What it really takes to scale artificial intelligence

https://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/digital-blog/what-it-really-takes-to-scale-artificial-intelligence

Changing company culture is the key—and often the biggest challenge—to scaling artificial intelligence across your organization.

 

 

June 18, 2019It’s an exciting time for leaders. Artificial intelligence (AI) capabilities are on the precipice of revolutionizing the way we work, reshaping businesses, industries, economies, the labor force, and our everyday lives. We estimate AI-powered applications will add $13 trillion in value to the global economy in the coming decade, and leaders are energizing their agendas and investing handsomely in AI to capitalize on the opportunity—to the tune of $26 billion to $39 billion in 2016 alone.

Meanwhile, AI enablers such as data generation, storage capacity, computer processing power, and modeling techniques are all on exponential upswings and becoming increasingly affordable and accessible via the cloud.

Conditions seem ripe for companies to succeed with AI. Yet, the reality is that many organizations’ efforts are falling short, with a majority of companies only piloting AI or using it in a single business process—and thus gaining only incremental benefits.
Why the disappointing results?
Many organizations aren’t spending the necessary (and significant) time and resources on the cultural and organizational changes required to bring AI to a level of scale capable of delivering meaningful value—where every pilot enjoys widespread end-user adoption and pilots across the organization are produced in a consistent, fast, and repeatable manner. Without addressing these changes up front, efforts to scale AI can quickly derail.

Making the shift

To scale up AI, companies must make three shifts. First, they must transition from siloed work to interdisciplinary collaboration, where business, operational, and analytics experts work side by side, bringing a diversity of perspectives to ensure initiatives address broad organizational priorities and to surface user needs and necessary operational changes early on.
Second, they must switch from experience-based, leader-driven decision making to data-driven decision making, where employees augment their judgment and intuition with algorithms’ recommendations to arrive at better answers than either humans or machines could reach on their own.
Finally, they must move from rigid and risk averse to agile, experimental, and adaptable, embracing the test-and-learn mentality that’s critical for creating a minimum viable product in weeks rather than months.
Such fundamental shifts don’t come easily. In our recent article, “Building the AI-powered organization,” published in Harvard Business Review, we discuss in depth how leaders can prepare, motivate, and equip their workforce to make a change. Here we summarize the four key areas in which leaders should focus their efforts.

Set up for success

To get employees on board and smooth the way for successful AI launches, leaders should devote early attention to several tasks, including the following:

• Explaining why AI is important and how workers will fit into a new AI-oriented culture.
• Anticipating and addressing from the start their firm’s unique barriers to change.
• Budgeting as much for AI integration and adoption as for technology (if not more). One of our surveys revealed that 90 percent of the companies that engaged in critical scaling practices spent more than half of their analytics budgets on activities that drove adoption, such as workflow redesign, communication, and training.
• Balancing feasibility, time investment, and value to pursue a portfolio of AI initiatives with different time horizons (typically over three years) and combining complementary efforts with different timelines for maximum value.

Organize for scale

In our experience, AI-enabled companies have two things in common when it comes to structuring roles and responsibilities—both in terms of who “owns” the work and how the work is executed.
First, they divide key roles between a central analytics “hub” (typically led by a chief analytics officer or chief data officer) and “spokes” (business units, functions, or geographies). A few tasks—such as data governance, managing AI systems and standards, and establishing AI recruiting and training strategies—are always best owned by the hub. And a handful of responsibilities, including end-user training, workflow redesign, and impact tracking, are almost always best owned by the spokes. The rest of the work—which includes, among other responsibilities, setting the direction for AI projects; building, designing, and testing the tools; and managing the change—falls in a gray area and is assigned to either the hub or spokes based on each firm’s AI maturity, business-model complexity, and pace of innovation. (Generally speaking, the greater the AI maturity and more data experts available, the more these responsibilities can be shifted to the spokes, while higher complexity and a need to innovate rapidly may shift these responsibilities to the hub).
Second, when it comes to execution, they put in place a governing coalition of business, IT, and analytics leaders that shares accountability for AI initiatives and sets up interdisciplinary teams within the spokes—drawing from talent in both the hub and spokes to build, deploy, and monitor new AI capabilities.

Educate everyone

To ensure the adoption of AI, companies need to educate everyone, from the top leaders down. To this end, some companies are launching internal “analytics academies,” which provide leaders a foundational understanding of AI, enable analytics experts to continue sharpening their hard and soft skills, build translator expertise to bridge technical and business requirements, and prepare both frontline workers and strategic decision makers, such as marketers, to use new AI tools in their daily work.

Reinforce the change

With most AI transformations taking 18 to 36 months to complete (and some lasting up to five years), leaders must also take steps to keep the momentum for AI going. Following are some of the best ways we’ve found to do this:

• Role modeling. For example, leaders can (and should) attend analytics academies as well as actively encourage new agile ways of working and appropriate risk taking by highlighting what was learned from pilots.
• Making the businesses accountable. A scorecard that captures project-performance metrics for all stakeholders, for example, is an excellent way to align the goals of analytics and business teams.
• Tracking adoption so teams can correct course as needed.
• Providing incentives for change, such as shining a spotlight on employees who have helped make the company’s AI program a success.

All this work (from the initial setup activities to the reinforcement mechanisms) not only helps organizations get more value from AI in the near term but also creates a virtuous cycle: the growth of interdisciplinary teams, test-and-learn approaches, and data-driven decision making that comes with the building and adoption of new AI capabilities leads to more collaborative practices among employees, flatter organizations, and greater agility. This provides fertile ground for even greater innovation, enabling companies to thrive as AI advancements barrel full speed ahead.
For a deeper look at how leaders can drive the cultural and organizational changes necessary for scaling AI, read “Building the AI-powered organization,” on hbr.org.
Tim Fountaine is a partner in McKinsey’s Sydney office and leads QuantumBlack, a McKinsey company, in Australia; Brian McCarthy is a partner in the Atlanta office and coleads the knowledge development agenda for McKinsey Analytics; and Tamim Saleh is a senior partner in the London office and heads McKinsey Analytics in Europe.