By: Sid Bhatia, Regional Vice President, Middle East & Turkey, Dataiku
If there is one thing all regional technologists agree on, it is the presence of a shortfall in high-end talent. Digital skills in general are in short supply, but the ratio of available roles to qualified candidates for expert positions is particularly high. Across the region, governments, businesses, and non-profits are crying out for AI specialists and data scientists. The future depends upon leveraging AI to solve problems that humans, on their own, cannot.
Objective 5 of the United Arab Emirate’s National Strategy for Artificial Intelligence 2031 focuses on the attraction and training of AI talent. The country’s AI ministry is intent on delivering AI training courses to government employees and the working public, as well as upskilling students by selecting one third of the nation’s STEM graduates each year — which is around 2,000 people — for special training.
The UAE has long been a proponent of leveraging AI to do just about anything that will upgrade society and governance. Its AI ministry was the world’s first, and the government’s laser focus on using smart technologies for everything from smart grids to public health will require a bedrock of talent in the years to come.
The road ahead
Acquiring the skills to implement all the AI projects in the pipeline currently means engaging the services of either a third-party software company or a consulting firm, but rarely do either of these improve the talent pool of a company in the long term. Once these options are discarded, recruiting companies are left with hiring the best data scientists available, but the talent acquired may not be of the desired quality.
Today, especially given the support provided for professional upskilling by the government, UAE businesses are best served by developing in-house skillsets and letting them explore AI in tandem with adding value to the company. So-called “academy companies” like this include some international giants such as PepsiCo, Goldman Sachs, JPMorgan Chase, General Electric, Amazon, and Netflix. The talent factories built by these organizations have three elements in common. First, they acquired an easy-to-use AI technology platform and made it available to everyone. Second, they provided self-service upskilling programs to suit every level. And third, they built a platform capable of identifying nascent talent.
Successful implementations have three layers in common.
- Model factory
Tools must integrate seamlessly into the technology stack and cover the entire AI product lifecycle, from data collation and cleaning to model development and governance. The talent factory should ensure that teams comingle and learn from one another, and that the platform caters to all types of skills. This is important because, typically, model factories serve a workforce that is 5% data scientists, 10% software engineers, 20% data engineers, and 65% other analysts.
Apart from making tools accessible to all skill levels and easy to pick up by those that do not do AI full time, organizations should ensure AI solutions can demonstrate productivity gains over previous technology. Some modeling factory productivity gains reach 300%, with the right technology.
- Adoption program
An AI program can quickly result in negative ROI if procured tools are not widely used. Would-be AI talent factories must initiate adoption programs to encourage use of the tools and build an AI culture. These programs should include development workshops, AI maturity assessments, hub-and-spoke organizational design, business value assessments, onboarding, training, hackathons, badges, laptop stickers, and a peer-to-peer learning community.
- Upskilling program
It is only when a wide enough sample of employees are using AI tools that upskilling can begin. Business analysts can cross train to become data engineers. Data engineers can segue into data science. And data scientists can elevate their delivery to video, images, audio, natural language, and deep learning solutions.
But as this upskilling accelerates, governance must not be forgotten. IT leaders must be able to monitor and manage data access, compute costs, projects, model releases, teams, and analysts. Not only will this allow the controlling of costs, but it will permit stakeholders to single out analytics adepts for further development.
Resource allocation
Much of the disappointment in AI implementation comes from the supposition that once a homogenized data lake and a modeling factory are built, that value should pour forth naturally. But AI culture is not built on technology alone. In fact, analytics budgets should arguably be split evenly between technology and adoption and upskilling. This is the very nature of the AI talent factory — to recognize that technology does not run itself. It needs to be used often and thoroughly; and those that do so must be skilled enough to add value in many different ways while continually topping up their knowledge.
Remember, it may be counterproductive to trawl through regional labor markets for the fully formed, relatively expensive data scientist who may or may not possess the ability to add value. A decade ago, employers blindly did just that — acquired AI specialists and taught them business and industry knowledge. Results were varied. But now that AI best practices are better understood, the most efficient route to ROI is to build an AI talent factory. Business and domain talent is already there, so why not use it? And with the right strategy on tool procurement, adoption and upskilling, companies can drastically boost productivity and become an AI academy. After that, the sky’s the limit.
