Borrowing a Mental Model from AI for Business Strategy

I recently started a certificate program in Artificial Intelligence from Columbia University. Although the course itself is technical in nature, as I progress through I've started making connections to apply in the world of strategy. I'll be sharing these thoughts and observations on the connections between the technical material and business strategy here.

One of the first lessons I've completed after the introduction lesson is on Search Algorithms for solutions.

Think about search algorithms in this way; when you play a game like poker, chess, video game and any other game you can think of there are multiple approaches to figuring out how to get to the END state or GOAL state as you would call in the world of AI. There are many different search algorithms that have many different approaches when they solve problems because there is not one path to the end goal but many different paths.

Now in AI there's many types of search agents that can be leveraged based on the scenario such as uniformed search agents (Breadth First Search, Uniform Cost Search) there's also informed search agents (this is my next lesson) and many other algorithms (these are in much later lessons) that can be used to solve problems.

However one thing I quickly realized in this lesson is that the PROBLEM must be defined very precisely and accurately before an agent can be applied (using any one of the search agents and/or other types of algorithms) to solve the issue, aka the search agent must be CLEAR about its goal (state).

Here is where I had an "aha" moment.

In strategy and in the AI space there are many different types of frameworks and agents respectively. Solutions are abundant but problems are unique almost all of the time. In AI there is a precise model to help define the problem (below), in strategy at times I've seen many departments that jump to solutions quickly which may eventually get them to their GOAL but perhaps leaves the best possible solution on the table. Strategy could learn from the AI discipline and learn that defining the problems they wish to solve is just as important if not more than the solutions.

Below is a method used for "problem formulation" in AI which helps design and/or use the most rational and best agent (s) possible:

• Initial State : The state in which the agent starts or is implemented in.

• States : All states that are reachable from initial state by any sequence of actions or all possible states that the agent can take.

• Actions : All possible actions that the agent can perform. Specifically, it provides the list of actions, that an agent can perform in a particular state. This is also referred to as Action space.

• Transition Model : This property describes the results of each action taken in a particular state.

• Goal Test : A way to check, whether a state is the goal.

• Path Cost : A function that assigns a numeric cost to a path or a performance measure

I've taken this model and leveraged it to apply it for business problems:

• Initial State: The state of the landscape and/or business in which our solution is going to be implemented or the strategic inception point well defined. i.e Market Landscape Analysis, Current State Business Maps etc.

• States: Various and/or all possible outcomes after the implementation of the strategy. i.e Future State Market Projections, Future State Business Maps etc.

• Actions: All possible tactics that our strategy can execute. Specifically, we'd create a list of tactics that would help us implement our strategy. i.e Rapid brainstorms, SWOT analysis of tactics and business cases created.

• Transition Model: Here we would define KPI's for each tactic (aligned to Business Goals)

• Goal Test: ROI projections and predictive models for all tactics

• Path Cost: Cost Benefit Analysis for the overall strategy

The biggest observations to take away here is that we should spend more energy, focus and attention defining the problem to get to the most efficient solutions. At my current agency we use many different tools like relationship maps (journey maps), 5 C's analysis, stakeholder interviews etc. to truly identify and deeply understand business problems before we move forward with strategies and solutions.

Feel free to borrow the above approach for your next project.