How to Build an AI Strategy

A recent article by Harvard Business Review opens with “every serious technology company has now an Artificial Intelligence (AI) team in place.” The author raises an important point when claiming “Please don’t hire a Chief Artificial Intelligence Officer” by highlighting: “For them, AI is a competitive advantage, but not part of their core product.”
This brings up an important question: How do companies focus on machine learning and artificial intelligence? There are three broad types of companies that put AI at the core of their value chain:

  • R&D companies: Similar to pure academic research, these companies aim to further the field of artificial intelligence, often without a specific product or application in mind. Examples are non-profit organizations, think tanks or investor backed R&D companies.
  • AI Champion: These companies put AI at the core of their value creation. While they do invest heavily in research and development, their research efforts are focused on developing a product that fully exploits the capabilities of AI for a specific application.
  • Digital Enterprise: These companies use AI as a competitive advantage and improve their own value chain but the development of AI itself is not part of their core product. For example, a retailer may use AI-based products to automatically set the best price for each product each day in each store or automate their logistics and replenishment network. AI-based solutions are used by the real business to solve a real need.

How can Digital Enterprises benefit from AI?

How to build an AI strategy

Digital enterprises can build up in-house competences and grow a team that focuses on Data Science and Artificial Intelligence. Ideally, indeed, the team would be supervised by a Chief Artificial Intelligence Officer (CAiO) who pushes digital innovation from the perspective of the senior management. At first glance, this strategy looks very appealing: All competences are in-house, external dependencies are minimized and all team members have access to all company data (hopefully), which allows them to integrate their solutions with the existing IT infrastructure. However, on closer inspection, this becomes less compelling. As development of AI is not the core of a Digital Enterprise, there is, at least initially, little expertise, meaning that the first few hires are critical. Without thoughtful and unbiased mentoring, ideally by external experts, risks are high to hire the wrong CAIO: They might, for example, be an extremely talented AI researcher – but with little exposure to the business needs or experience in handling the challenges the particular company or vertical faces.
On the other hand, chances of finding someone who has a deep understanding of a specific vertical or business as well as being well-versed in AI technology and its applications are slim at best. The same applies to the initial Data Science/AI team: Hiring many recent computer science graduates may be possible, but that won’t push the company forward. These hires have little work experience and are not familiar with a specific business vertical – or know what to focus on in a business environment to start with. Hiring AI specialists is difficult: Why should they come work for you when they could join technology companies like Google or Facebook with planetary resources and a large, established AI community? What is more, growing a Data Science and AI team takes time: Not only do they need to start working as a team, trust each other and get to know each other’s strengths, they also need to establish themselves as part of the company and build fruitful relations to all other – already existing – teams.
Pressure will be high: Unless establishing an AI team is part of a long-term effort of transforming the company to put data first, the business likely has some pressing business needs that are to be addressed. Bringing in a new AI team isn’t a silver bullet to magically fix all problems – and it also invites a “told you so” response where this effort is seen as chasing the latest hype rather than building a sustainable future.

What should a Digital Enterprise do to include AI into their core business processes?

The first hurdle is to realize that integrating data-driven solutions into the core of the business is driven by a business need. Senior executives should not go looking for the next thing and just stick AI in front of it but rather investigate which existing or new business cases can benefit from the potential AI brings.
A “data first” strategy will likely come with a large amount of homework for the business, regardless of which specific application is tackled. The organizational structure of the business will have to change to reflect the way data-driven decisions are formed, propagated and executed within the company. In particular, operational decisions will either be supported or automated by AI systems with little to no human intervention. For most companies, this is a significant shift in the way the business is done day-to-day: Conventionally, human operators and experts are responsible for a large number of operational decisions that have to be made every day. For example, a retailer has to determine the optimal order quantity of each individual product for each store, taking constraints from the supply chain network, as well as a myriad of influencing factors into account. The sheer number of such decisions alone indicates that those decisions are rarely optimal as the human experts simply don’t have the time to deliberate each decision and consider all the factors that make the best decision into account. An AI-based replenishment system, however, could automate up to 99 percent of all order decisions, freeing the human experts to think about the remaining 1 percent of operational decisions that still require human intervention or focus on other tasks, such as improving customer service.
So how does AI enter the picture and how are the decisions made? This is where the second type of companies, the AI Champions, come in: These companies develop specialist AI solutions or products that address a particular business need such as automating the store replenishment or setting the optimal price for all products for a retailer. As they specialize in specific sectors, they have extensive expertise in AI research, domain specific knowledge, as well as experience in turning R&D efforts into working products that create value for the customers. By commercializing a product or service, they also know which ideas or use-cases will work out in a commercial setting, bring value to customers, are realistic to tackle on a specific time-frame – or are more likely an interesting research question that won’t help the customer forward. The fact that these AI Champions are external to the customers is actually an advantage: Each company can focus on what they do best and further their core competencies. This leaves the data exchange defined by a well-designed API that details which data are needed (and which are not) as well as where and how data are exchanged between both partners.

Why should all the Data Scientists and AI experts join the AI Champions? All the other cool kids do. More seriously, these companies focus on driving AI further by a strong R&D effort – and keep in synch with the “real world” by applying their developments to real-life use-cases and help real customers.

Dr. Ulrich Kerzel Dr. Ulrich Kerzel

earned his PhD under Professor Dr Feindt at the US Fermi National Laboratory and at that time made a considerable contribution to core technology of NeuroBayes. He continued this work as a Research Fellow at CERN before he came to Blue Yonder as a Principal Data Scientist.