Another view: sunset your enterprise architecture every two years

'Temporary EA' may better reflect the reality of today's business technology, while injecting fresh energy into planning processes.

Enterprise architecture (EA) is something that is intended to outlast any technology it is initially designed around. In an ideal world, EA should provide a roadmap for the years to come, with a technology vision, as well as practical guidelines for purchases, implementations, and standard-setting well into the foreseeable future.

Photo: NASA

That's the ideal, at least. But with businesses and technology changing so fast, it often doesn't work out that way. EA plans tend to fall apart quickly. Maybe it works better to accept that new EA plans need to be built from scratch every two years -- with any previous plans sunsetted.

That's the view of at least one industry visionary, who suggests that accepting a "temporary EA" approach may be less costly and more flexible in today's environment. Robin Meehan, chief technology officer with Smart421, put forth a proposal that organizations "stop trying to make EA practices that are self-sustaining over the long term," and opt to put a 24-month time limit on EA plans.

That will sync EA more with the cyclical nature it already tends to follow -- affected by changing business environments, budgetary cycles, and executive and staffing changes. Eighteen to 24 months seems to be the natural course of even the best-laid plans, before they begin to calcify or are rendered outdated or irrelevant by the changes swirling about.

Meehan's proposal has a ring of government program management to it. That is, many programs -- local, state and federal -- have "sunset" provisions attached to them, meaning that no further funding will be made for the upcoming year until the program's reason for existence is justified. As Meehan explains it:

"Accept that EA is a cyclical function... So let's embrace it rather than resist it - accept that rather than trying to create long-running functions, create EA functions that are designed for maybe 18 months maximum duration. Focus on building artifacts that consider a maximum of a 24 months' time horizon, don't discard them once the EA team ramps down, but deploy only a limited capability to maintain them. To use an analogy, consider all outputs from the EA function to be like sandcastles - designed to survive only up until the next tide (aka the wave of change) and not over-engineered to survive longer. Accept that these outputs will need to be refreshed/recreated in two years' time, with a new team of probably largely new people that bring fresh energy."

Is it reasonable to put a shorter time limit on EA planning? Is a two-year turnover of plans too long, and should be more frequent? Will it even make a difference?