Digital Processes Are Not Automated Processes: What Is the Difference?

Almost every SME now runs on digital processes. Very few run on automated processes — and the difference between the two decides whether the software you bought turns into hours saved or just into a change of medium.

A PDF invoice arriving by email is a digital document. The process handling it is still manual.

Between the email and the entry in the ERP sits a person opening the attachment, reading six fields and retyping them. The paper is gone. The work is not.

That confusion is expensive. It leads companies to conclude they have automated when all they did was swap paper for a screen — and leaves them puzzled as to why buying the software freed up nobody's afternoon.

Digital processes vs automated processes: the difference is who takes the next step

A process is digital when its documents and data exist in electronic form. A process is automated when the move from one step to the next happens without anyone deciding to take it.

The practical test: if you remove the person in the middle, does the process stop?

If it stops, you have a digital process. If it carries on and only calls a human when something falls outside the rules, you have an automated process.

Everything else is vocabulary. A spreadsheet shared on SharePoint is digital. A form that generates an email someone has to read is digital. A dashboard nobody opens is extremely digital and automates precisely nothing. Digital processes can be every bit as slow as paper ones; they simply do not leave ink on your fingers.

EU law has drawn this line for a decade

The clearest proof that "digital" and "automatable" are different things does not come from technology. It comes from legislation.

The official EU definition of an electronic invoice, set by Directive 2014/55/EU, describes an invoice "issued, transmitted and received in a structured data format which allows for its automatic and electronic processing".

Note what that excludes. A PDF is electronic, travels by email and is perfectly readable — by a human. It is not a structured format. For the purposes of the directive, it is not an electronic invoice.

The deadlines are already set. Under the ViDA package, adopted by the Council in March 2025, structured invoices following the EN 16931 standard become mandatory for in-scope transactions from July 2030. In Portugal, PDF invoices remain acceptable as electronic invoices until 31 December 2026, with qualified digital signatures required from 1 January 2027.

Regulators understood the difference between digital and automatable well before most companies did.

The numbers show how wide the gap is

In 2025, 46.5% of EU enterprises with 10 or more employees used an ERP system, according to Eurostat data on the integration of internal processes. Among enterprises with 10 to 49 employees, the figure drops to 41.1%.

Now the other side of the ledger. In the same year, only 5.35% of EU enterprises used AI technologies to automate workflows or support decision-making, according to Eurostat data on AI adoption. In Portugal the figure is 3.98%.

Nearly half of all companies have the core system installed. Fewer than one in twenty have anything running by itself between systems. It is the most direct measure available of the distance between digital processes and automated processes.

The problem is not a shortage of software. It is a shortage of connections between the software already bought. This is also why manual processes keep growing even in companies that have gone digital: every new system adds one more place where data has to be re-entered.

Three signs you have digital processes and not automated processes

Someone copies data from one screen to another. The most reliable sign of all. If a colleague keeps two windows open and moves information between them, you have two digital systems and zero automation between them.

The process stops when one person is away. Holiday, sick leave, resignation. If the answer to "who handles this?" is a name rather than a workflow, the process runs on human memory, however neatly the files sit in the cloud.

Email is the integration layer. Requests circulating by email, approvals given as replies, data travelling as attachments. Email is excellent at carrying messages and terrible at carrying state: it does not validate, retry, log, or chase anyone who forgets.

A worked example with numbers

Take a company receiving 300 supplier invoices a month, all by email, all as PDFs.

Digital: someone opens the email, opens the attachment, copies supplier, number, date, net amount, VAT and total into the ERP, checks it against the purchase order and files it. At four minutes per invoice, that is 1,200 minutes a month — 20 hours, roughly 240 hours a year.

Automated: the workflow picks up the email, extracts the fields from the PDF, validates against the purchase order, posts to the ERP and sends the confirmation. Only mismatches reach a person. If 12% of invoices need review at three minutes each, that is 36 invoices and just under two hours a month — around 22 hours a year.

The four minutes, the 12% and the three-minute review are our own estimates, not study data. The structure of the calculation is what matters: the gap between 240 and 22 hours does not come from doing the same work faster, it comes from no longer doing most of it. To turn hours into euros honestly, we set out a step-by-step method for working out what a manual process actually costs.

Measure your own four minutes before trusting ours.

Digital processes are the first step. The mistake is stopping there

The order matters and is not negotiable: you cannot automate a paper process. Digital processes are the foundation, not the goal. First the data has to exist in a structured format, then the rules have to be written down, and only then can systems talk to each other.

The third point is where projects fail, and they fail over one technical detail that gets expensive when ignored: "the ERP has an API" is not the same claim as "the ERP has an endpoint that creates a purchase document, accepts batches and returns the identifier of the record it created". The first sentence sells projects. The second decides whether the project is viable.

Ask for your systems' API documentation before you approve the budget, not halfway through the build. Half an hour of reading removes the single biggest source of overrun in automation projects.

When digital processes are enough

Not all digital processes should become automated processes, and pretending otherwise would be selling you work you do not need.

Automating a process that runs three times a year rarely pays: building and maintaining the workflow costs more than it saves. Nor does it pay while the rules are still changing month to month — automation freezes a decision, and freezing the wrong one costs more than doing the work by hand. And in some processes the slow step genuinely is human judgement: there, automating the gathering of context helps, automating the decision does not.

The criterion remains volume multiplied by stability. If you need concrete candidates, we keep a list of 20 business processes almost every company should automate, organised by department.

The question for your next meeting

Pick one of the digital processes your company already treats as solved and ask a single question: in the last 30 days, how many times did someone copy a piece of data from one system into another inside this process?

If nobody can answer, measuring that is the first project. If somebody can, you have just identified the digital process most likely to be worth turning into an automated one.