You’ve probably received the advice at some point in your life that “context is everything.” While it’s probably wise to avoid relying too heavily on clichés to guide your life, when it comes to using AI successfully, context does in fact matter a lot.

Enterprises often seem to operate between two bipolar views of AI. On the one side, there’s the optimistic idea that you can just plug in an AI platform and it will immediately start spitting out predictions to radically reshape the business. On the other, is a more fearful, pessimistic view of AI’s potential, in which the black box of AI algorithms means that we should be skeptical because we can’t understand how the technology arrived at its predictions.

In reality, as with so many things in life, the truth lies somewhere in between. AI isn’t perfect, but it shouldn’t be feared. And when it is fed with not only sufficient data, but also the right context to make sense of that data, it can produce dramatic insights for businesses.

Recently, I had to the chance to speak with Anil Kaul, CEO of Absolutdata. Since the company’s founding 17 years ago, Absolutdata has focused on predictive analytics. And in that time, Kaul and the organization as a whole have developed a unique theory based on their client experiences about what makes AI successful. His theory is well thought-out and worth exploring in depth here.

We all need a coach

In my previous work on implementing AI in the enterprise space, I’ve focused on analytics ops, which is a term for the need to continually train and evolve AI models to produce reliable results. Kaul has a similar view that AI cannot operate in a vacuum and advocates companies embracing a concept he calls “AI coaching” in which AI technology is provided with crucial business context to ensure its predictions are safe. After all, AI doesn’t know everything on its own. It needs guidance to be truly effective.

The idea of AI coaching has become more central to Absolutdata’s approach as the company has evolved over the past few years. As Kaul told me, “Everybody talks about AI. Some people know what they’re talking about; some don’t. But the focus of the discussion is more on AI than the techniques that lie underneath. What people need to realize is that when you use AI in an enterprise setting, with a focus on using the predictions for scalable business impact, you need more than just the AI itself. Creating that scalable business impact is not just about the AI.”

Kaul said that companies need four things to come together to really make AI operate at its best. AI coaching is at the core of this four-point strategy, which I’ll outline below.

  1. Understand how to use core AI techniques: Like so many spheres in the tech landscape, if you don’t understand the techniques that drive a new technology, you’ll likely not be able to use it optimally. Companies need to take the time to invest in truly comprehending what drives AI, including a large amount of data needed to make predictions reliable, and the patterns it seeks to recognize in those datasets. This not only sets the stage for using AI wisely, it also helps to increase trust for those who will actually be using it.
  2. Develop analytical frameworks: Kaul emphasized that operating within an analytical framework is vital to successful AI. “There are lots of frameworks to solve fairly complex business problems,” he said. “If you don’t use one of these, you’re making your life more difficult.” As an example of a framework, Kaul offered a company trying to predict future customer purchases based on that customer’s history and journey thus far. It’s a foundational approach of a specific business problem that the company hopes to solve by using AI. This helps guide the entire AI process. “Often companies try to make predictions blindly, with total faith in AI,” he said. “They do not understand that AI needs a lot of data before it can perform well. And that you need to provide context for the AI. An analytical framework helps to ensure you have enough data and context in place for AI to work well.” He recommended mapping out the structure of data gathering that would take place prior to using the AI and how the AI prediction will be used as a necessary step in establishing an analytical framework.
  3. Establish the business context: This is where context and AI coaching comes into play. Once that framework is laid down, companies need to ensure they provide the AI with the business context for the particular problem it’s trying to solve. “AI can make predictions, but it is not good at context on its own,” Kaul said. “For instance, it’s not good at incorporating everyday business rules. Anyone in sales knows that new products often have more appeal more than those that have been around for ten years. But AI doesn’t know this. It thinks all products are equal. So you have to constantly teach AI context so it’s not just pushing the most successful products over the past ten years to customers instead of new products that would be more interesting.” Kaul offered other examples of business context and rules that must be integrated into the AI model, such as knowing that customers don’t want to receive too many offers and that they have a budget and so it is only worthwhile to send product offers that fall within that budget.
  4. Technology puts all of these things together: Finally, Kaul stated that AI must be scalable to be effective. The models have to be able to be put into production across the business. Once a model proves itself with one product, companies need a way (such as an AI platform) to then take that model and easily expand it to other products and use cases.

Context Trains the Models

The idea of AI coaching is worth further exploration. The fundamental idea here is that AI needs to be built with guide rails and in a way that makes retraining and evolution as easy as possible. But Kaul and I agreed that this training doesn’t need to be a constant headache and in fact, should not be done manually.

“The first thing you should do when implementing an AI system is to ensure you are building self-learning models,” he said. “Then you don’t need to have people go in and retrain the model.” The algorithms used should be self-correcting.

And then once those models are trained, AI coaching can play its full role. Kaul said an AI coach could be a person who is in charge of setting the guide rails up to work automatically. The coaching brings in the context for the specific business problem to make sure that the models are not biased or predicated on obvious contextual mistakes. This person tries to minimize the magnification of biases in the models, monitors the results, and ensures the model is working effectively by assessing things like whether all people are getting the same product recommendations (obviously, they shouldn’t be).

Additionally, the coach also seeks out new sources of data to feed into the models. “Your AI is only as good as the information you put into it,” Kaul said. “You can’t be complacent. Once you start with AI, that’s not the end of your journey.”

Finally, Kaul said the coaches should be looking for new techniques, as AI is a rapidly evolving field with new ideas constantly springing up.

Ultimately, the role of the coach shows that for AI to be successful, there has to be an integration of technology with culture.

The Business Impact

Kaul said that companies should be looking at AI as more than just a way to do what they’re already doing. Ultimately, companies should be asking themselves “How are we using the technology to do things we’ve never done before?”

“The conventional approach is to take AI and use it to do something you have been doing better than you have in the past,” he said. “But AI is even better for things you’ve never done before.” He outlined four questions to rethink the approach to AI to get the most out of it in a sales context:

  • What is our message to customers?
  • What channel are we using to connect with them?
  • When is the right time for us connect with them?
  • How can we get them to respond?

AI can be used to make recommendations for each of these questions. Kaul said Absolutdata works with customers to create scalable business impact by combining these four focus questions with the approach I covered above. “And we’ve seen fantastic outcomes when you follow this strategy,” he said. “You can identify leads you didn’t know you have and discover new opportunities.”