Process automation: where to start without making a mess

Almost every company we work with has one task somebody does the same way every week: moving orders from one place to another, building the same report, sending the same email with two figures changed. Nobody chose it, it just stayed. This article is about starting process automation with that one task instead of a grand plan that never gets going.

What automating a process means, and what it does not

Automating means a sequence of steps a person does today gets done by the system, with the same rules and without anybody having to remember. It can be as small as an online order landing straight in the invoicing tool, or the sales report arriving every Monday in the inbox of whoever needs it.

What it is not: replacing every tool in the company, laying anyone off, or a project that takes half a year to show signs of life. In practice most of what we build goes live within weeks and pays for itself with the hours it frees up.

The three tasks that almost always go first

• The bridge between two tools. The shop does not talk to the management software, or the CRM does not talk to email. Somebody exports, pastes and checks. It is the most common case and the most profitable.
• The weekly report. Same data, same format, same day. A solid half hour that turns into an automatic email with the file already attached.
• The alerts that depend on somebody's memory. An order stuck for three days, a payment that never arrived, a client who needed a call. If nobody looks, nothing fires.

How to tell whether a task can be automated

Two quick questions. First: is it done the same way every time, following the same steps? Second: if that person is off sick, does it simply not get done? Two yeses and you have a candidate. If the task changes case by case and needs judgement, automate the mechanical part and leave the decision to the person — usually the best combination.

Where to start without making a mess

• Watch how it is really done today. Sit next to whoever does it rather than asking by email. There are almost always steps that appear in no manual.
• Write down the rules and the exceptions. What happens if the order arrives with no address, if the client already exists, if the amount is zero. Exceptions are what break badly built automations.
• Build the smallest piece that already helps and run it on real data, in parallel with the manual method for a few days.
• Leave a log and an alert. Anything automatic has to record what it did and shout when something fails. Without that, nobody notices the day it breaks.

If your company meets the requirements, part of this work can be funded through the Spanish Kit Digital scheme, for which we are an accredited digitalising agent.

Three everyday examples

So this does not stay theoretical, three cases that keep coming up in small companies and are solved with short pieces of work:

• From the shop to the invoice. Every order on the website creates the customer if they are new, issues the invoice with the right series and emails it out. No more morning ritual, and no more invoices carrying last year's price.
• From the quote to the follow-up. Sending a quote schedules a reminder five days later on its own. If the client replies, the reminder cancels itself; if not, somebody gets the nudge with the phone number attached. This one alone recovers work that gets lost to forgetfulness.
• From chaos to a report. Every Monday at seven an email goes out with the week's sales, the open orders and anything stuck for more than a week. Five minutes of reading instead of half an hour of assembling.

None of the three is a big project. All three are noticeable in the first week, and that is exactly the criterion for choosing where to start. Spain's data protection agency publishes clear material for small companies in case the process touches customer data, which it usually does.

What it costs and what it saves

Automations are quoted by what has to be built, just like custom software: there is no per-task rate because connecting two modern tools with an open API is nothing like extracting data from an old program with no way out. What we always do is agree the scope and the price in writing first.

The quick sum

Take the task, measure the minutes it eats each day and multiply by the working days in a year. Forty minutes a day is over a hundred and fifty hours. Multiply those hours by what an hour costs in your company and you have the real budget available to get rid of it. Add the errors that slip through today — the outdated price, the row that never got copied — and the maths usually makes itself.

What can go wrong

• Automating a broken process. If the current flow does odd things, automating it only makes it do them faster. Tidy first, automate second.
• Nobody watching. Every automatic process needs an alert when something breaks and somebody who reads it. A silent flow that has been down for two months does more damage than doing it by hand.
• Personal data. If the process moves customer data, you need to know where it ends up and who can see it. Spain's INCIBE publishes free guides for small companies on exactly this, and they are worth an hour before connecting anything.

Off-the-shelf tools or custom development

Plenty of cases are solved with the tools you already pay for: management, invoicing and email platforms usually ship with automations and connectors nobody ever configured. Starting there is the cheapest route.

Custom development comes in when the process is genuinely yours, when the old program has no way to communicate, or when three systems that know nothing about each other have to be joined. Volume matters too: what costs an hour by hand costs a full-time role at a hundred orders a day. And if you want to lean on AI models to classify or summarise, that falls under AI integration, with its own rules about what information leaves the building.

How we work

We are three partners based in Palamós, working with companies across Girona and Barcelona. Ivan handles applications, integrations and process automation; Jordi runs marketing; Izan looks after cybersecurity and data. The pricing page publishes the recurring fees; automations are quoted per project.

If a task repeats and is eating your month, describe it in detail: how many times a day, who does it and with which tools. That is usually enough to see whether it is worth automating, whether configuring what you already own solves it, or whether it is better left alone — which happens too.