The 5 Things Your AI Unit Needs to Do29 January, 2018 / Articles
Not a day goes by without the announcement of the appointment of a new VP of Artificial Intelligence (AI), a Chief Data Scientist, or a Director of AI Research. While the enthusiasm is undeniable, the reality is that AI remains an early-stage technology application. The potential is vast, but how managers cut through the AI hyperbole to use its power to deliver growth?
In our consulting work, we often encounter managers who struggle to convert AI experiments into strategic programs which can then be implemented. Michael Stern (not his real name), for instance, is the Head of Digital for a German Mittelstand office equipment company. Michael is used to starting new projects in emerging areas, but feels unable to fully understand what can AI can do for his business. He started a few experiments using IBM Watson, and these produced some clear, small tactical gains. Now Michael is stuck on how to proceed further. How can he create cross-functional teams where data experts work with product teams? And how will they pick project ideas that produce real ROI? Michael wonders if his firm even knows what new business models can be explored with their existing datasets — let alone which new ones might be made possible by AI.
Michael is not alone. As more and more companies invest in AI-driven units, many newly appointed managers face these challenges – especially in companies with little or no previous experience with cognitive technologies. Part of the trouble: in many companies, the role of these teams is undefined. Very little research has been done to design the mission and scope of these new units.
At the European Center for Strategic Innovation (ECSI), we examined numerous corporate AI initiatives among large organizations, and identified five key roles that can help AI units to develop the right mission and scope of work to succeed.
- Scouting AI technology, applications, and partners. This role is about setting up a core team of “AI sensors” in charge of monitoring new trends, identifying disruptive technologies, and networking with innovative players — mainly startups. The automobile-parts supplier Bosch and the tech and engineering powerhouse Siemens are two prime examples of this. With a planned investment of $300 million, Bosch has established three AI corporate centers focused on IoT and other AI-related fields in Germany, India, and Palo Alto. Siemens, similarly, has included AI in the company’s list of innovation fields to be monitored through its network of innovation outposts with offices in California, China, and Germany.
- Experimenting with AI technology and applications. This role is about understanding through quick, small AI pilots how to develop or adopt cognitive technologies to the company’s business and operational models. Although off-the-shelf AI tools and open-sourced systems are available, they have limited transformational potential compared to customized ones. At Deutsche Telekom, the development of its own AI solutions is an important priority. Instead of buying AI chatbots from vendors, Deutsche Telekom has its own developer teams. With the support of partners, they design, train, and fine-tune AI solutions for the company.
Rather than concentrating efforts on a single big win, AI units and teams should embrace a portfolio approach to their experiments. The power of AI should be tested across functions and business areas. There are three types of experiments that are worth paying particular attention to:
Experiments in the driver’s seat are typically conducted by the company’s AI unit or internal developer teams. In the last few years, Deutsche Telekom has tested internally three different AI-backed chatbots and virtual assistants to improve the company’s corporate and private customer services.
Experiments with others in the driver’s seat involve joining forces with innovative players such as start-ups, research centers, and universities. In general, such experiments are focused on cutting-edge technologies or applications requiring in-depth expertise and skills that companies do not have. This is a common strategy among large organizations: Mercedes-Benz entered a partnership with the Computer Science and AI Lab of MIT; Associated Press collaborated with Automated Insight, a specialized AI firm; Deutsche Telekom partnered with the German Research Centre for AI, called DFKI.
Experiments by learning from others are common among companies interested in pioneering AI technology and applications, but too premature for their industry. Observing others translates into funding ventures or start-ups innovating at the frontier of AI. This is the case at German insurance company Allianz, which funded Europe’s first global AI equity fund to position itself as a “pioneer in AI investments.”
- Supporting business units in applying AI technology. This role is about building internal capabilities through a specialized network of AI experts who can support business units in the integration and application of AI tools and solutions (from basic data visualization and chatbots to the automation of entire processes like claims management). The success of AI applications lies not in the technology per se, but in the ability of a company to align it with its business and operational models.
The Data and AI Lab is one of the most visible BNP Paribas’ AI efforts. The Lab is responsible for the development of AI tools that can improve the internal processes. At BNP Paribas, the AI team is in charge of accompanying and supporting business units all along the way, from the identification of potential applications to the experimentation and fine-tuning. It’s essential that these labs be tightly integrated into the organization, not in a far-off lab. Constance Chalchat, Head of Change Management at BNP Paribas says, “Data scientist teams need to work in close partnership with both the business and IT.”
- Getting the entire organization to understand AI. This role is about the ability of the AI team to educate the organization on the opportunity to harness the power of AI. Why? Because AI is ultimately a tool. Organizations need to build solid foundations that enable people to actually use and secure value from AI technology. As passion for AI is rising to the top of large organizations, this applies also to the C-suite and Board. Executives need support to cut through the complexity of AI-driven discussions and find ways to extract value.
Embedding AI in the company’s culture and core skills set can be done at two levels. First, internal communication initiatives can help raising awareness and acceptance of AI technologies, in particular those with a high transformative potential, while creating a common AI language and culture. Second, targeted education efforts allow building basic, standard capabilities of people, who are not AI experts in the organization. AirBnB is a prime example of this. By setting up an internal Data University, AirBnB is teaching employees data science with the goal of making the transition to a more AI-aware organization easier and faster.
- Attracting and retaining talent. This role is about addressing the AI skills gap. A dedicated AI unit should work in close cooperation with the HR department to identify the right skills and capabilities required, and define strategies for talent retention. Companies are currently adopting different AI talent acquisition strategies. Edouard d’Archimbaud, Head of the BNP Paribas Data and AI Lab is gradually expanding his 25-member team. “We’re recruiting around ten people a year […] we’re very careful and only like to hire the right people,” he explained. Other companies have invested more significantly. This is the case of Airbnb that recently “acqui-hired” a team of seven data engineers from Changecoin, a start-up with deep knowledge of blockchain technology.
The framework in action
Sometimes these newly created AI teams will be investing time and effort in all the five roles. The challenges at other companies can be quite different. Plotting the five roles on spider graphs like the one shown here can help companies figure out where they are currently focused and where they may need to increase or reduce their efforts. They can, for example, compare what they are currently doing with what they should be doing, given their company’s strategic intent and their capability and organizational issues.