Prioritizing AI Investments That Create Real Value

In the wake of ChatGPT’s launch three years ago, executives are under pressure from their boards and investors to have a clear AI strategy that will keep them from falling behind peers—and that will actually yield ROI. Creating such a strategy requires organizations to move beyond the bottom-up experimentation phase and the rush to include AI, in any form, in their products. To do that, leaders need to reexamine basic questions about how their firms create value for customers and understand how the constantly evolving technological landscape is changing the answers. So how do you identify which opportunities to invest in? Here are three questions to discuss with your executive team to target your approach. Questions to Ask Where is value created in our industry? And how is that changing? When confronted with a new technology, many established firms attempt to quickly incorporate it into their current processes and offerings. But this misses a step and means that real transformation may pass them by, according to John Winsor, a strategic adviser, serial entrepreneur, and fellow at Harvard Business School’s Digital, Data, and Design Institute. To truly reinvent their business, Winsor argues, leaders need to understand where competitive advantage lies in their industry and how the new technology might change this. Traditional value chain–mapping exercises can help by providing a structured way to visualize how value is created, delivered, and captured in your industry. By revealing dependencies and areas of friction or scarcity, they can point to opportunities for improvement or differentiation. Next, look at how AI is already changing value creation and control across that value chain. Where are your current efforts no longer going to be effective? This process isn’t just about identifying threats. Are there gaps in the evolving value chain that you have the capabilities or assets to address? Reconnect with your mission and purpose. Ask yourselves: What business are we really in? What are we really good at? Is there a broader definition of our purpose that connects with these new opportunities? Take Shutterstock, a stock photography firm. Traditionally, the stock photo industry created value based on a scarcity of high-quality images. But in the age of AI, Shutterstock realized that the abundance of cheap, synthetic images was not only diminishing that source of value, it was leading to a collapse of trust in the value of authorship and copyright, which are critical for any content to be licensed commercially. The company has since shifted its focus to trust and rights management. It’s built on its existing capabilities in those areas, as well as its relationships with photographers and clients, to create a system that makes AI-generated content safe for enterprise use. It has established data-licensing partnerships with OpenAI and Nvidia, compensation models that ensure fair payment for contributors, and indemnification frameworks that protect customers from copyright risk. It is moving from being an aggregator of photos to a governance engine. (Example adapted from “What Executives Get Wrong About AI.”) Too often, businesses only do this kind of strategic work when they face an existential threat. Take a step back and understand how you might fit into your changing industry before it’s too late and you’ll have more abundant possibilities available. Where can AI add value to our customers’ experience? While many companies are focused on automation and efficiency, growth opportunities will eventually come from improved and entirely new customer offerings. But many early market-facing AI offerings from established firms have fizzled because they were launched without consideration of customers’ needs. For example, take the case of Snapchat. The platform originally gained popularity among its young user base with features such as disappearing messages that fostered a sense of privacy, and with it the casual, frequent sharing of authentic moments. In 2023, the company introduced My AI, a gen AI chatbot. The feature was pinned to the top of users’ feeds, where they were used to interacting with their real human contacts, and was perceived to be a kind of unwanted observer. While the company shared that many of its users enjoyed the feature, the flood of negative reviews that cited it told a different story and the app’s U.S. iOS rating plummeted to 1.67 stars. (Example adapted from “Make Sure Your AI Strategy Actually Creates Value.”) To avoid the experience of Snapchat and other firms, don’t let your eagerness to offer an AI feature overtake a disciplined focus on customer need. Revisit what you know about your customer journey and consider how it might be changing—or how it could be improved—with AI in a way that differentiates you from your competitors. Evaluate AI product or feature ideas with these questions front and center, but also use this exercise to generate new ideas that are born of what your customers value and want rather than what the technology can do. Perhaps in response to the first wave of unsuccessful AI-driven product features, companies are already kicking off such projects less frequently than they did earlier in the gen AI era, with more attention being given to natural-language interfaces for existing products and improved product performance, according to a late 2024 client survey by Bain. Where are our biggest opportunities for cross-functional productivity improvement? Process automation is the first lever many companies have pulled to create business value from AI. But because the first phase of AI adoption has been about bottom-up learning, these use cases have developed in isolated pockets in functional silos, limiting their impact. Moving to a more coordinated, cross-functional approach can yield greater gains. To find the areas of the most opportunity, begin by mapping how data and information flow through your organization. With technology moving this quickly, there isn’t time to do this analysis for the whole firm. Instead, begin with your biggest functional investments and capabilities. Understand what tasks take the most effort and where they extend into other functions. These are prime targets for further workflow mapping—and potential automation. Once you have a list of key areas of effort across the business, consider which of those tasks AI is best suited to do today. Take into account the cost of any errors and the type of data needed and whether it is readily available. Ultimately this effort should be about disciplined workflow redesign as much as it is about AI automation. For example, Bain parter Arjun Dutt described how the firm worked with a Fortune 100 hardware company looking to consolidate its AI automation efforts. The hardware company conducted several time studies to assess employee effort across domains such as R&D, sales, and marketing and found that more than 20,000 team members across marketing, sales, product, and other functions were spending up to half of their time developing marketing content. This pointed to marketing copy as a key potential area for automation. The current generation of off-the-shelf gen AI tools also seemed well suited to the text composition tasks involved, and risk was perceived to be relatively low since humans in the loop would still review the material. Finally, leaders also hoped that automating this work could demonstrate value rapidly because time spent would be easily measurable against the benchmark they had already established. The work involved both AI automation and developing new, streamlined workflows with clearer process guardrails. Ultimately, the effort led to a redesigned process with an 80% reduction in time needed from conception to publication for text-heavy content and a 30% reduction for image and video-heavy content. However, the use cases most ripe for automation may not be simply those that are easiest to measure. While the largest fraction of spending on AI experiments currently goes to the sales and marketing functions, back-office automation is what actually tends to produce the biggest ROI in terms of productivity improvement—legal, finance, and other services that have less well-established productivity metrics. Casting a wide initial net can help you zero in on the workflows that will yield the highest gains. Go Deeper The Gen AI Playbook for Organizations By Bharat N. Anand and Andy Wu HBR, November–December 2025 (37-minute read) What Executives Get Wrong About AI By John Winsor and Sangeet Paul Choudary HBR.org, November 10, 2025 (11-minute read) 3 Questions to Ask About Your Strategy in the Age of AI By Vijay Govindarajan, Venkat Venkatraman and Bill Achtmeyer HBR.org, November 6, 2025 (12-minute read) Make Sure Your AI Strategy Actually Creates Value By W. Chan Kim, Renée Mauborgne and Mi Ji HBR.org, September 5, 2025 (12-minute read) How to Make Enterprise Gen AI Work By Melissa Valentine, Daniel J. Politzer and Thomas H. Davenport HBR.org, September 18, 2025 (14-minute read) Credits Arjun Dutt (partner, Bain), Stefano Puntoni (professor, Wharton School and faculty co-director, Wharton Human-AI Research), Melissa Valentine (associate professor, Stanford and senior fellow, Stanford Institute for Human-Centered AI), John Winsor (serial entrepreneur and executive fellow, Digital, Data, and Design Institute, Harvard Business School).

Prioritizing AI Investments That Create Real Value

Summary.

Exclusive Benefits

Partner Center