Let’s proceed with order: can we give a better definition of planning and artificial intelligence?
I’ve worked in Finance for many years and I’ve seen that when it comes to budget and forecast cycles, many companies spend a huge amount of resources (people, time and sometimes multi-millionaire softwares) to complete these tasks. The flaws are always the same and mainly they’re about the mis-conception that budgets and forecasts aim is to predict the future, while their scope is about setting targets. In a simplified way, traditional planning is a two phases activity: setting a vision (a representation of the future) and understand how and when to allocate resources to get there. Then execute and keep track of success / failure and understand the impact of the changing internal and external environment.
The way it has been represented by scholars is about maximizing utility, reduce uncertainty, providing benchmarks to track progresses.
Artificial intelligence is not as easy to define, because it cannot be identified in a final product or service and has many (interdisciplinary) fields of application like learning, language processing, speech recognition, motion, data mining, gaming, robotic etc… Minsky says that “Artificial intelligence is the science of making machines do things that would require intelligence if done by men.”. To me, the definition is not the central point, the evolution of AI is: today AI exploded because of cheap parallel computation, big data and better alghorytms. This gave birth to powerful neural networks and expert systems and it’s easy to say it’s going to continue in this direction.
So we have a complex recipe with vision, resource allocation, utility, uncertainty, decision making, neural networks and expert systems. My idea is to look at planning as a set of moments and think if AI can help or not in each of them.
Setting a vision / a representation of the future: artificial intelligence will be able to analyze time series and estimate the impact of a variety of factors; it can probably return percentages of success and if the programmer can factor many parameters into the system, it can help to optimize decisions, but for sure the reality will always be more complex, articulated and un-predictable than any tool can say. If we think short term, an expert system, where all possible sets of available knowledge and rules are listed and interconnected, can be sufficient. But when it comes to long term, a neural network is more suitable, because it has the advantage of not requiring prior assumptions about possible relationships, it’s going to discover them itself and learn as they happen. The conclusion is, AI can assist in decision making, but not replace humans easily.
Understand how and when to allocate resources: the “simple” calculation of resources needed in theory does not require AI, but simple computational power. It can be improved if big data is available: if internal resources and figures are there and external data are accessible, AI can identify and discover hidden patterns, not-so-obvious relationships and connect the dots, making allocation more accurate. Let me do an example. When you have to staff your call center, you don’t need artificial intelligence to understand that when you do promotional activity you are going to receive more contacts. You just need to read the data to discover at what time of the day your customers are calling. But you can apply AI to discover a link between the calls and the clarity of your digital communications; if the way you display your message, the colors you use, the words and the sentences you write are somehow misleading and people call to receive information and clarification, that’s not immediate to recognize; then advanced tools can definitely help.
Forecasting: I think this is the area where AI can help the most. The standard company forecasts in time units made of months or week or sales canvasses. Some activate the process every quarter, some proceed with rolling forecasts and provide mainly monthly updates. The dream would be having a daily view. Companies are struggling to build daily dashboards just to recognize the actual performance and we are still far from updating a forecast on a daily basis; this happens mainly for two reasons: accounting, even when it’s fast, does not capture all the internal business phenomena on a daily basis (this is to have an updated starting point) and forecasts do require a lot of human interactions between departments, divisions and sometimes different countries. Artificial intelligence can help to spot external trends, run micro analysis on certain kpi’s and see how they are affected by changing environment and support the decision making process (especially in what if scenarios). But this is only half of the story. Even if AI can assist in optimizing the decision making process of selected decision makers and the way they influence each other, companies will then still need to optimize the execution after the preferred path is selected. In the end we can say that even when you know the right road to take, you still need a very disciplined labour force, reactive middle management and excellent communication along the command chain, which are not easy at all to achieve; and, last but not least, you should need to entirely change the way targets are managed, because if the new decision / tactic enters in conflict with goals previously set, people will simply resist and make the process less effective.
Artificial Intelligence Techniques Enhance Business Forecasts
ARTIFICIAL INTELLIGENCE (AI)
The Future of AI: a Ubiquitous, Invisible, Smart Utility