Professionals in all industries and business units are filling up lists and lists of potential AI Use Cases they could implement. Being able to classify these use cases for an efficient meaningful result, requires practise and domain expertise. AI Use Cases can be broken down to a vast amount of technical and non-technical variables. The more thorough the segregation process, the more successful the implementation will be.
Not being aware of the very first and most basic clustering of AI use cases, is synonymous with your company being at the very rear of the AI race for competitive advantage. Artificial intelligence is not only able to increase your customer satisfaction by developing better products and services, but also to improve your internal operations. Building an app capable of depicting the spoken wishes of your customer into an image, will probably increase sales; training a model capable of identifying your personas so that you can better target your campaigns, will definitely do so. Making sure that electricity is intelligently exchanged between a charging point and a car, will lead to sustainability; reducing equipment failure by applying predictive maintenance will have a similar effect. Generating a digital twin to nimbly improve the integration of your development and design, will reduce R&D costs; extracting knowledge from your customer feedback by means of NLP, will accomplish the same goal. It is important for anybody who intends to generate some sort of value with AI to not only look to the outside, but also to the inside. Therefore, managers should always holistically consider how AI could help their companies become better.
Understanding the previous argument is just the very first step towards mastering the art of AI Use Case classification, further clustering and breaking use cases down is where the real challenge begins. Consider this: is applying computer vision to check burrs on plastic parts the same use case as checking part completeness? What about analysing driving incidents compared to analysing main city events for dynamic routing? Do we classify them as the same use case or are they different? Does “predicting a component breakdown and alerting the driver in advance” fall into R&D, maintenance, security or user experience clusters? Does it fall into all of them? Is it more important to consider the processing power, algorithms, cloud computing services, development environments and types of data sources when disseminating a use case, or is it better to consider business units affected, measurable ROIs, social-political impact, and use case affinity with company strategy and goals? In my opinion there are two important concepts when it comes to scatter AI use cases: taxonomy and variables.
When it comes to taxonomy, the idea is to group the use cases in a top-down approach, starting with a broad taxonomy level (such as the industry), narrowing it down to families of use cases. The following graph shows an example for the automotive industry that we created in AI Shepherds:
Table 1: Use Case Taxonomy
Of course there are no written rules when it comes to define a use case taxonomy. Based on my experience though, going deeper than a 3rd level, would make things too complex to follow. At least in the initial stages of use case taxonomy. In regards to the variables, one should first and foremost differentiate between technical and business related. Here some examples:
- Technical variables: AI discipline, type of learning, algorithms & libraries used, type and amount of data needed.
- Technical variables: industry, business function, type of ROI (economical, social, political), ethic matters (such as bias).
Just like with taxonomy, there is not a specific formula to define AI use case variables. Every firm has its own culture, its way of doing and measuring business. Therefore their list of variables should be a representation of that.
I believe that by starting with three taxonomy levels and around 10 variables is a secure way to avoid complexities, and an easy way to begin practising the art of use case classification.
Regardless if it is a startup or a massive incumbent, AI illiteracy is still omnipresent. Properly identifying and classifying use cases is one of the many activities that will lead to a higher AI maturity in your firm, therefore ensuring better positioning in the market. For instance, something as simple as helping employees classify AI use cases between machine learning, computer vision, natural language processing or IoT disciplines, would produce a massive difference in the mindset of your employees, as they will be able to identify and properly communicate AI opportunities within their daily tasks. Now imagine having listed out a total of 50 potential AI use cases for your company. Imagine that this list is divided into three taxonomy levels and contains the 2 business variables “problem solved” and “use case benefit”. Publishing this list throughout your company would once more increase the AI maturity among your peers and therefore your competitive advantage.
There can be no doubt that AI has arrived and that it is not only here to stay, but to take over. It is going to be present in every single industry and business unit, affecting every single one of us. A report by Accenture calculated that thanks to AI, business productivity will increase by 38% in 2035. Firms are aware of it and have started a race towards its implementation. In our experience as AI consultants and specialists, the main aspect any company should focus on is to increase their AI literacy, and a fantastic way to do so is by dominating the Art of Use Case Classification. The reason why we believe this is because, at the end of the day, regardless of the science and project, it all starts with the use cases. In AI Shepherds, we are offering the service AI Use Case Identification, a 14 hours workshop where we help our clients to understand the different concepts of AI on one hand, and to identify and classify a first set of AI use cases on the other. Additionally, we have also created the brochure 10 things you must know about AI , which will give you a better understanding of what AI entails in less than 20 minutes.