Digital public infrastructure has a public-relations problem: its flagship stories are so large they read as unrepeatable. A billion identities. Payment rails moving a continent’s salaries. Registries at civilisational scale. A minister in a five-million-person nation reads the case studies and reasonably concludes: not us, not now, not at that price.
The conclusion is wrong, because the lesson of DPI was never the scale. It is the shape: a small number of shared rails — identity, payments, registries, data exchange — that every service stands on, so that no ministry rebuilds them and no vendor owns them. Shape is free. Shape is a decision.
The four rails and the missing fifth
The canon lists four rails: digital identity that answers “who is this?”; payment rails that move money with receipts; registries that hold the facts the state already asserts — people, companies, land, vehicles; and exchange standards so systems ask each other instead of asking the citizen twice. Mid-size nations can stand up credible versions of all four in a budget cycle, and many have.
What the canon under-names is the fifth layer, the one citizens actually meet: the workflow layer — the permits, licences, benefits, inspections and cases that consume the rails. This is where DPI programmes quietly die. The rails get built; then every ministry procures its own front-end sprawl on top, each one a bespoke integration, each one a new dependency, and the “infrastructure” ends up wearing forty different unaffordable hats.
The playbook
The sequence that works for a mid-size nation is unheroic and repeatable. First, treat identity and payments as procurement of proven patterns, not research projects. Second, pick the three registries that unlock the most services — usually persons, businesses, addresses — and give them owners, laws and APIs. Third, and this is the decisive move: deploy one governed workflow engine as shared infrastructure, so every ministry configures its services on the same kernel instead of buying its own platform annually. Fourth, sequence services by queue length, not by prestige — the trade licence before the moonshot. Fifth, certify a public cadre to run the configuration, because DPI operated by foreign contractors is DPI in name only.
Costed this way, national digital infrastructure stops being a mega-programme. The rails are bounded procurements. The workflow layer is one platform decision with a knowable engagement model. The services are fourteen-day configuration cycles run by your own certified team. The first citizen-visible result lands inside ninety days, which matters, because DPI programmes are funded by patience and patience is funded by evidence.
Ownership is the strategy
One test separates infrastructure from procurement theatre: who owns it in year seven? Rails rented from a vendor are not public infrastructure; they are a public subscription. The mid-size advantage is agility — a nation of five million can decide, deploy and own faster than a continent can convene a committee. The playbook is not to imitate the giants’ scale. It is to beat them on sequence, and to own every layer the citizens stand on.
Small nation, shared rails, one governed engine, your own cadre. That is the whole playbook. The rest is execution — and execution, unlike scale, is available to everyone.