5 Actions to Achieve AI at Scale

Artificial intelligence (AI) has entered the mainstream and executives expect great things (84% believe AI will give them a competitive advantage), but most companies have yet to effectively exploit its full potential.

This is owing, in large part, to the "AI paradox": it is deceptively easy to launch AI pilots but extremely difficult to achieve scale.

It's an intriguing problem. Companies achieve excellent results with their pilots and naturally want to scale them up. But as soon as they try to broaden the impact across the organization, things get messy. This is because AI algorithms learn by ingesting data. When AI systems interact with other digital systems, they change the systems they come into contact with. For pilots confined to a small unit within the company, the messiness is reasonably manageable. But when AI systems interact with dozens of units throughout the company, myriad entanglements arise.

These entanglements can be tricky, but they can also be overcome. Companies need to plan for and implement an AI at scale transformation, which can be managed through five steps.

Build a solid AI architecture. Whether you purchase an AI platform from a vendor or build on your own, you're going to need a robust architecture to manage, document, and monitor workflow. The architecture will not only manage issues of entanglement, but it will ensure the security and integrity of AI systems.

Select the right vendors. As your organization gets up to speed in AI, you'll need to partner with vendors. It's the only way to quickly access the AI talent and capabilities necessary to scale up. But selecting vendors isn't a simple decision. AI algorithms thrive on data—sometimes highly sensitive or proprietary data—which means companies must choose carefully. Consider two key questions when selecting a vendor: How valuable is the vendor's process or offering to your future success? How strong is your ownership, control, or access to high-quality, unique data, relative to the AI vendor’s? These questions can shed light on the ways vendors are poised to strengthen or weaken your competitive advantage.

Prepare for a human-machine world. The introduction of AI will affect your workforce in several ways. First and foremost, you will need to hire people with specialized skills, such as AI data scientists and systems engineers. But it's also important to anticipate the ways AI systems will impact the existing workforce. Low-impact AI systems serve to augment employees in their existing roles; therefore, change is minimal. High-impact AI systems, on the other hand, can learn with little or no human intervention and function independently; therefore, change will be significant. As you move along this continuum, you'll need to upskill and reskill the workforce to keep pace with change—and managers will need to adjust to a combined human and machine intelligence.

Design appropriate governance structures. When designing for AI, three essential elements should be centralized: data management, expertise (by creating excellence hubs, for example), and governance. But the AI systems themselves should be managed by cross-functional teams, including AI experts and topic specialists, located within business units. Of course, the employees who interact with the AI systems on a day-to-day basis will be decentralized, carrying out their work at the ground level, but they should be supported by change specialists who can offer training and resolve issues on the spot.

Manage the transformation process. In our work with clients, we have seen excellent results when leaders take the following steps to manage an AI transformation. First, use surveys and interviews to conduct a maturity assessment. Take inventory of all existing AI initiatives and review the current operating model. Prioritize AI initiatives and, if necessary, implement changes to the operating model. Develop a transformation roadmap and create a program management office to oversee the transformation. Implement the AI at scale program by detailing the work streams, responsibilities, targets, milestones, and resources.

***

BCG recently conducted a survey to assess the current state of AI adoption. In China, 31% of survey respondents say they work in organizations that already use AI, followed by North America (26% in Canada, 24% in the US) and then Europe (20% in the UK; 18% in Spain, 16% in France, and 15% in Germany). The vast majority of employees say they are excited about AI and expect it to have positive implications for their organization and themselves. But there's still much work to be done. For AI to fulfill its promise, companies need to overcome the AI paradox and implement AI at scale. Only then will we tap into its truly transformative potential. To learn more about scaling up AI, here is an excellent report from my colleagues Philipp Gerbert, Sukand Ramachandran, Jan-Hinnerk Mohr, and Michael Spira.