Three tips for successfully building out of box (prepackaged) AI Apps and use cases in your product

Three tips for success in prepackaging out of box AI apps and use cases.

As business applications and platform vendors add more AI based use cases to their products, many have started wondering if the squeeze is worth the juice. If adoption of such use cases by their customers is a measure of success, many have failed and very few can claim success. Having led product teams on both sides of the aisle - vendors of AI use cases and consumers of AI, let me share three tips for success for the vendors of AI apps and use cases.

Tip 1: Experience your customer before you build experiences for them: Before you build any experience for your customer, whether it is a UI based experience or a "behind the screen decision support", experience your customer. Do not just understand them, experience them! And while four-hour interviews give you some insights, I have found them insufficient. First, ask if your product managers, data scientists, engineers, go to market team members or anybody else in the chain of delivery have worked as a customer in their career? If you have someone, consider yourself lucky. Learn from their real life experience about the needs of the customer, how they generally organize to meet these needs, what value do they perceive, how important is their voice overall in the organization, do they have sufficient funding to support AI use cases, how do they get human in the loop while using AI, how do they balance human learning with machine learning, what are the day to day issues, what kind of reporting is needed, are there common vested interests and conflicts of interest between personas,.. As you can see there is so much to learn to build AI that works. There is no way a four-hour session would suffice. It may be a good start though. Check with your customer if your team members can shadow real life personas for a week, month, maybe more. Also, you are mass customizing AI. Have that lens always on when you shadow. Not all customers have the same way of working. And that leads me to the second tip.

Tip 2: Determine your Customer Decoupling Point (CDP): I define CDP as the point at which the out of box product ends and customer needs to start customizing. This must be looked upon from a customer's point of view. It is the set of out of box product features where mass customization ends, and each customer starts implementing their unique "to-be" scenario. It is that point at which a product starts transforming into a customer solution. Anything beyond CDP is effort, resources, and budget that the customer must pay for. There are many options here: sometimes it is better to keep the CDP away from the customer if you think every customer has unique needs; you could also create industry verticals and bring the CDP closer to the customer; another way would be to standardize certain drop-down lists which brings the CDP closer. After you determine and execute on the CDP, the solution journey starts for your customer. Many call this "the last mile".

Tip 3: "The last mile" between CDP and go-live: I have always believed that success of AI (or any other product) depends on how much support is provided in the product for the last mile. The realm of AI products gets a little more difficult when it comes to the last mile. Here's why: Some of your customers might be new to AI environments, others might have matured with AI implementations from home grown data science or other vendors, while others might be on the learning path. You need to provide support in your product that can adjust to all three levels and assume various environments. You might be the only AI apps vendor, or you live with other home grown/ other vendor implementations, or you might be replacing existing implementations. There is a human aspect to this as well. Are you replacing organizational memory (knowledge learnt by folks tweaking deterministic rules), or supplementing it? What does success mean to your customer? How does your product support measure and monitor success? And as you might have guessed from your own experience, I have hardly scratched the surface of "things to consider" here. Simply put, build features that remove the anxiety of implementation/ maintenance, can co-exist with various AI environments, and help your customers measure and communicate value.

There are many more tips to share, and more depth to explore in each tip. For now, I would ask to remember the phrase "Experience your customer before you build their experiences", and that should take you on the path of building "Practical AI apps" for your customers - the folks that sign the check.