Digital Transformation Priorities: How to Choose Without Stopping the Business

Most digital transformation plans do not fail on the technology. They fail on the list.

A company identifies fifteen problems, has budget for three and real capacity for one. It picks the three most visible ones, starts them all at once, and six months later has three projects at 60% and an operation slower than it was before.

Setting digital transformation priorities is not about choosing what delivers the most value. It is about choosing what the operation can absorb while it keeps invoicing.

The constraint is not budget, it is the attention of people who already have a job

A 2025 study in Frontiers in Sustainability ranked nine digital transformation enablers for SMEs, using the Best-Worst Method with a panel of eight experts. The result contradicts the usual intuition: technology and digital infrastructure came fifth, with a weight of 7.39%.

Ahead of it sat digital leadership (22.55%), adaptive digital strategy (17.86%) and dynamic resource management (17.30%).

Read the study carefully: the sample is ICT-sector SMEs in the Middle East, and the weights do not transfer cleanly to a manufacturing company in northern Portugal. But the ordering matches what actually happens on site. What stalls an automation project is rarely the tool. It is that nobody has three free hours a week to steer it.

Which makes the useful question not "what should we automate first?" but "how many hours a week do we genuinely have, once the operation is done?".

Before prioritising, make sure you are measuring rather than guessing

A priority list built from opinions always produces the same outcome: the process belonging to whoever complains loudest wins.

If you are not yet certain there is a structured problem to solve, the seven signs that a company needs digital transformation work well as a prior diagnosis. Prioritising before diagnosing is choosing in the dark.

For each candidate process you need three numbers:

• Frequency — how many times a week or a month it runs.
• Time per run — measured, not estimated from memory. Timed once, under normal conditions.
• Cost of error — what happens when it goes wrong, and how often it goes wrong.

The third is the one most often skipped and the one that most changes the ordering. A five-minute process that produces one wrong invoice a month can cost more than a thirty-minute process that never fails.

The arithmetic that orders the list

The method is simple arithmetic. An example, with illustrative figures rather than market data:

Process A — supplier invoice entry. 40 a week, 6 minutes each.

40 × 6 = 240 minutes a week, or 4 hours.

4 × 44 working weeks = 176 hours a year.

Process B — holiday requests. 12 a month, 3 minutes each.

12 × 3 = 36 minutes a month.

36 × 12 = 432 minutes, or 7.2 hours a year.

Both turn up in the same meeting with the same level of irritation attached. One is worth 176 hours, the other less than a working day per year. Automating the second one first is a style exercise.

These numbers are a worked example, not a benchmark — what travels is the method, not the values. To turn hours into euros using official labour cost data, the calculation of what a manual process really costs covers the second half of the sum.

The second criterion: can the process be touched mid-month?

This is where most lists fall apart. A process can be worth 200 hours a year and still be a bad first project, if it cannot be touched without risk.

Four questions, per process:

• Can it run in parallel with the manual version for two weeks, without duplicating records?
• If the automation fails at three in the morning, what happens at nine?
• Are the rules written down, or do they live in one person's head?
• Is there a period of the month when touching this process is out of the question — month-end close, order peak, campaign?

A process that answers badly to two of these is not a bad candidate. It is a candidate for the third quarter, once the rules are documented.

Shadow, parallel, cutover: the sequence that keeps the operation running

The way to automate without interrupting anything is never to switch off the manual path before the automated one has proved itself. In practice, three phases:

1. Shadow (1 to 2 weeks). The automation runs, but its output goes nowhere. Every day you compare what it produced against what the person produced. The differences are your fix list.
2. Parallel (2 to 4 weeks). The automation becomes the authoritative version. The person stops executing and starts validating, which typically costs ten to fifteen minutes a day instead of hours.
3. Cutover. The manual path is reserved for exceptions. It still exists, documented, for the day a supplier integration changes without warning.

This phasing has a real cost: it stretches the project by four to six weeks and forces duplicated work during the shadow phase. In exchange, it removes the scenario where the company has no working process at all on a Wednesday morning.

If you need concrete candidates for phase one, these 20 business processes almost every company should automate are organised by area, with criteria for deciding where to start.

What stays off the list, and why

An honest priority list needs a second list beside it, naming what will not be done this year.

Typically excluded:

• Processes with unstable rules. If the rule changes every quarter, automating it means buying permanent maintenance.
• Low-frequency processes. Below two or three runs a week, the effort of keeping an integration alive rarely pays for itself.
• Processes that depend on replacing the ERP. Automating on top of a system due for replacement within a year is disposable work.
• Processes where judgement is the work. Negotiating with a difficult supplier is not a workflow. It is a conversation.

And then the limit that matters most: automation does not fix a badly designed process. It reproduces it faster and more consistently. If an approval chain has four signatures too many, automating it gives you a four-signatures-too-many chain that runs on its own.

The European context, without the rhetoric

According to the 2025 edition of Eurostat's publication on digitalisation in Europe, 73% of EU businesses reached at least a basic level of digital intensity (2024 data), ranging from 50% in Bulgaria to 93% in Finland. The Digital Decade target is for more than 90% of SMEs to clear that bar by 2030.

Nationally, the 2026-2027 action plan of Portugal's national digital strategy keeps SME digitalisation as one of its axes, with funding attached.

What this means for a decision-maker: money is the easy part. Funding lines exist, and consultants exist to handle them. What no funding call delivers is the internal hours of the people who understand the process — and that, not the budget, sets how many fronts you can open at once.

The decision to make tomorrow

Take your list and ask it one question: which of these processes can run in parallel with the manual version for two weeks, without risk and without duplicating records?

If one can, you have your first project — start by measuring its frequency and time per run this week.

If none can, the problem is not prioritisation. It is documentation, and that is the real priority.