If you want to implement EAI, inventory all the system interactions in your company--you'll be surprised at how many there are--and identify the most likely EAI candidates. Perhaps you have systems that are not communicating reliably, or systems for which "the guru" is no longer with the company. Then there are systems so critical that to lose them would cost big bucks every hour they're down. These are all good candidates for the first phase of an EAI deployment.

Now pick a couple of low-threat interfaces and use them as learning experiences. Choose test systems where you have in-house experts in the source and target systems, so mistakes can be patched easily. Run a sample implementation with these low-risk interactions, and then start Phase 1 of the rollout.

Getting high-risk integrations into Phase 1 is a tactical choice: The biggest business benefit is going to come from stabilizing unreliable interfaces or replacing homegrown integration code that no longer has an owner. The process of integrating your enterprise will take an investment of many labor hours, so showing a positive result early on is important.

Finally, dedicate staff to your EAI system long-term. It will function as the switch in your application network, handling all your app interactions. If it goes south, your apps will be without connectivity.

Also, have some trained EAI developers at the ready. Upgrades and system replacements will generate plenty of work for them. By using such EAI professionals, you're in effect not adding staff, but merely shifting responsibilities to ease the project teams' workload.

Don MacVittie is an application engineer at WPS Resources. Write to him at dmacvittie@nwc.com.

Post a comment or question on this story.