Artificial Intelligence for Business

CHAPTER 7Thriving with an AI Lifecycle

Once the system has been developed and deployed, the work is still not complete. As with all other software, an AI project requires regular upkeep and maintenance. Even if you choose never to implement another new feature, there will always be bug fixes, server updates, and other forms of maintenance to keep the system going. Although this commitment is typically not as large as the system's initial implementation, an ongoing investment is still required to “keep the lights on.”

When building an initial project estimate, you must also ensure you have enough resources secured to maintain the system once it goes live. It is typical to see a spike of issues shortly after launch, when end users are first accessing the system. This is so common in fact that there is a name for support when a system first goes live: hypercare. It would be a mistake to assume that the system developed will be 100 percent correct and a guaranteed success right out of the gate. Although a good 80 to 90 percent of the project will likely be correct, users will get hung up on the balance. As time progresses, maintenance resource requirements should stabilize. This stable point is what you should use as the estimate to secure the next year's maintenance budget, augmenting it accordingly if large updates or migrations are planned. A few critical activities must be performed once a system is live in order to ensure that the project is healthy and that the benefits from ...