Digital Transformation in Government: Sequencing the Work

The big-bang replacement problem

Replacing a legacy system in a single cutover concentrates all risk at one date, years after funding began. If the date slips — and it usually does — the program has produced no delivered value to defend.

Incremental modernization inverts this: each increment moves one workflow, one data set, or one interface, and each increment is independently useful.

A sequence that works

The order below front-loads learning and defers irreversible decisions.

• Document the current process as performed, including the workarounds staff rely on.
• Expose read-only APIs over the legacy data before changing anything.
• Rebuild the highest-friction user-facing workflow against those APIs.
• Migrate data domain by domain with reconciliation reporting at each step.
• Retire legacy functions explicitly as replacements reach production.

People decide the outcome

Staff who have operated a legacy system for a decade hold requirements that exist in no document. Involving them as design participants rather than training recipients surfaces those requirements while changes are still cheap, and materially improves adoption.

Measuring the right thing

Track service outcomes — processing time, error rate, rework, and applicant experience — alongside delivery milestones. A modernization that ships on schedule without improving those numbers has not modernized anything.