Companies--an estimated 80% of them--are investing in artificial intelligence in some way, but those investments are made more difficult by a competitive staffing environment. That environment means significant salaries for those who do have these in-demand skills, and a focus on training up existing staffers at big tech companies.
Part of that competitiveness might come from the narrow definition of how “AI talent” is being defined. “The previous ‘unicorn’ definition of the data scientist is beginning to decompose,” said Alex Fly, CEO of Quickpath.
A recent Harvard Business Review article argued that firms are ignoring a variety of key AI roles, to the detriment of their ability to implement and scale up artificial intelligence technologies. Engineers, data czars, translators and business leaders are all important parts of an organization’s AI staffing cohort, HBR argued, but are being overlooked.
Important AI Staffers to Consider
Certain skills are particularly in demand as the technological needs of organizations change. A recent report from Microsoft and IDC found that digital skills, analytical abilities and continuous learning capabilities will be the three highest in-demand skills for business leaders in Asia Pacific, for example.
More than one kind of employee can demonstrate these skills, and hiring or training for them can broaden a team’s ability to meaningfully bring AI into an enterprise’s operations. For example, Fly said, on the business skills side, analytic translators are beginning to serve as AI product managers, with their combination of a solid business background and an understanding of the practical capabilities of artificial intelligence.
“This allows them to act as a liaison between the business stakeholders and the data science team effectively," Fly said, adding that these business managers can look at AI use cases and decide if they represent strategic investments, quick wins or innovations worth an enterprise’s time.
Things are changing quickly in data science, which is a challenge--but not an insurmountable one--on the IT side, Fly said. “The evolution of data engineers from batch ETL data processing to more API and streaming-based applications has helped with the transition to supporting AI-enabled applications,” he said. With machine learning engineers also on board, they can provide a greater understanding of analytic products from the data science side.
It’s also important to pay attention on the infrastructure side, said Ken Zamkow, general manager of Run:AI. Having the right compute infrastructure is needed for scalable AI, he said, including newer hardware like GPUs or new ASICs, as well as updated software and workflows.
“It’s important for organizations to appoint IT leaders and personnel to focus on this developing infrastructure world,” Zamkow said, “in order to support the new workflows and tools needed for launching AI applications at scale, as well as for efficiently supporting the ML and deep learning teams that develop them.”
Think “Dream Team”
Harvard Business Review missed some AI roles that are important to consider, said Marc Vanlerberghe, CEO of Rulai, a SaaS company working in AI-based virtual assistants. Business leaders, mentioned by HBR, are needed, he said--specifically those who understand how artificial intelligence is going to be used, what the impact is on business processes that will be automated, and what the KPIs are that will get optimized.
“Without a clear understanding of these, most AI projects will fail,” Vanlerberghe said.
However, also critical are conversational designers, particularly in the virtual assistant space, Vanlerberghe said. “This is a person who understands best practices in how to create natural conversations between a human and a bot,” he said--such as how to guide the human and how to handle bot confusions Many companies lack experience in this area and will rely on web or app designers, he added, but designing a website is a different job than designing a bot.
Rulai offers courses to fill these gaps, Vanlerberghe said, teaching companies about best practices in conversational design and deployment methodology. The demand is there, he said, and the first course--developed with professors from Carnegie Mellon, University of Washington, and University of California--sold out quickly.
Fly recommends going with a “dream team” approach, something he’s seen work in creating AI factory models for Quickpath’s clients.
“Blended product teams of business, data science and IT,” he said, “are able to align priorities, funding, and the necessary skill sets to execute and scale AI initiatives more successfully.”
