Digitising an SME with a warehouse: the right order
Short answer
Start with master data (products, customers and prices in order), then sales, then the links between sales and the warehouse, then the warehouse floor, and only at the end AI. That is the order Plurifil, a Portuguese company, followed; its ERP had 35 user accounts in nine roles on 3 August 2026. Reversing the steps turns months into years.
Digitising a company with a warehouse is a staircase, however much it is sold as a leap, and the order of the steps matters more than the speed. It is the order Plurifil, a Portuguese company, followed (online store, ERP, production records, picking), and reversing it is what turns projects of months into projects of years. The ERP on that staircase had 35 user accounts in nine roles on 3 August 2026.
The order that works
The order that works is master data, sales, the links between them, the warehouse floor and only at the end AI.
- Master data first. Products with unique codes, customers without duplicates, prices in one place only. Nothing digital survives messy data.
- Sales. An online store or B2B channel that works, with the local payment methods customers expect (in Portugal, MB WAY and Multibanco references) and certified invoicing connected.
- The links. Orders flowing from sales to the warehouse without retyping; stock flowing back the other way.
- The floor. Picking with confirmation, production records entered at the moment they happen. Paper is the last thing to go.
- AI. Once the data flows, AI has somewhere to work: reading documents, sorting, proposing. With messy data, it only multiplies what is already wrong.
The three mistakes that cost years
The three mistakes that cost years are starting with the software, leaving the project open-ended and automating approval.
The first is starting with the software instead of the process: buying a tool before mapping who does what guarantees that the tool adapts to the chaos instead of fixing it.
The second is the open-ended project. “We will digitise as we go” means there is never an end date, a budget or a person in charge. Each phase needs a scope closed in writing, with an end date and a named owner.
The third is automating approval. The system can propose everything; releases to production and decisions involving money need a person who approves. Skipping this principle lasts until the first scare.
Where to start this week
Start with the process that annoys the team most: follow it end to end, once, with a sheet of paper in hand, and note where it begins, who touches it and where it ends. That one-page map is worth more than any software demo, and it is the first deliverable of Draypec’s operations diagnostic.
Questions
Can we start with AI and clean up the data later?
It does not pay. AI works on the data that exists; with duplicate product codes and prices kept in three places, it only multiplies what is wrong. Clean master data is the first step, and the others depend on it.
How long does it take to digitise a company with a warehouse?
It depends on the order and the scope. With phases closed in writing, each with an end date and a named owner, the project is measured in months. Left open-ended, it is measured in years. At Plurifil the sequence was online store, ERP, production records and picking.
What is an operations diagnostic?
It is Draypec's first step before proposing any software: follow one process end to end, with a sheet of paper in hand, and return a one-page map of where it begins, who it passes through and where it ends. That map sets the scope of the first phase.