Digital transformation for SMEs: connect your systems, digitize your processes

For an SME, digital transformation works best as a sequence of small projects, each fixing one process and leaving a working system behind. Start with a map of how work actually flows, connect the systems you keep before replacing any of them, and digitize the steps that cost the most manual effort. Replace a core system only when connecting it would cost more than migrating. This keeps the budget in phases you can approve one at a time, and the business keeps running throughout. Our digital transformation consulting ends in working systems, not in a strategy document.

Replace, connect or extend: what should happen to legacy systems?

Keep and connect a legacy system when it still does its job and the vendor supports it; the problem is usually trapped data, not the system itself. Extend it with custom tools when the core works but processes around it don’t. Replace it when support ends, it cannot exchange data at all, or workarounds dominate daily work. Most SMEs need the first two.

Approach	When it fits	Watch out for
Keep and connect	System works, data is trapped	Old interfaces; test with real data volumes
Extend with custom tools	Core is fine, the processes around it are not	Keep one clear system of record per data type
Replace	Support ends, no way to exchange data, workarounds dominate	Data migration and retraining are most of the effort

How is a phased transformation put together?

Phase one is a process and system map: every system, every data flow, every manual step, drawn with the people who do the work. From it comes a list of changes ranked by effort and impact. Each following phase takes one item from the top: an integration that removes double entry, a portal that replaces email orders, a digital form that replaces paper, a dashboard that replaces a monthly Excel. Each phase ends with something in daily use.

The engineering underneath is the same we use elsewhere: integrations that are idempotent (a retried run never creates duplicates), logged and monitored; clear ownership of each type of data; and AI where it saves effort, such as sorting incoming emails, with people confirming uncertain cases. The startup discovery and outreach platform we built for a SaaS client runs a daily multi-step pipeline on these principles.

Example: a family-owned wholesaler in Ticino

Illustrative scenario, not a client project.

A wholesaler of building materials with 35 employees supplies contractors in Ticino. The ERP dates from 2009 and has no API, but it can import order files from a folder. Orders arrive by phone, email and sometimes fax. Delivery notes are printed, signed on site and scanned back weeks later.

Phase one maps the order-to-invoice process and shows that most errors come from retyping orders. Phase two adds a B2B order portal in Italian and German: contractors see their prices and past orders, and each order is written as a file the ERP imports every fifteen minutes. Phase three replaces paper delivery notes with a tablet signature and a PDF sent to the customer. Phase four adds a dashboard of orders, deliveries and open invoices. The ERP stays; its replacement is reconsidered when vendor support ends.

What should SMEs plan for?

• Data protection. The revised Swiss FADP, in force since 1 September 2023, requires privacy by design and by default for data processed in Switzerland, as the GDPR does in the EU. Decide hosting location, access roles and retention during architecture. If data is stored abroad, your privacy notice must say so; the FADP requires naming the countries.
• Payments and invoices. A structured payment reference on each invoice, such as the one on the Swiss QR-bill (in use since 30 September 2022) or the Creditor Reference in the SEPA area, lets bank credit notifications (camt.054) be matched to open invoices automatically. That is often an early, low-risk win.
• Standard business software. Many SMEs run accounting tools such as bexio or Abacus in Switzerland, or Xero or DATEV elsewhere; most publish APIs, so they can be connected rather than replaced.
• Languages. Portals, documents and messages often need more than one language, such as German, French and Italian in Switzerland. Build this into the data model from the start.

How does a project run?

1. Discovery. Workshops and interviews produce the process and system map and the ranked list of changes.
2. Architecture. For the first phase: target process, data ownership, integrations, hosting and data protection.
3. Build. In short iterations, tested with real data and the people who will use it.
4. Launch. One team or location first, with the old process as a fallback.
5. Iterate. Review the result, then decide on the next phase.

After discovery you receive a fixed-price or phased proposal. If your most pressing gap is an internal tool, see internal business systems; for how we work in general, see how we work.