Building an AI team can be an expensive chore. The talent market is hot with most companies recruiting, with only so many valuable candidates out there. Let’s assume that you already have a data science team on its journey into machine learning and artificial intelligence (ML and AI). How can you get the best ROI? Are they creating value already? Unfortunately, quite often the reality leaves much to be desired. Here are a couple of thoughts I’ve gathered at Silo.AI and at our clients on how to set up a capable and efficient AI team.
DISCLAIMER: In many organisations, you can substitute AI team with data science team or ML team, if they are working on ML and AI. Much of the following applies to other aspects of data science as well.
Jaakko Vainio is the Head of AI Solutions, UK and manages the natural language processing team at Silo.AI. He has years of experience in consulting in AI and data science in various roles as well as a lifetime interest in empowering teams.
Mindset and skills in a successful AI team
The key to a successful AI team is to take into account both the nature of the work and the business drivers in the creation of the team dynamics. Naturally, there are many different ways to organise successful teams but there are certain things that should be found in most of them.
There are many interesting articles and posts on team setups, but in this post, I would like to concentrate on one part of the team composition, which is not so often addressed. A successful team setup should have access to certain key roles.
I have seen in various organisations that the lack of these roles can lead to a multitude of problems. These problems would warrant a post of its own, but most of it boils down to the absence of direction. In different companies and organisations, these people might go by different names, but at Silo.AI, we call these roles AI Mentor and AI Solutions Strategist. These do not need to be job titles, as the persons can include for example AI Scientists, AI Engineers or AI Team leaders in the teams. The important thing is that certain mindset and skill sets are there.
AI Mentor is a person with a wide understanding and experience in machine learning. He or she has been around long enough to develop an intuition into what might work and what doesn’t. This knowledge is readily shared and used to guide more junior AI Scientists to be more efficient and get them past any methodology hurdles. With this knowledge, the team also has an understanding of what technologies are the most promising.
One key aspect of the AI Mentor is that this is a person who can keep up with the cutting edge research and is able to read the latest scientific papers. This is crucial in any rapidly evolving technological field. To keep the hands-on touch to applications an AI Mentor should not be aloof but also getting hands dirty in a project or two. However, in order to get the most out of the time, the AI mentor could be advising on several other projects. The background of this person might be at the academia or he or she could have long experience in the industry, more probably in academia.
Caveats: Intuition can be a tricky thing, it can also be wrong. Therefore, an AI Mentor must not have too big an ego, and be ready to accept that he or she is not an AI god with a capital G. As knowledge sharing is a key task in any team, steer clear of the people who got stuck in the academia and forgot their people skills.
AI Solutions Strategist
AI Solutions Strategist is a person who sees the bigger picture and understands how AI can fit into the surrounding business. Despite all the hype, AI models themselves are very rarely valuable. Rather, the how and where they are used matters.
The AI Solutions Strategist helps bridge the gap between research-oriented AI projects and working business solutions. An important part of achieving this is design thinking in understanding the current business processes that will be affected by AI. Equally important is to find out the best way to make the solution help the users, instead of making their lives more complicated.
Caveats: As this is a combined business and technology role, having a good balance or at the very least understanding the bias, is key. Otherwise, projects and solutions can end up being too modest or way too ambitious. AI Solutions Strategists can be difficult to find as they need both technical and business (preferably industry specific) understanding.
Assuring your AI team’s success
Both these roles can support several AI teams, but as a rule of thumb, one should aim at 3 or fewer teams per AI Mentor and AI Solutions Strategist. Otherwise, the time will become too fragmented and focus will suffer. Of course, this depends on how wide a range of topics and business segments these teams cover. One extra benefit of these roles is that they tend to attract the right sort of people around them. E.g. a good AI Mentor helps attract younger AI Scientists that have the passion to learn and to improve. For this reason it is also possible (while not the only way), to start creating teams this way.
How to get started
AI Mentors and Solutions Strategists can be grown, but this tends to take some time and nurturing. As an interim solutions it is also possible to acquire the expertise of these roles from the outside. We at Silo.AI often work together and as part of our clients’ teams as part of successful project but also to provide knowledge transfer and guidance in AI both in technical and business terms. This is one efficient way to cater for the need for AI Mentor and Solutions Strategist.