Gone are the days when artificial intelligence (AI) was merely a “someday” fantasy of science fiction.
Every day, developments in AI, machine learning and even robo-advisors appear in mainstream media headlines. Although most companies are just beginning their AI journey, the momentum in this field is growing rapidly.
While there is broad interest in its potential, achieving commercial results from AI in a socially responsible way is easier said than done. Most companies are still thinking through the implications of AI; some executives are still grappling to understand its relevance to their organizations, while others are considering how to manage its inherent risks. Among these is the potential impact on the world’s workforce. For example, the 2016 MIT Technology Review revealed that 70% of Asian executives expect AI to result in substantial job losses in the next five years. Many are wondering if our society is moving fast enough to address that impact and the needs of an AI-enabled workforce.
There are other risks that have not been as widely publicized as potential job losses. For example, there is potential for systematic bias in decision-automation algorithms, either by accident or design. This can result in adverse social consequences, such as denying a segment of the population access to insurance.
To navigate a successful path through the benefits, opportunities and risks of AI, organizations will need to map out a thoughtful strategy that aligns technology with purpose. Then, they’ll want to anticipate and prepare for the inevitable challenges in executing AI-specific programs. Here are three steps along the journey to consider:
1. Align AI with purpose
When imagining how AI might change their organizations, many business leaders become too narrowly focused on the tactical impact. In doing so, they risk missing the big picture. AI is fundamentally changing the role of people at work and thereby the very reason why organizations exist. When the AI focus is purely tactical, there is no impetus for the whole organization to respond to both the opportunities and the risks, across the value chain.
Instead, to derive the full value of AI, organizations will need to think more broadly about how, where and when to deploy AI technologies and processes to deliver meaningful top- and bottom-line results. To do that, they will need to align their strategy and their technologies with the company’s overall purpose. Defining and developing a coherent AI strategy aligned with purpose will be essential to guide successful transformation.
2. Manage input to generate good output
Enterprise technology programs are not easy to execute in the best of circumstances. Implementing AI solutions is even more difficult, thanks largely to three key differences. First, AI solutions are (mostly) trained, not coded. That means that a significant amount of development work that would have traditionally fallen to an IT team now must shift to subject-matter experts who may have limited capacity to spare. Second, AI solutions consume unstructured data, such as documents and rich media. Managing the “ground truth” in real time becomes critical, while complying with data privacy regulations may become harder because sensitive data may be embedded in free-text documents. Third, the pace of AI innovation is unprecedented, and software vendors are under continuous pressure to stay ahead of the curve. Each of these characteristics will shape the way organizations will select and deploy AI solutions.
Meanwhile, in our work with a range of EY clients, we’re seeing significant differences in the capabilities of AI software and their compatibility with established data and analytic platforms. Understanding these characteristics, and identifying which technologies and processes best suit the needs of the organization are at the root of developing successful AI solutions.
Finally, there is no point in creating a great solution if no one uses it. In this new territory of AI, it is not uncommon for companies to work with providers to develop hundreds of predictive algorithms yet implement only a handful of them. Companies will need to build strategy and capacity for operationalizing AI processes and apply a rigorous approach to steering AI transformation programs to successful impact.