The ultimate AI project team
Managing the three phases towards success
When you look at the reasons why AI projects fail in their implementation, most people tend to think of missing data, missing know-how in the development of machine learning models, wrong strategy or insufficient budget.
In the vast majority of cases, however, the error occurs in advance and lies in the composition of the project team. The three most important lessons learned can be briefly summarized:
1. You always need two types of AI specialists: technical and application experts as well as dedicated translators between the two.
2. A strategic concept should be available for every AI initiative, clearly describing its later application and benefits.
3. AI projects should be managed in three phases: Concept and development, productization, operation — the corresponding resources need to be ensured.
The following article shows you the elements of success including which roles to fill in order to successfully utilize AI even in critical business processes.
Technical machine learning vs. applied machine learning
It is often assumed that only developers or machine learning engineers are true AI experts. They understand the technical details of the models, and can mathematically understand the steps necessary to achieve a desired result. They are also able to use ML frameworks and translate requirements into code. All this is perfectly correct and makes Machine Learning Engineers a rare and coveted species these days. But this alone is not enough to realize successful AI projects.
A simple example: I would like to open up a restaurant. Before I talk to the technical experts such as a kitchen builder, an interior architect, and installers, I should first become clear about the key strategic issues:
- What is my value proposition to the end-customer?
- Who exactly am I targeting?
- How do I communicate the advantage over competing solutions?
- Which location makes sense for this?
- How large should the project be?
- Which monetary gain do I aim for in order to be profitable?
- What team do I need to set up and operate the project?
The project is therefore divided into different phases in which different key competencies are important.
Phase I — Concept & Development
For the strategy development and planning, I need one decision-making authority (decision maker) at the highest reasonable level (trade-off between time-commitment and decision-authority). Secondly, I need a structured innovation process to identify key requirements to the above mentioned questions.
A thorough analysis of my assets and capabilities (data scientists) and coordination with the building owner (IT executive) have shown that the technical equipment and interior design are particularly difficult and important in this case because the building is not actually designed for gastronomy and a smooth process is eminently important for business success.
In this phase, my main focus is on the technical specialists who propose a suitable architecture, select and purchase high-performance equipment (data architects and engineers) and take the complete installation and security precautions (machine learning engineers and developers). As soon as the concept is available, the practical users: chefs and service personnel should also be involved in order to verify that the theoretical concept is consistent (domain experts).
Critical questions should always be answered by the decision maker. The implementation process should be coordinated by a capable project manager who is able to precisely translate the requirements into work steps (project manager and translator).
Phase II — Productization
If the infrastructure is built up in a first resilient form, I need an acceptance test of the modifications (testing). However, many other aspects are required for a successful launch: interior designers (UX / designer), precise financial planning (controlling and pricing), suppliers (supplier and partner management) and, of course, a unique menu (product management).
Phase III — Operation
During operation, everything then focuses on the end-user. In order to optimally design the user experience, I have to make sure that everyone who participates in it is optimally educated (trainer). I need a restaurant manager who is able to optimally implement the concept and the possibilities offered by my infrastructure (team lead and translator). Finally, I must ensure that everything runs smoothly and that internal and external requirements are met (legal and compliance specialist).
There are usually three critical project phases. In the first phase, I need four different roles on the technical side (for small projects these roles can be taken over by less than 4 people) and two business experts (they should have separate roles) plus one decision-maker. This means that a relevant part of my team has to understand exactly what is at stake, but does not necessarily need many years of experience in setting up and developing AI models.
On the other hand, one should also avoid using rare technical experts with tasks that do not correspond to their profile. In order to move from the strategic-technical nucleus to successful application, further functions are needed which need to be far less involved in the topic of AI. The better the very central role of the project manager and translator is performed, the less important is the prior knowledge of the team members in phases II and III.
Also, in the overall project coordination, it is often necessary to fill new roles over the project period. An excellent project manager and translator in phase I does not automatically have to become a good product manager and he or she may not have the competence to lead a large team in phase III. In addition, the initial role of the decision-maker gets less critical when the first steps are made and the way forward becomes clearer. High-level management attention will therefore shift from the project and requires a strong team lead to be put in charge.
We always welcome your feedback and questions.
Feel free to reach out at appanion.com