The success of your artificial intelligence initiatives may depend as much upon art and philosophy as it does upon data science and machine learning. That’s because the effective deployment of AI in the enterprise will require building well-rounded teams that include people from a wide range of backgrounds and skill sets, including non-technical roles.
“Any AI initiative requires a marriage between an IT expert and a domain expert,” says Moshe Kranc, CTO of Ness Digital Engineering. “The IT expert understands the machine learning toolkit: Which algorithm families are most likely to solve a particular problem? How [do we] tune a specific algorithm to improve the accuracy of the results? The domain expert brings domain-specific knowledge: Which data sources are available, how dirty is the data, what is the quality of the ML algorithm’s recommendations? An IT expert cannot possibly answer these questions without input from a domain expert.”
The early takeaway here: AI success indeed relies upon a team rather than any one individual or role.
”When building an effective AI team, we can either look for a superhuman or a super team,” says Keith Collins, EVP and CIO at SAS. “Teamwork wins the day. A diversity of disciplines is the key to success in AI.”
Four core types of AI talent
Collins sees four core disciplines as necessary for AI teams:
- Someone who understands the business processes critical to establishing real-world scenarios and valuable outcomes.
- Someone who understands analytics – machine learning, statistics, forecasting and optimization – that leads to using the right techniques.
- Someone who understands data: Where does it come from? What is the quality? How should it be handled to preserve security and trust?
- The final ”someone” is an AI architect who knows how to put analytics in action by operationalizing for outcomes.
Collins, like other IT leaders and AI experts, notes that these core disciplines or roles can draw from a variety of backgrounds. He points to music, chemistry, and physics as examples.
“Those disciplines encourage understanding the scientific process and thinking in terms of complex, interactive systems,” Collins says. “They’re often good at the critical thinking skills required to establish good experiments and the outcomes of applied machine learning.”
The value of a diverse AI team
The value of a diverse team ranges far and wide: It can help your organization better combat against AI bias, for instance. It’s also important to solving business problems – including your largest and toughest problems – which is presumably one reason you’re developing an AI strategy in the first place.
“It is generally accepted that diversity of opinions is critical to all complex problem-solving,” says Jeff McGehee, senior data scientist and IoT practice lead at Very. “Diversity is all about life experience, and professional background is a large part of most individuals’ life experience, which can add dimension to AI projects and provide new perspectives to finding innovative solutions.”
McGehee also points out that building diverse teams – AI or otherwise – requires an active effort on the part of your company as part of recruiting and hiring practices. Sitting back and assuming diversity will find you is not a viable team-building strategy.
[ Related read: Why diverse IT teams have a competitive edge. ]
With that in mind, let’s look at a range of experts and roles – including non-technical roles – that can be valuable to an AI team.
1. Domain experts
You could also think of these as subject matter experts. Regardless of the term you use, it bears mentioning again their importance to your AI initiatives.
“Developing an AI system requires a deep understanding of the domain within which the system will operate,” McGehee says. “Experts in developing AI systems will rarely be experts in the actual domain of the system. Domain experts can provide critical insights that will make an AI system perform its best.”
Kranc from Ness notes that answering the question of which domains is necessarily specific to your organization and strategy.
“The type of domain expert needed depends on the problem to be solved,” Kranc says. “Whether the desired insights are in the area of revenue generation, operational efficiency, or supply chain management, a domain expert is required to answer questions like [these]:”
- What insights would be most valuable?
- Can the data collected about the domain be trusted to be the basis for insights?
- Do the derived insights make sense?
We’ll break out some specific domain examples later, but first, let’s look at some other key roles on an AI team.
2. Data scientists
This is the first of three key needs for AI teams working on any greenfield project, according to Dave Costenaro, head of artificial intelligence R&D at Jane.ai. Example projects include a chat agent, a computer vision system, or a prediction engine.
“Data scientists come from all sorts of backgrounds, like statistics, engineering, computer science, psychology, philosophy, music, etc. – usually with a strain of curiosity that compels them to dig into systems to find and use patterns,” Costenaro says. “They provide the ‘what’ for an AI project, determining what it can do and training it to do that.”
