AI and Analytics at Scale

Chapter 6. Where to Go from Here

The first step forward from now is to decide whether or not your own organization might benefit from a more scale-efficient approach to AI and analytics.

We know from many conversations that some people who have not yet tried or observed these approaches assume they are implausible or that they do not produce substantial benefits over what they’re already doing. But we also know from the companies we’ve worked with and from what we’ve observed that many old trade-offs—cost and scalability, or scalability and performance, or any of those with reliability—are unnecessary. Truly scale-efficient and cost-efficient AI and analytics systems are feasible with the methods we’ve presented. In short, things can be very different with the right approach.

If you do want to take advantage of greater scale-efficiency, how do you put these ideas into practice?

Perhaps you are lucky enough to be building a system from scratch. In that case, you can take a greenfield approach to putting the solutions we’ve described for efficient large-scale development and production systems into practice. But for so many others, who already have a system in play, how do you go forward? In essence, how do you rebuild the plane while you’re still flying it?

The good news is that you don’t have to tear everything down and rebuild your entire system all at once. The inherent flexibility afforded by the approaches and technologies we describe also lend themselves to adoption in ...