Don’t just hire PhDs.
AI is almost everywhere in the news today, and the drive to create and implement AI solutions is creating an enormous talent gap. An estimated 80% of companies are already investing in AI and most are facing challenges hiring the capabilities they need to implement a useful AI application or product. It’s clear that there is an intensively competitive market for artificial intelligence and machine learning specialists. Many companies first attempt to hire Ph.D.-level data scientists with expertise in AI algorithms and “feature engineering.” Some analysts have even equated “AI talent” with such researchers.
However, AI talent goes far beyond machine learning Ph.D’s. Equally important and less understood are the set of talent issues emerging around AI product development and engineering. Most firms have not filled these roles, and their AI projects are suffering as a result.
The AI Engineer Role
Because some others have already realized their importance, let’s focus first on engineering skills. A very useful articlerecently pointed out the difference between machine learning researchers and machine learning engineers. The key takeaway is that most companies need engineers to help develop products and production applications, rather than a researcher to help push the boundaries of AI technique and technology.
These engineering skills include creating technology architectures that scale, writing and deploying bulletproof software, and integrating AI capabilities with existing systems. The people in AI engineering roles need to know something about AI, but just as much about programming, computing, and corporate IT environments. Such skills are becoming increasingly important over time as AI knowledge and tools mature, and as algorithms and techniques become commoditized.
The AI Data Czar Role
AI initiatives also need data experts. We’ve also argued elsewhere that the machine learning race is increasingly a data race in which unique data, rather than cutting-edge modeling, is what creates a valuable AI solution. Unfortunately, sourcing and managing data is a skill set that does not often overlap with algorithm development. The AI data czar is typically a position that is created over time through experience, rather than hired out of school, although education in computer science or statistics can be very helpful. The role encompasses such capabilities as:
- Knowing what data sources are useful to address an AI question or problem;
- Being aware of how data is used in algorithms;
- Assessing data quality;
- Cleaning and treating data;
- Having a focus on detail (and being a stickler for data quality);
- Possessing the strength to push back at technical teams;
- Knowing the typical ways to transform data.
Data management also requires business knowledge. Let’s discuss a simple example. Our startup uses machine learning to bring automation to strategy consulting, and one of the key inputs we use is the financial information included in annual reports. This data is inherently full of gaps. Not every company reports the same set of metrics, and the reason for failing to report is most often that there is nothing to report in the category. For example, many companies do not bother reporting their research and development spending because they have none! This means that the best course of action for filling most of these gaps (which must be filled for the algorithms to work their magic) is to fill them with zeros, representing no spending.
However, in the world of data science and machine learning, zero-filling data is extremely uncommon and filling with the median value is a generally accepted best practice. Our application is the rare case where median filling actually introduces errors into the data set—for example by assigning an average amount of research and development investment to every company, when 70% of the market actually spends nothing on R&D. If we had handed off data management to the tech team, as many companies do, we would have headed down the wrong path. Instead, by having an informed business team deeply involved in the AI development process, we are able to catch potential problems.
The Business Leader and AI Translator Roles
AI groups also need a role at the intersection of business strategy and AI methods. Such a person, usually a somewhat senior executive, is able to translate strategic objectives and business models into the types of AI that can advance them. Unfortunately, the role of a business leader with some understanding of AI techniques is rarely discussed, and even less often filled.
The result is that AI is often used to create either off target or sunk cost projects where the technology investment does not yield the ROI anticipated by the board or the leadership team. Our experience working with boards and leaders is that creating a solid AI product that provides either customer, employee, operational or investor value is about 40% problem and product definition, 40% data sourcing, cleaning, filling, and merging, and only 20% algorithm development.
Solving these problems requires ongoing partnership between business and technology. Yet most companies do not have a clear point of view about how AI can help organizations make better, more informed and faster decisions, or smarter products and services. Automating parts of decision-making and product development requires a person that can work at the intersection of strategy, business models, code development, algorithm creation, and product development—a rare breed.
Having someone in this role even pays dividends when it comes to algorithm development. From rules-based systems to logistic regression to neural networks and beyond, the algorithms that are used in AI each have different characteristics, good and bad. Although we wouldn’t expect all business leaders to know these, we would expect a good AI business leader/translator to engage with developers on these pros and cons to help drive the right decision. For example, neutral networks, though powerful, lack explainability. It is difficult to say exactly why the model returns the results that it does. For many products, a “black box” solution won’t do—the users want or need to understand how it works.
Many companies rush into the AI race without clear objectives, hope a brilliant AI researcher and a technology team can create something great without guidance, and end up with little to show for it. Recruiting an AI quarterback to provide the business input, and ensuring success with well-defined metrics, is the most important job that most companies miss entirely. Some have argued for the importance of the translator function for traditional business analytics, but given the complexity of AI it is even more important with that set of technologies. Indeed, many large AI groups will need multiple people to play the translator role.
The businessperson who fills this role does not need to become a programmer, know the best AI tools from vendors, or delve into the nuances of neural networks versus logistic regression. He or she does, however, need to understand the basics of how different types of AI work and the data sets that will be deployed with them., Such individuals should also have a desire to get deeply involved and work iteratively with the AI team rather than throwing requirements “over the wall,” leaving the machine learning team with the tough decisions. In addition, they need to create a clear economic use case and product road map that produces value for customers, employees, partners or investors. In most cases, these individuals should lead the AI group, and the researchers, engineers, and data czar should report to them.
Having someone on board that who is in or reports directly to the C-suite with an understanding of these topics, and who can oversee the other important AI roles we have discussed, will help the organization achieve its core objective—value for stakeholders—while avoiding the costly, unproductive cycles we often see in poorly managed AI development.