This is the third, and last, post of my series about AI Talent. In the previous two posts, I discussed how to source the right talent and how to assess the fit between the candidates and your organisation in order to maximise the potential of AI in your organisation.
You never know what you have until you lose it, Anonymous
This post focuses on how to structure the talent you have acquired and how to provide an environment to ensure their retention and enhance their development.
When to Create Data Science Teams
A common question that has been discussed many times is where to put your data scientists in your organisations to optimise their impact. This is a critical point, but I have already written about it before. Instead, I will focus this section on an even more important question, especially for start-ups. Namely, when to create a data science team.
Do you have a clear problem to solve?
Without a clear vision and a defined set of problems to solve from the tactical point of view, the data scientists will not be able to focus their efforts and it is likely that a large part of their work and effort will be lost. As a result, they will become frustrated and will probably leave. This does not mean that companies cannot (or should not) pivot, but that the goal of each pivot should be well understood at any given point by the whole organisation.
Do you have data available already or can you collect it?
Data is a key ingredient of Machine Learning and AI and without it, or at least the possibility of gathering it, the data scientists will be limited on how much they can impact the organisation. As a result, the value of a data science team will be lower than expected, increasing the frustration both within the team (e.g., “If I only had this data point I could build this amazing feature”) and the organisation as a whole (“We have invested heavily on AI and we have seen no impact so far”). If this situation persist over time, your data scientists will leave.
Do you have any infrastructure in place?
If you have a clear problem and relevant data, your data scientists will be able to work. However, depending on your problem and scale, they might not be able to work effectively or to productionise their findings without the right infrastructure in place. This question is more subtle than the previous ones and, in some cases, it will make sense to start the team even before having some infrastructure in place but the risk is to have people working in a suboptimal way where Data Scientists are either not able to work effectively, or they will spend a large part of their day doing engineering tasks. Some people will embrace that challenge, while others will be annoyed. In the best case scenario, you will be utilising the skills in the team ineffectively but supporting personal development for those data scientists who want to gain more engineering skills. In the worst case, you will have a team working very ineffectively that is unhappy about the nature of their day to day work. As usual, reality tends to be somewhere in the middle.
Retaining and Developing your team
It is always a difficult moment when a member of your team leaves. In a startup, or even a scaleup, this could be terrifying, especially if the person is very influential in your organisation. The bad news is that, no matter what you do, even with a perfect company, management and culture, people will leave at some point, and you should be ok with it. They might want to travel the world, or maybe they want a job that is just not available in the company, or they may have simply reached a point where they want to explore new adventures in a different company or industry.
Retaining talent is conceptually very simple, but very difficult to execute. People joined you for a reason, therefore, the main approach to retain them is to make sure the company keeps the same characteristics that attracted them in the first place. This might sound easy but ensuring this as the organisation scales and grows is a challenge. For Signal AI, we have seen that culture, impact while working on interesting problems, academic collaborations and personal development plans have been some of the most impactful retaining factors for people working in Data Science:
Culture: Signal AI has more than double in size over the last couple of years and we have invested a lot of time making sure we have a fantastic office environment and culture where people feel valued and recognised. From the biweekly demos to share our last product improvements and to reiterate our company vision to our browbags to encourage the spread of knowledge between different teams and functions.
Impact and Challenging Problems: Data Scientist are happier when they are challenged working in interesting problems and when they can see the impact of their work. At Signal AI, we process and gain insight from millions of textual documents which poses immense challenges on engineering and research. In addition, the impact of any research improvements are clear on the value provided to the users and their satisfaction. If your data scientists do not feel they are working on interesting and impactful lines of work, there is a risk they will get bored and leave.
Academic Collaborations: Other important aspect for us is to keep our academic collaborations and our reputation and knowledge in the space. Some approaches that have worked well for us include weekly research guilds, our constant publications, conference participation and organisation and our external speakers. All of this keeps changing, growing and adapting as we grow as a company as the expectations and complexities grow but one truth remains: In order to attract and retain people who were (or will be in the future) in academic environments, you need to keep your reputation and publication record in your community.
Personal Development: Opportunities to develop further and a clear career path for your data scientists is also critical for their well-being. Managing expectations and supporting and mentoring are key for them to feel valuable and heard. In some cases, they will want to become “as good as they can” as data scientists while in other cases they would like to manage teams and departments. In more extreme cases, they might want to move to a completely different role and department (e.g., Product Manager). Managing this situation while being transparent and fair with all of them is necessary in order to have a productive and happy team.
This post focused on the questions to be asked before starting scaling your Data Science function or teams and on retention and employee development. My main two takeaways are the following:
- Do not rush to build a data science team until the organisation is ready for it.
- Make sure the company keeps the same characteristics that attracted people to join in the first place in order to retain your talent.
This is the last post of the series and I will also use this section to close the whole series about Talent in AI. I believe that talent is one of the keys to unlock the power of AI in organisations across the globe and I hope the thoughts I shared in this blogpost series have been helpful. The main takeaways from me are, conceptually, very simple:
- Make sure you have a clear problem that AI can support
- Hire for more than just technical expertise
- Make yourself known in the community
- Understand and take care of your Data Scientists
But most importantly, always remember that a company is nothing without its people.