Where should you incorporate AI?

Artificial intelligence (AI) is clearly a growing resource for technology adopters. We see it taking center stage in our influential conferences; we see articles telling us about many successful use cases; and we see it everyday in TV, movie & commercial placements. "AI" is everywhere. Except where it's not.

As an advocate for technology solutions at a large enterprise, I see a definite rise in proposals that are leaning right into that FOMO reaction to AI. Does anyone else feel like we entered another period that strongly echoes one we felt not so many years ago about "Big Data"? (Kudos to the marketing teams who are so good at pushing industries toward change using data!)

I get it. AI is a useful resource to help us solve problems. But it cannot be a resource that is only available to a few. As an advocate who also manages to a budget, I created this strategic "anchor visual" for practical implementation patterns of AI in an enterprise. To further explain:

1. We want to see the largest amount of implementations embedded in our workflows and existing systems, with an approach that provides an ease of adoption to the widest number of people.
2. When that approach fails, then think about using a 2nd tier. For example when we really do need more data to create a decision or we really do need functionality that can't be available in our systems. In these scenarios, we prefer to utilize an enterprise level "2nd tier" environment that allows us to utilize the black-box or pre-built engines. (Yes. Queue the "to the cloud!" advertising here!) Think of it as standing on the shoulders of giants! Yes, we want to do that!
3. When that approach doesn't work, and we truly believe it's time to create some "secret sauce" intelligence, then we invest in those specialized data science teams, approaches, and tools. We need to solution our AI implementations from the outside in!

Too many times, I see teams investigating or vendors proposing the solutions in the opposite order. They start at #3 with customization, inside out.

In your next AI solution or design session, I ask you all to stop and think: Is there a way to provide an advantage of economies of scale with a different architecture implementation? Can you work outside in? If you're a technology creator, is there a way you can embed the AI resources in our existing platforms and collaborate with our other application providers to keep our users in their workflow?

As we move into the adoption phase of this highly marketable technology, remember that scale and maintainability are key considerations in our solution architectures. I can't wait to see what's next!