metamorworks - stock.adobe.com
While nearly every company has adopted digital transformation buzzwords in recent years, the actual implementation of these disruptive technologies remains another story. For a variety of reasons, new technologies often fail to meet expectations. One example is artificial intelligence, which is an advanced technology that companies cannot ignore in an era of increasing digitalization and remote work.
Many organizations deploying AI projects through tactical initiatives that seek to deliver an immediate payback have learned that this approach rarely delivers meaningful value. AI is a powerful tool, one that has the capacity to redefine entire industries. But delivering on that potential is hard and will not be achieved through piecemeal projects.
Organizations that limit their AI investments to incremental and opportunistic deployments without a broader strategic vision risk losing their competitive advantage to those that define a new vision for their enterprise and pursue that with purpose and drive. That's why visionary leaders are the key to unlocking the latent power of AI.
Taking the next step
A common first step for an organization embarking on its AI journey is to establish an AI Center of Excellence (CoE). The CoE harvests business cases from across an enterprise and identifies willing partners in the business. Typically, these AI projects are conceived and designed at the lower levels of an organization and then pursued by talented (but junior) AI engineers.
All too often, there is poor alignment between what a business wants and what its technical team can deliver. Well-designed projects close this gap iteratively as both the business and the AI team learn what is possible to achieve and what is truly needed to deliver value. Unfortunately, it's more likely that neither side has the experience, sophistication or sponsorship to bridge this gap. As a result, both sides depart frustrated.
Even projects that deliver value for a given business often run aground when it comes time to scale. Scaling requires collaboration from multiple cross-functional teams: IT to provide the infrastructure, risk management, HR to provide training and senior business stakeholders to provide sign-off. Unfortunately, one-off projects with limited value struggle to obtain bandwidth and priority from these teams.
Given these challenges, it's no surprise that many AI projects fail to deliver on their intended benefits. However, AI engineers at the ground level aren't to blame. The problem starts with a lack of vision from senior leadership.
The effort to elevate the strategic focus of AI starts by engaging people who own big business problems. Then find goals that are sufficiently compelling for everyone to support the investment and pain of enacting them. This sets teams up for success and gives them the flexibility to adjust their goals as necessary. To achieve its potential, AI needs leaders with a clear vision that will help align stakeholders, justify their investment of time and money and motivate their teams to bring the persistence and collaboration required to make their projects a success.
Adapt to achieve meaningful outcomes
Deploying AI can be an incredibly messy and creative endeavor. Traditional, top-down organizations often have difficulty managing these types of projects, especially when it comes to technology that is as new and transformative as AI.
While IT projects rarely follow the plan, this is even more true for AI projects, for numerous reasons. It may be too technically challenging, data may not be as readily available as originally thought or it may create unforeseen ethical risks.
To succeed, AI demands creativity and leaders who won't let perfection get in the way of good. One of the major challenges in deploying AI is that it's very hard to know what type of data you will uncover. That's why it is crucial that leaders develop an agile approach that empowers teams to adapt to new information as they move forward.
This means moving away from managers leading teams, to experts leading experts. In this scenario, experts provide the guidance and overall framing so teams can make tradeoffs as they progress. A visionary leader will provide an overarching goal and clarity of outcomes -- giving teams the freedom to pivot from intractable objectives.
Let's look at an example from the banking industry. A team trying to classify transactions to determine the appropriate sales tax treatment might find that attaining 99% accuracy would require excessive time and expenses. On the other hand, an 80% accuracy rate would easily be within reach. This is when a team leader who understands the business context might realize that 80% is good enough if the remaining 20% can be handled by people and still deliver significant cost savings.
Alternatively, if full automation is required for the larger strategy, then the extra investment may be justified. A leader who prioritizes the vision and overall goals of the organization would be able to make that decision and implement a more effective strategy.
Inspire from the onset
When an organization pursues piecemeal approaches to AI implementation, trade-offs for each project are made independently and investments cannot be used across the organization. Without sufficient business context, individual project teams will not understand what "good enough" looks like or make appropriate tradeoffs.
Setting a compelling vision and following through with an agile approach is the best way to move forward and integrate AI more effectively into business strategy and transformation. This demands collaborative teamwork across all levels of an organization, not a typical top-down approach. This may take more time, but the payoff can be transformative.
About the author
Nigel Duffy is a technologist and entrepreneur serving as EY global AI leader. He is responsible for projects driving strategic transformation of how EY operates, competes and provides services. Previously, he was a founder and executive in deep technology startups that use AI to run hedge funds, design pharmaceuticals, control computer games and improve online retail. Duffy has built research organizations and revolutionary products in multiple fields and is a highly cited author with papers in machine learning, computer science, linguistics and other areas. His original research includes the earliest theoretical papers on gradient boosting. He received a master’s degree in mathematics from University College Dublin and a Ph.D. in machine learning from the University of California, Santa Cruz.