More companies in every industry are adopting artificial intelligence to transform business processes. But the success of their AI initiatives depends on more than just data and technology — it’s also about having the right people on board.
An effective enterprise AI team is a diverse group that encompasses far more than a handful of data scientists and engineers. Successful AI teams also include a range of people who understand the business and the problems it’s trying to solve, says Bradley Shimmin, chief analyst for AI platforms, analytics, and data management at consulting firm Omdia.
“The technologies and the tooling that we have available is skewing more and more toward enabling and empowering domain professionals, the business users, or the analytics professionals to take direct ownership of AI within companies,” he says.
Carlos Anchia, co-founder and CEO of AI startup Plainsight, agrees that AI success rests largely on establishing a well-rounded team with a diverse range of advanced skills, but doing so is challenging.
“Identifying what makes a highly efficient AI team may seem like an easy thing to do, but when you examine the detailed responsibilities of individuals on successful AI teams, you quickly come to the conclusion that building these groups is extremely hard,” he says.
To help you assemble your ideal AI team, here is a look at 10 key roles found in well-run enterprise AI teams today.
Data scientists are the core of any AI team. They process and analyze data, build machine learning (ML) models, and draw conclusions to improve ML models already in production.
A data scientist is a mix of a product analyst and a business analyst with a pinch of machine learning knowledge, says Mark Eltsefon, data scientist at TikTok.
“The main objective is to understand key metrics that have a major impact on business, gather data to analyze the possible bottlenecks, visualize different cohorts of users and metrics, and propose various solutions on how to increase these metrics, including making a prototype of the solution,” says Eltsefon, who adds that, when working on a new feature for TikTok users, it’s impossible to understand whether the feature benefits or alienates users without data science.
“You don’t understand how long you should test your feature and what exactly you should measure,” he says. “For all of this, you have to apply AI methods.”
Data scientists may build the ML models, but its ML engineers who implement them.
“This person is tasked with packing the ML model into a container and deploying to production — usually as a microservice,” says Dattaraj Rao, innovation and R&D architect at technology services company Persistent Systems.
The role requires expert back-end programming and server configuration skills, as well as knowledge of containers and continuous integration and delivery deployment, Rao says. “An ML engineer is also involved with validation of models, A/B testing, and monitoring in production.”
And in a mature ML environment, ML engineers also need to experiment with serving tools that can help find the best performing model in production with minimal trials, he says.
Data engineers build and maintain the systems that make up an organization’s data infrastructure. They are crucial to AI initiatives because data needs to be both collected and made suitable for consumption before anything trustworthy can be done with it, says Erik Gfesser, director and chief architect at Deloitte.
“Data engineers build data pipelines to collect and assemble data for downstream usage, and in a DevOps setting, they build pipelines to implement the infrastructure on which these data pipelines run,” he says.
The data engineer is foundational for both ML and non-ML initiatives, he says. “For example, when implementing data pipelines in one of the public clouds, a data engineer needs first to write the scripts to spin up the necessary cloud services which provide the compute necessary to process ingested data.”
If you’re building a team for the first time, you should understand that data science is an iterative process that requires a lot of data, says Matt Mead, CTO at information technology services company SPR. Assuming you have enough data, “about 80% of the effort will be related to data engineering tasks and approximately 20% will be the actual data science-related work,” he says.
Because of this, only a small percentage of your AI team will work on data science efforts, he says. “The rest of the team will identify the problem being solved, help explain the data, help organize the data, integrate the output into another production system, or present the data in a presentation-ready manner.”
A data steward oversees the management of a company’s data and makes certain it is accessible and of high quality. This important role makes sure data is used consistently across an organization and that a company complies with changing data laws.
Data stewards ensure data scientists get the right data and that everything is repeatable and clearly marked in a data catalog, says Ken Seier, national practice lead for data and AI at technology company Insight.
A person in this role needs a combination of data science and communications skills to collaborate across various teams and work with data scientists and engineers to ensure stakeholders and business users can get access to data.
A data steward also enforces an organization’s policies around data usage and security. “The data steward is making sure that only people who are supposed to get access to secure data get that access,” says Seier.
The domain expert has in-depth knowledge of a particular industry or subject area. This person is an authority in their domain, can judge the quality of available data, and can communicate with the intended business users of an AI project to make sure it has real-world value.
These subject matter experts are essential because the technical experts who develop AI systems rarely have expertise in the actual domain the system is being built to benefit, says Max Babych, CEO of software development company SpdLoad. “Domain experts can provide critical insights that will make an AI system perform its best.”
When Babych’s company developed a computer-vision system to identify moving objects for autopilots as an alternative to LIDAR, they started the project without a domain expert. Although research proved the system worked, what his company didn’t know was that car brands prefer LIDAR over computer vision because of its proven reliability, and there was no chance they would buy a computer vision–based product.
“The key advice I’d like to share is to think about the business model, then attract a domain expert to find out if it is a feasible way to make money in your industry — and only after that try to discuss more technical things,” he says.
Moreover, domain experts can be vital liaisons between customers and the AI team, says Ashish Tulsankar, head of AI for edtech platform iSchoolConnect.
