When I hear statements like “Every company should have a Chief Analytics/Digital/Innovation officer.”, it makes me cringe. Why? Not because these job roles are not necessary. But by saying that EVERY company needs them, we are taking the route of force-fitting a single formula on every organization across every Industry.

No two organizations will be the same regarding organizational structure, culture, current capabilities, etc., so what may work for one (like having a chief analytics officer) might not work for another. Following the “one size fits all” path can be dangerous for any organization.

Trying to replicate what everyone is doing in terms of AI strategy can be suicidal in the long term (or maybe even in the mid-term)

Your organization is unique.

In my perspective, every organization has a unique DNA. Hence any AI Strategy formulated needs to account for the unique challenges associated with THAT specific organization. In the subsequent sections, we will review a framework that can be used to evaluate if your organization has the “ingredients” for a successful AI strategy.

Ingredients to create a successful AI strategy

So, what are some of the circumstances/ingredients that need to exist for a successful AI Strategy? There are many schools of thought on this, but as per the framework I like, they are the following [1]:

• A strategic Dilemma or trade-off must exist.
• The nature of the problem is driven by uncertainty.
• An AI algorithm can reduce uncertainty enough to change the balance in the strategic dilemma.

A Real-world example: Optimal Inventory Placement

I plan to use the example of an e-commerce company to go through the framework; The company in this example is Otto, a German e-commerce venture. Now let us review the three ingredients in this example.

Ingredient 1: A strategic Dilemma or trade-off must exist.

Otto was experiencing delayed deliveries to the customers. Whenever there was a spike in delayed deliveries, it was also accompanied by a significant increase in returns. Impatient consumers of the modern age would buy that item in the store and return the product from Otto whenever they received it. Even when Otto had sales, returns added to its costs.

Otto obviously could not have stocked everything at every distribution center to avoid delayed deliveries. So, their dilemma was –

What Inventory do we need to hold and where?

Ingredient 2: The nature of the problem is driven by uncertainty.

The uncertain nature of customer demands led to Inventory stock-outs and delays.

Ingredient 3: An AI algorithm will be able to reduce uncertainty enough to change the balance in the strategic dilemma.

Leveraging a database of 3 billion past transactions and hundreds of other variables, Otto created a prediction algorithm that can now predict with 90 percent accuracy what products it will sell within a month.

This high level of prediction accuracy allows Otto to set up a new way of organizing logistics, Inventory, new warehouse locations, local shipping, and customer delivery guarantees.

It is not that simple as well.

This is just an example, and the ingredients mentioned here are high level to indicate that a structured and more thoughtful approach is needed. It is more prudent to take a year or two to do an extensive exercise like this for “dilemmas” across your operations/organization, run pilots, and finalize the AI capabilities you need to develop.

Only when you have finalized a list of AI capabilities can you decide what type of analytics organizational structure you want to develop, what roles you need, and what the desired skill sets should be for those roles. There is also an approach for that but it is not within the scope of this article.

Customize and create…. don’t just replicate!

1]: Gans and Goldfarb 2018