On a recent Insight Live, co-presenter Amol Ajgaonkar and I discussed the genetic makeup of an all-star data science team and the necessary skill sets needed to bring a data and Artificial Intelligence (AI) initiative to completion. Watch the webinar on demand to learn more.
Building your all-star data science team
IT executives are quickly discovering that without the right team members in place, data and AI initiatives fall flat. As the success of data and AI becomes more widespread, top-performing companies are racing to establish or develop existing data science teams within their organization. A recent poll uncovered that more than 52% of respondents already had a data science team within their organization.
But what exactly does it take to build a winning data science team?
The strength of an all-star data science team lies within the various roles tying the project together seamlessly. These roles, both technical and nontechnical, are what ensure a data science model successfully reaches the finish line.
1: The business analyst
The first checkpoint in the AI lifecycle is to confirm you are solving the right problem. This includes defining the problem clearly, understanding the impact of the problem on the business, what is important to your organization, and how success will be defined. Clearly outlining the outcomes you want to achieve with an AI solution will help you gain internal buy-in to invest and solve your business challenge.
A business analyst has the right expertise to pose these mission-critical questions:
- Are we solving the right problem?
- What is the question that we're trying to answer?
- How accurate must our prediction be to provide business value?
- What are the risks of false positives and false negatives?
2: The data architect and/or engineer
The data engineer is focused on transporting, cleansing, enhancing, and presenting the data in a way that it can be easily digested and used (often by data scientists) to influence decision-making. This role is responsible for the “plumbing” ─ moving things around efficiently and effectively and in a way that if a part of the operations breaks downstream, there are clear flags and warnings to prevent errors from amassing and self-repair procedures are applied over time.
Through the looking glass of data science, this role is a bit different.
In a traditional enterprise data environment, like a data warehouse, teams agree on the metrics to be supported, what data to bring in, what transforms to use, and what schedule to refresh. Because it is centrally planned and owned, it is relatively easy to keep governed. However, in a data science team, data scientists need to be able pull down data, merge it with other data, apply transforms, repeat, back up, and try again. Additionally, the data scientists’ findings need to be cataloged for other data scientists to benefit from.
These different patterns require different engineering that is typically outside of most data engineers’ experience without training.
3: The data scientist
What's a data scientist?
Data science refers to assessing historical data, attempting to extract patterns and predicting the likelihood that identified patterns will apply in the present and the future.
Data scientists take this process a step further by ensuring the data is used to produce new solutions. Data scientists pull cleansed and enhanced data collected from the team’s data architects and engineers and apply it to solve unique use cases. The data is then used to devise innovative solutions that are provided to clients through simple-to-use platforms.
4: The project and program manager
Project and program managers are essential members of your data science all-star team. These members work consistently to ensure your AI models turn into full production. Project managers can apply Agile practices to keep the team going, Scrums as needed, lead daily meetings, and define and manage milestones using probabilistic reasoning. Project and program managers also make sure stakeholders are communicating throughout the project.
5: UX designer and strategist
An all-star data scientist team is incomplete without the expertise of a UX designer and strategist. This role is solely focused on the experience of the end user and poses questions like:
- Who is the end user?
- How will the end user utilize the product?
- What problems can we anticipate if the product was launched today?
- How will this product simplify daily tasks?
The goal of the UX designer and strategist is to ensure the AI solution will simplify the lives of the end users and cultivate champions/product evangelists.
Creating a Center of Excellence
Uniting these roles under an AI Center of Excellence (CoE) is one of the best ways to conjure multiple efforts and get your all-star data science team to value quickly. Assembling a CoE pulls expertise from various departments under the same organization to ensure IT projects can scale and perform as needed.
As my co-presenter Amol Ajgaonkar said in our recent Insight Live session on LinkedIn:
“… a CoE brings all of these teams in around a data science function and now applies it to different problems in the business… so now you come back to, ‘Well, we have these standard practices.’ That is how I think we should define what a standard or best practice is — it’s the best practice that works for you and for that business. It's not across the industry.”
Members of your CoE should understand the value of AI and how their independent roles unite to bring an AI model to completion. These are internal experts with leading-edge expertise in various areas of the business. Members of the AI CoE should receive data science-specific training to help them operate through the lens of data and AI.
What you need to get started
So, how many players do you need to get your data science initiative off the ground? To get started, your team must begin with a data scientist. A single data scientist can make a difference to your organization, but it can take up to 20 weeks to build a good algorithm.
Though, as more requirements are revealed, the need for certain skill sets within your team will become increasingly vital. Very often, especially in the case of building sustainable models, a singular data scientist trying to encompass the entire success lifecycle will have challenges scaling. Over time, this one-person team winds up doing more work on maintenance than on the discovery of new value and new data.
Regardless of industry, establishing a data science team under a CoE takes time and companywide buy-in, selecting the right players from several departments. But done successfully, a data science team has the power to launch a data and AI solution that drastically improves company performance, cuts costs, and boosts innovation at scale.
Our data science experts take a deeper dive into building high-performance data science teams in this webinar — watch it on demand to learn more.
Ready to get started? Learn more about how Insight can help you set up an AI Center of Excellence to propel your data initiatives forward and make real change in your organization.