budiman.ai

· 1 min read

Enterprise Architecture, Year Three

A reflection on what the three-year enterprise architecture plan at JCorp actually looked like in practice — what got built, what got cut, and what the year-three exit shows about the original year-one bets.

  • enterprise-architecture
  • leadership
  • case-study
  • johor

We wrote a three-year EA plan at JCorp because the alternative was a three-month plan that got rewritten every quarter. The plan was deliberately under-detailed in the later years and over-detailed in the first — the only honest way to plan that horizon, because the unknowns compound. Year three has arrived, and the post-mortem is more useful than the original plan.

Three things from the year-one bets held up. The Group-wide data backbone, the four-layer EA model (business / data / application / technology) as the operating spine, and the explicit refusal to centralise semantic ownership. Each of these survived twelve calendar quarters of business change because we had bet on a direction rather than a destination.

Two things from year one got cut. The first was a planned governance layer that turned out to be friction without value — the people who needed the discipline already had it, and the people who didn't were not going to acquire it through a steering committee. The second was a vendor lock-in we had quietly accepted for the sake of speed; year two made it expensive enough to undo, and year three is healthier for the undoing.

The year-three exit is not a victory lap. It's a recalibration. The next plan won't be a three-year plan again. It will be a six-quarter plan with explicit re-plan gates, because the AI-era cadence does not tolerate three years of unrevised commitments. The original plan was right about its direction and wrong about its tempo. That's a useful thing to learn.