Once a niche research specialty, AI is fast becoming a vital aspect of the IT strategy at many businesses. The maturity of data science and machine learning tools, as well as the rise of readily accessible machine learning platforms in the cloud, are fueling this trend, enabling businesses to explore new ways to extract high business value from existing and accumulating data.
But to make the most of AI in the enterprise, you have to have a strong team of AI practitioners in place. Here, we take a look at how four organisations are taking on the challenge of putting together a top-flight AI team to tackle new projects.
What successful AI looks like
When first building an AI department, it’s important to know that successful AI requires multiple roles with differing skillsets.
“At first, we attempted to recruit for a single role – a data scientist – who had the all of the capabilities we needed. That approach did not work out,” says Chris Brazdziunas, vice president of products at LogRhythm, a security intelligence company.
“In our experience, we found that an AI group needs at least three distinct roles: a data engineer to organise the data, a data scientist to investigate the data and a software engineer to implement applications.”
Over the past 12 months, LogRhythm has hired five people into its AI group and plans to grow the group to 10 people over the next year.
For professional services firm EY, AI roles break down along three lines.
“In our approach to AI, we currently see three parts: generating information, interpreting information and making judgment about that information,” says Martin Fiore, EY Americas tax talent leader. Thus far, for EY, “the generating information capability is currently the strongest,” Fiore adds.
Thomson Reuters’ AI efforts have evolved to the point where the interpretation piece is in production. Developed in cooperation with Reuters journalists, Reuters News Tracer was built using AI.
“This application consumes information from Twitter and filters out news from the noise. It can distinguish between a rumor and a fact with approximately 70 percent accuracy,” explains Khalid Al-Kofahi, head of the cognitive computing center at Thomson Reuters.
Here, Thomson Reuters has already translated AI skills into a product in support of its business.
Recruiting and retaining AI talent
AI professionals are in high demand. To assemble -- and maintain -- an AI team, retention and recruitment are key. But that doesn’t necessarily mean having to look outside the organisation.
Developing AI talent internally is part of EY’s approach.
“In EY’s tax group, we provide extensive training on technical tax matters. However, we are also starting to add training on automation and AI. While recruiting a graduate with degrees in tax and AI is excellent, there is a significant talent shortage. That is one of the reasons we put resources in upskilling our people,” says Fiore. In the past year, EY has hired over 20 professionals focused on automation and AI.
Recruiting AI talent in a hot hiring market often requires going directly to academic institutions.
“Being active in the community – especially presenting at conferences and publishing papers – has supported our recruiting efforts. We have also presented at Columbia, MIT, and other leading organisations,” explains Thomson Reuters’ Al-Kofahi.
But bringing a new hire in the door is not enough. You also need them to stick around.
“When it comes to retention of AI professionals, a few factors make an important difference. First, it has helped us to think through the career path and show what other opportunities are available. Second, we encourage our staff to actively participate in the professional community, including presenting at conferences, attending Meetups and other activities,” LogRhythm’s Brazdziunas says.
Professional development is not the only way to grow: Autonomy is also important.
“In addition to external activities, we support giving our AI talent time at the office to carry out their investigations and generate new ideas,” Brazdziunas adds. For professionals interested in research, such support will go a long way.
How to organize your AI efforts
In the technology industry, Bell Labs and Xerox’s PARC loom large as examples of corporate support for research. In addition to broad research units such as Google X, several organisations are creating AI-focused organisations. In Canada, Royal Bank has invested heavily in AI by establishing an AI research lab in Edmonton. Pure research is not the only way to structure your AI organisation. Consider the approach that Thomson Reuters has followed.
“Our organisation has three major groups: traditional research, application development, and user experience,” explains Al-Kofahi. “My philosophy for the group combines two themes: following the business and leading the business. That means that we deliver incremental improvements for the business and create entirely new ideas and products.”
When it comes to moving AI concepts and products from the world of computer science to business use, paying attention to user experience is key.
“In my view, AI is the new UI,” explains Elliott Yama, assistant vice president of machine learning at software provider Apttus. “Max, our AI assistant, is designed to be used conversationally and ask to follow up questions,” Yama adds.
User experience design is a focus at Thomson Reuters as well.
“If you want to build applications that will change how professionals get their job done, the user experience is a big part of this story,” Al-Kofahi says.
Alternative AI staffing solution: Freelance marketplaces
What if your organisation is just not ready to recruit a computer science Ph.D.? There are other ways to get started in AI.
Anand Kulkarni, founder and CEO of Crowdbotics, has hired three machine learning specialists from Upwork, a large talent marketplace. Crowdbotics is far from alone. According to Upwork’s Q1 2017 Skills Index report, demand for AI skills is the second fastest growing skill set. With multiple opportunities available, contractors may be a good addition.
Using freelance talent, Crowdbotics’ AI efforts have been able to improve the accuracy of project estimates for its clients. “If you are building something like a database-driven web application, we have the capability to improve development. This includes suggesting using specific libraries in development. In some cases, we can cut total development time in half,” says Kulkarni.
But Kulkarni offers caution to organisations starting out with AI. In Crowdbotics’ experience, improved estimation and automation only succeeds in cases where the AI has meaningful historical information to reference. So, if you are developing something completely new, consider whether AI is truly right for you.
Here, tapping contractors may be the best first step in finding out.