tools that connect

Three questions:

• Are the steps the same every time?
• Does the information start digital? Email, web forms and exports are easy. Paper and phone calls are not.
• Can your tools share data? Or only through a person with a keyboard?

Pain (25 points): mistakes, one-person dependence, urgency

• Mistakes that cost money or trust.
• “Only one person knows how.” That’s a readiness problem and a risk. If that person is off sick, the work stops.
• Urgency: work that can’t wait.

What the score suggests

Treat the bands as next-step guidance, not a verdict:

• 0 to 39, tidy up first. Sort out the process and the data before automating. An Ideas session can turn that into a plan.
• 40 to 69, quick wins. A good, contained first candidate. Usually a single workflow.
• 70 to 100, big win. Worth automating, and worth planning properly first, because it touches more steps, people or systems. What a first workflow looks like is a useful place to start.

Readiness signals: is this task ready to automate?

As an automation readiness checklist, a task is usually ready when:

• Someone owns it, and would notice if it broke.
• There’s a clear trigger. It starts with an event, like a form, an email or a date, not “when someone remembers”.
• The normal path fits on one page, and people describe it the same way.
• The inputs are digital and consistent: the same fields, and few duplicates.
• Exceptions are known, and most cases follow the normal path.
• You can say what “better” looks like: hours, errors or response time.

Here’s a real one. We built a desktop tool that turns one long interview or podcast recording into captioned vertical clips for social media. It worked because the rules were written down first: minimum and maximum clip length, and the most clips it may suggest. It checks at startup that everything it needs is in place, so it fails early and clearly. A person still reviews the clips and picks what gets exported, because “interesting enough to post” is a judgement call. One recording it handled ran about 1 hour 45 minutes and roughly 18,000 words. Automate the repetitive middle, not the judgement.

If two or three of your signals are shaky, that’s not a no. It’s a list of what to tidy up first.

Tidy up first: fix the process and the data

“Tidy up first” is a respectable result. Practical steps:

• Remove steps that no longer add anything, and merge duplicates.
• Agree one way of doing it, and write it down.
• Pick one place where the data lives, one you trust, and stop keeping copies in spreadsheets.
• Move paper or phone inputs to a simple digital form where you can.
• Share what’s in one person’s head. Have a second person do the task once from the written steps.
• Know where personal data goes. If customer details are involved, be clear where they’re stored and who can see them. Rules like POPIA in South Africa and GDPR in the UK and EU apply, so know this before you automate. This isn’t legal advice.

Here’s what that looks like in practice. A growth-analytics platform client’s product owner sent change requests by email, often as Word documents with screenshots. Every request followed the same path: read it, change it, show it, get approval, release it, report back. Before we automated it, I did all of it myself. In four weeks there were 25 client emails, eight on a single day.

It scored well on every test, but it only worked because the boring parts were settled first:

1. The client was told up front that an AI assistant would help, and it’s clearly signed as my assistant.
2. One preview link, where the client checks one piece of work at a time.
3. Clear permission levels. Routine changes go ahead. Anything touching hosting, settings, logins or money comes to me first.
4. Approval means a clear “approved” or “go live”, agreed in advance. A vague “yes, that works” gets a follow-up question.
5. A shared log links every email to its change, status and release date.
6. Bigger product decisions stay human.

One request was approved 49 minutes after it came in and was live 12 minutes later. We also learned something the hard way: two of the three sites had no preview link, so the client couldn’t check those changes. Set up the checking step before you automate.

If a messy process is the sticking point, a one-hour Ideas session is a good way to turn it into a written plan.

What to prepare before anything gets built

A short list you can act on today:

• your weekly task list, with hours × people
• the written steps for your top one or two tasks
• three to five real examples: sample emails, forms, invoices or exports, including an awkward one
• the tools involved, and who has admin access
• the owner, and who checks the result
• a baseline: how long it takes now, and how often it goes wrong
• what “done well” looks like

What happens next

I build most automations in code, as connected services with AI layers where they help. I use the best tool for each job, not one tool for everything. I only use no-code tools like n8n or Make to prototype an idea quickly. Whatever the build, clear steps and clean data make it simpler and more reliable.

Good preparation pays off. The Accordium compliance case study shows where it can lead: real-time sensor data and automatic HACCP logs, replacing manual paper logs and fridge checks.

Fixed-price workflows start from R3,950, and the Automation page has the full detail.

I reply to every message within 24 hours.