3. Data engineers
These are the programmers who apply the “how” for an AI project, Costenaro says. “They take the ideas and models and algorithms from the data scientists and bring them to life by formalizing code, making it run on servers, and talk successfully to the appropriate users, devices, APIs, etc.”
4. Product designers
The final of Costenaro’s three key needs also speaks to the value of non-technical expertise on AI teams.
“Product designers also come from all sorts of backgrounds, like art, design, engineering, management, psychology, philosophy,” Costenaro says. “They provide the ‘why’ and plot out the roadmap for what is desired and useful.”
5. AI ethicists and sociologists
This may be an especially critical role in certain sectors – think healthcare or government – but it seems likely to grow in importance across a broad scope of use cases.
“A big part of an AI system is understanding how it impacts people, and whether or not under-represented groups are treated fairly,” McGehee says. “If a system has unprecedented accuracy but does not produce the desired social impacts, it is doomed to fail.”
McGehee also sees a separate but related need for legal expertise in this burgeoning field. “The GDPR has set a precedent for laws around algorithmic decision-making,” McGehee says. “As governments become more aware of how AI is being used in industries, it is reasonable to expect that more laws will emerge. A lawyer who is well-versed in this area could be a valuable asset.”
Since domain experts are so crucial, as Kranc and McGehee spell out, it’s worth looking at some tangible examples of domains – including technical and non-technical – that should be part of your AI team building, depending upon your particular goals and use cases.
“Since AI is often simply an enabling layer that supercharges an existing business use case,” notes Costenaro of Jane.ai, “the full stack of teammates that have already supported that use case in the past are still valuable and essential for the same reasons.”
Costenaro offers five examples of roles that may be valuable AI contributors – and explains how each may need to adapt and augment their existing roles in an AI context.
7. Executives and strategists
Executive leadership will “need to consider which portions of the business model can be automated and improved with AI [and] weigh new opportunities and risks surfaced from other teams below, like data privacy, human-machine interaction, etc.,” Costenaro says.
8. IT leaders
Don’t let the value of non-technical roles confuse you: Your organization’s AI strategy isn’t going very far without IT. Costenaro notes that IT teams will need to solve problems like these: “If large data stores are being accumulated for model training, how do they ensure the privacy and security of that data? Also, how do they store it and serve it quickly and reliably from servers to customers’ devices?”
Costenaro adds that this will also fuel the already growing demand for DevOps pros and people with expertise in cloud-native technologies like containers and orchestration. He also sees an opportunity for IT to use AI tools such as chatbots to streamline internal service.
[ Also read: Kubernetes jobs hunt: How to land that role. ]
9. Human resources leaders
“Similar to IT, a lot of opportunities exist for HR to become more efficient by employing AI tools like chatbots to serve their internal clients,” Costenaro says.
Moreover, HR appears likely to become an important player in evaluating the impacts of AI within an organization, not unlike McGehee’s inclusion of roles like ethicists and lawyers.
10. Marketing and sales leaders
As Kranc pointed out, if your AI initiative is tied to, say, revenue generation, you should most certainly consider adding domain expertise from areas like sales and marketing.
Costenaro also notes that sales and marketing professionals may need to augment their existing skills and processes with technologies like sales automation tools and robotic process automation (RPA) as part of an AI project.
11. Operations pros
Within the overall IT department, operations and DevOps pros have a particular domain expertise to bring to bear on AI initiatives. Costenaro cites the following questions as examples of where operations expertise will be needed:
- What can be automated and improved?
- If we are using machine learning models, how do we create new data collection processes to continually train and improve those models?
- Are there off-the-shelf, pre-trained models and/or datasets we can grab from open source repositories to get a giant head start? Will API services offered by third-party vendors do some of the tasks and use cases in mind?
While AI can solve some significant problems, it’s also virtually certain to create new challenges. This is fundamentally why the makeup of your team matters.
“People with different backgrounds and personalities tend to focus on different project details and constraints,” McGehee says. “This is useful because it raises the likelihood that all important details will be addressed, and provides a holistic approach to identifying solutions.”
[ Want lessons learned from CIOs applying AI? Get the new HBR Analytic Services report, An Executive’s Guide to Real-World AI. ]