“This person can communicate with the customer, understand their needs, and provide the next set of continuous directions to the AI team,” he says. “And the domain expert can also keep track of whether the AI is implemented ethically.”
An AI designer works with developers to ensure they understand the needs of human users. This role envisions how users will interact with AI and creates prototypes to demonstrate use-cases for new AI capabilities.
An AI designer also ensures that trust is built between human users and an AI system, and that AI learns and improves from user feedback.
“One of the difficulties organizations have in scaling AI is that users don’t understand the solution, disagree with it, or cannot interact with it,” says Shervin Khodabandeh, co-lead for consulting firm BCG’s AI business in North America. “Organizations that are getting value from AI — their secret is actually just that they get the human-AI interaction right.”
BCG thinks about it in terms of a 10-20-70 rule, which is that 10% of the value will be algorithms, 20% is the tech and data platforms, and 70% of the value will come from business integration or tying it to the strategy of the company inside the business processes, he says.
“That human-AI interaction is absolutely key and is a huge part of that 70% challenge,” he says, adding that AI designers will help you get there.
The product manager identifies customer needs and leads the development and marketing of a product while making sure the AI team is making beneficial strategic decisions.
“In an AI team, the product manager is responsible for understanding how AI can be used to solve customer problems and then translating that into a product strategy,” says Dorota Owczarek, product manager at AI development company Nexocode.
Owczarek was recently involved in a project to develop an AI-based product for the pharmaceutical industry that would support the manual reviewing of research papers and documents with natural language processing.
“The project required close collaboration with data scientists, machine learning engineers, and data engineers to develop the models and algorithms needed to power the product,” she says.
As the product manager, Owczarek was responsible for implementing the product roadmap, estimating and controlling budgets, and handling cooperation between the tech, user experience, and business sides of the product.
“In this particular case, as the project was initiated by business stakeholders, it was especially important to have a product manager who could ensure that their needs were met while also keeping an eye on the overall goal of the project,” she says, adding that AI product managers should have both technical skills and business acumen.
“They should be able to work closely with different teams and stakeholders,” she says. “In most cases, the success of an AI project will depend on the collaboration between the business, data science, ML engineering, and design teams.”
AI product managers also need to understand the ethical implications of working with AI, Owczarek adds. “They’re responsible for developing internal processes and guidelines that ensure the company’s products adhere to industry best practices.”
The AI strategist needs to understand how a company works at the corporate level and coordinates with the executive team and external stakeholders to ensure the company has the right infrastructure and talent in place to produce a successful outcome for its AI initiatives.
To succeed, an AI strategist must have a deep understanding of their business domain and the basics of machine learning; they must also know how AI can be used to solve business problems, says Dan Diasio, global AI leader at EY Consulting.
“Technology was the hard part years ago, but it’s now reimagining how we wire our business to take the best advantage of that AI capability or AI asset that we create,” he says, adding that an AI strategist can help a company think transformationally about how it uses AI.
“To change the way [a company makes] decisions requires somebody with a significant amount of influence and vision to be able to drive that forward,” Diasio says.
AI strategists can also help organizations obtain the data they need to fuel AI effectively.
“The data that companies have inside their systems today or inside their data warehouses really only represents a fraction of what they will need to differentiate themselves when it comes to building AI capabilities,” he says. “A part of the strategist’s role is to look out into the horizon and see how more data can be captured and utilized without overstepping privacy considerations.”
Chief AI officer
The chief AI officer is the lead decision-maker for all AI initiatives and is responsible for communicating AI’s potential business value to stakeholders and clients.
“The decision-maker is someone who understands the business, business opportunities, and risks,” says iSchoolConnect’s Tulsankar.
The chief AI officer should know the use cases AI can solve, where there’s the most significant financial benefit, and they should be able to articulate those opportunities to stakeholders, he says.
“They should also chalk out how these opportunities need to be achieved iteratively,” he says. “If there are multiple clients or multiple products across which the AI needs to be applied, the chief AI officer can break down client-agnostic and client-specific parts of the implementation.”
The executive sponsor is a C-suite manager who takes an active role in ensuring AI projects come to fruition and is responsible for obtaining funding for a company’s AI initiatives.
Executive leadership has a significant role in helping drive the success of AI programs, says EY Consulting’s Diasio. “The biggest opportunities for companies often are areas where they break across particular functions,” he says.
A consumer products manufacturer, for example, has a team that’s responsible for R&D, a team responsible for the supply chain, a sales team, and a marketing team, he says. “The biggest and best opportunities to apply AI to help transform business cut across all four of these functions,” he says. “And it takes strong leadership from the CEO or C-suite of a company to go after those changes.”
Unfortunately, senior management in many companies aren’t adequately versed in the potential of AI, says BCG’s Shervin Khodabandeh.
“Their understanding of it is quite limited, and they often think of it as a black box,” he says. “They throw it to the data scientist, but they don’t really understand the new ways of working with AI that are required.”
Adopting AI is a big cultural change for many companies who don’t understand how a high-functioning AI team works, how the roles work, and how they can be empowered, he says. “For 99% of the traditional companies adopting AI, it’s a hard thing.”