Most companies struggle to capture the enormous potential of their data. Typically, they launch massive programs that try to meet the needs of every data end user or have individual application--development teams set up customized data pipelines that can't easily be repurposed. Firms instead need to figure out how to craft data strategies that deliver value in the near term and at the same time lay the foundations for future data use. Successful companies do this by treating data like a commercial product. When a business develops a product, it tries to maximize sales by addressing the needs of as many kinds of customers as possible with it--often by creating a standard offering that can be tailored for different users. A data product works similarly. It delivers a high-quality, easy-to-use set of data that people across an organization can apply to various business challenges. It might, say, provide 360-degree views of customers, of employees, or of a channel. Because they have many applications, data products can generate impressive returns. The customer data product at one large bank, for instance, has nearly 60 use cases, and those applications generate $60 million in incremental revenue and eliminate $40 million in losses annually.

Most companies are struggling to realize artificial intelligence's potential to completely transform the way they do business. The problem is, they typically apply AI in a long list of discrete uses, an approach that doesn't produce consequential change. Yet trying to overhaul the whole organization with AI all at once is simply too complicated to be practical. What's the solution? Using AI to reimagine one entire core business process, journey, or function end to end, say three McKinsey consultants. That allows each AI effort to build off the previous one by, say, reusing data or enhancing capabilities for a common set of stakeholders. An airline, for example, focused on its cargo function, and a telecom provider on its process for managing customer value. Scaling up AI involves four steps: (1) Identify an area where AI will make a big difference reasonably quickly and there are multiple interconnected activities and opportunities to share technology. (2) Staff the team with the right people and remove the obstacles to their success. (3) Reimagine business as usual, working back from a key goal and then exploring in detail how to achieve it. (4) Support new AI-based processes with organizational changes, such as interdisciplinary collaboration and agile mindsets.