Teamwork makes the (AI) dream work
Author: Klaus Puchner (Program Manager AI & Team Lead)
Presentation of the team structure and role of AI Programme Manager
We get asked quite often what’s driving us as AI team and how to imagine the daily work with artificial intelligence in our team. Another frequently asked question is how the AI team is organised in order to master the challenges brought by the use of AI.
This blog post is part of a mini blog series through which we would like to answer all of these questions to give those interested a better insight. Each post is meant to highlight different aspects. Among other things, you can read from members of the AI team what their experiences and tasks in the team are. In this post, we will present our motivation, our ‘why’, but also how we are organised as a team.
What drives the AI team at XXXLdigital?
In everything we do, we are strongly convinced that we are enabling a better experience as well as simplifications for our customers and our colleagues in all areas of the company through our machine learning (ML) based solutions. The AI team makes this possible by developing ML solutions in the form of high-performing microservices and by designing them to be easily integrated into existing systems and processes via API.
Who is the AI team at XXXLdigital?
The AI team is divided into two specialised product teams (Computer Vision and Predictive AI). A product team consists of a Product Manager, AI DevOps Engineers as well as Data Scientists. Each team is responsible for the entire life cycle of the solutions developed by them. This includes all steps from conception to planning to operations and maintenance.
Organizational chart showing the structure of our AI team.
We develop all our AI solutions ourselves. There is no need for us to use ready-made solutions from external suppliers. This allows for greater flexibility and security and for creating valuable knowledge within our company. Currently, our team members are based in Wels and Vienna, with more locations planned.
What are the tasks as AI Programme Manager?
In my role as Programme Manager and Team Lead, it is my duty to create a work environment that is liveable, lovable and can be experienced both on a cultural and organisational level. In the cultural context, it is very important to me to not only support an appreciative cooperation as well as mutual learning and a culture where mistakes are dealt with constructively, but also to act as a role model. I do this for example by being available to my team members as needed to discuss or challenge their ideas.
Structured processes are also an important requirement for a successful team which can only be lived and reproduced through providing the right organisational framework. For this purpose, I am constantly in touch with the Product Managers and AI DevOps Engineers to evaluate existing processes (project management and CI/CD processes) and improve them, if necessary.
Another one of my exciting tasks is to identify use cases in which ML solutions can create added value. This, however, does not happen behind closed doors, but during continuous exchange with colleagues from different departments as well as the AI team. This task also involves planning and controlling the AI project and product portfolio in collaboration with the product teams.
What is the general project workflow in the AI team?
Each product team works in an agile setting. This means that every product team covers project operations, development, deployment, maintenance and improvement of its AI products. During each of these tasks, the AI Product Manager, the Data Scientist and the AI DevOps Engineer work closely together. Within the product team, it is not unusual to drive several AI projects at the same time.
The MLOps (Machine learning operations) life cycle, source: https://blogs.nvidia.com/blog/2020/09/03/what-is-mlops
AI projects are handled according to our tailor-made hybrid project management method. This method is based on our insights and experience that we were able to gain throughout the entire life cycle of already developed AI products. Even though we mostly deploy agile methods (10-week product increments, 14-day sprints, reviews, retros, dailies, kanban), we use classical approaches for the planning of product increments (project sheet with goals, non-goals, stakeholders, scope, non-scope, etc.).
Here, it is important to capture the essentials while keeping the balance for the level of detail in the documentation. This combination enables us to define the planned goal for the end of the product increment in as much detail as possible while working as agile as possible towards the defined goal during the sprints. It also allows us to quickly react to changes or unexpected developments in the project.
What are our biggest challenges in AI projects?
During AI projects, we are usually confronted with two challenges. For example, the term AI is not standardised, which is why there is no single understanding of what AI actually is. Film and television play their part in this, making it difficult to some extent to live up to the expectations for AI. Here, the AI team has chosen the approach of demonstrating to stakeholders with working prototypes what AI can do (and what it can’t do). Through the direct exchange, this approach also allows the AI team to better understand the stakeholders’ needs and requirements regarding a potential ML solution. This, in turn, has a positive impact on the quality of the delivered solutions.
SCRUM plan format typically used for our AI projects.
The other big challenge is dealing with the quality of the existing data. Although there is a large amount of data available for exciting projects, the data quality often is not high enough to be immediately used for AI projects. In this regard, the AI team is in intensive dialogue with the Data Owners in order to point out data quality issues and to develop proposals for a sustainable solution. Not only the AI team, but all data-driven departments and teams within the company benefit from this feedback.
Curious?
We still have many aspirations. That’s why we are looking for people to support us with their personalities and skills as well as their courage to learn new things and their motivation to help shape the future with us. We look forward to getting to know you during a personal interview.
More interesting articles
You also want to read the other parts of this AI mini-series? Here’s the article overview for you:
* Die deutsche Version findest du hier.