Question 1: How are emerging AI roles and required skills shaping the future of the labor market? Discuss the evolving demands for AI skills and the challenges of upskilling or reskilling the current workforce to meet demands.
The AI development process is different from a traditional software development process. Unlike standard software practice, an AI practice must account for application development and data management including data and model drift over time. This new development process will impact on job roles and responsibilities of the people who will be developing the AI applications. AI will create new technology positions while some existing roles will continue to evolve and specialize in specific AI models.
This paper will identify the key AI development roles. For each role, I will identify the corresponding existing role and discuss the common skill sets between the existing role versus the role for the AI development process. For the new roles, I will identify the critical skills required and discuss the potential reskill of existing personnel in that new role. The bottom line is that most of the AI roles can be addressed by current staff if the company takes the time to build thorough AI job descriptions, conduct skills assessments of their current personnel, identify the gaps, build a training plan to upskill their employees, and train those identified people. There will be cases where a company has holes in their current personnel, and this will cause them to look outside.
We will address each position as they are brought into the AI application development process. We will first start with objective goals, problem characterization, and requirements definition. We will then move to identifying the required data to support the desired outcome of the project. The next phase is the model selection and development followed by software development and the user interface. This process may appear to be a waterfall model, but it is not. These phases happened in parallel with potential lag from each other and there is a feedback loop between phases to account for necessary updates. I am segmenting the process this way to define how the roles impact the entire AI application development process.
Business Analyst: The business analyst is responsible for the problem definition, gathering the requirements, and identifying the success metrics of the project. In many companies this role is often done by the sales engineer. The sales engineer is the customer facing technical point of contact. They are often aligned to specific industry verticals so they also can act as the domain expert. The one area for upskilling the sales engineer is on the AI development process, the different AI models, and the AI business value.
AI Strategist: The AI strategist is a senior solutions architect who previously been engaged in AI projects. They will align the business objectives with an AI vision. They will also assist in defining how the AI project’s success criteria are measured and assessed. This role is new to many companies. Companies will need to take their more senior sales engineer/solution architect personnel and train them on AI models, business problems that AI model can address, and how to measure those outcomes.
Project Manager: Most companies have project managers that they can assign to many forms of engagements. Ideally, the project managers should be familiar with the AI development process and the areas of risk that will need to be mitigated. Consequently, the companies have these resources available today but they will need to be trained on the AI differences and areas of concern with the AI application development process.
Configuration Management Specialist: As soon as a project manager is assigned, a configuration management specialist (CMS) will also be needed. Like the project manager, companies already have this resource in house. They will need to be trained on the different items that will require configuration management. Historically, the CMS has only managed code versions and document artifacts. AI has the additional dimensions of data and AI models. The CMS will need to manage the different data sets (training data, validation data, operational data). The CMS will also need to manage the AI models that may be commercially off the shelf or custom developed for the AI application.
Data Engineer: This role already exists in many companies. The data engineer is responsible for data collection, data cleansing, transforming, and database management. The data engineer will manage three different sets of data for the project. Those datasets are training data, validation data, and operational data. The data engineer will also work very closely with the data scientist.
Data Scientist: Like the data engineer, the data scientist is an established position. The data scientist will determine the best AI model to use for this application. They will then train and validate the AI model using the training data set and validation data set respectively. Data scientists will also perform Sensitivity tests of the AI model by adjusting weights and calculations.
AI Ethics Officer: The AI ethics officer is a new position for most companies and should be engaged early in the development process ideally at the same time as the data scientist. Trust and explainability are critical to any AI application especially when that application is dealing with potentially life impacting or can do significant harm to an individual based on its determinations.
ML Engineer: This is a relatively new position for most companies. Even though machine learning has been around for a period of time, companies have not formalized the ML engineer position. They have used their more senior software engineers to act like ML engineers. The ML engineer will deploy the algorithms into the application. This will include developing any required API for sharing the results with the data visualization team. It is recommended that companies take their more experienced programmers and train them on developing ML algorithms.
AI Cloud Architect: The AI cloud architect is a role similar to the cloud architect that companies may already have on staff. They will define the compute and memory requirements for the AI application using CSP’s resources. The big difference is that the AI cloud architect must know the different AI software platforms that each of the cloud providers already has like AWS’ SageMaker. AI cloud architects need to understand the other cloud platforms like Google Vertex AI, Azure Machine Learning, and IBM Watson Studio platforms.
Data Visualization Engineer: The data visualization engineer (DVE) is a job position that companies already have. This person protypes, reviews, and finalizes the user interface (UI) with the customer. They often work in Power BI or Tableau. DVE works with the ML engineer to utilize the defined APIs to get and visualize the data for the customer.
Software Engineer: This is a job position that many companies have. The software engineer is equivalent to the utility player in baseball. They can address many different areas of application development. Some software engineers can grow into the ML engineer position. In smaller companies the software engineer will also be the data visualization engineer.
In conclusion, the labor transition to an AI powered ecosystem is well underway and does not require a company to hire new personnel to fill these roles. Rather, a company needs to assess their existing talent base and determine which engineers can make the transition to AI. Companies will need to have very good job descriptions for these new roles and a pragmatic training plan to help their current employees make that transition. Where a company has gaps in their talent, they will need to look outside the company to fill those holes.
A condensed version of this article is published on our Blog under the title AI-Powered Personnel Transformation.
