Why do only a small percentage of GenAI projects actually make it into production?

Only about 5% of GenAI projects lead to significant monetization of new product offerings.

Article hero image
Credit: Alexandra Francis

It seems like everybody in tech is working on a new AI project. But how many of these generative AI (GenAI) initiatives will make it into production—much less lead directly to a new revenue stream?

In all likelihood, not very many. A 2023 Gartner report found that while tech executives are full of enthusiasm for AI initiatives, actual deployment rates remain low. On our podcast recently, Google Cloud’s director of AI/ML and generative AI, Miku Jha, estimated that in only about 15% of AI adoptions does the organization have a clear idea of what it wants to accomplish using GenAI. In other words, 85% of the time, it can be more of a spend first, think later situation: companies are investing in GenAI projects without a clear understanding of how they’re going to use the technology to create business value. Only in about 5% of cases, Jha suggested, will GenAI projects lead to significant monetization of new product offerings.

Direct monetization isn’t always the goal for GenAI initiatives, but for businesses considering how (and how much) to invest in this new technology, the question of how to guide a GenAI project from initial enthusiasm all the way to successful implementation is highly relevant. Let’s talk about why GenAI initiatives can be so challenging for organizations to adopt and identify some commonalities across the projects that do make it to production.

A Harvard Business Review article by Terence Tse, Mark Esposito, Danny Goh, and Paul Lee dug into some of the reasons why adopting GenAI projects is so difficult. Three they highlighted:

  1. Companies are still figuring out the previous generation of AI tools. Plenty of companies are still figuring out how to integrate “traditional AI” (that is, non-generative AI; tools like machine learning and rule-based algorithms) into their business operations. Maybe they’re still exploring traditional AI, or maybe they’re completely at sea. They’re not in a position to leverage the next generation of AI tools while they’re still getting their arms around traditional AI.
  2. GenAI is designed for very specific use cases. GenAI is not only much more complicated than traditional AI; it’s also designed for highly-specific use cases. While GenAI “is able to write a 5,000-word report in no time,” per Tse et al., “it cannot, for example, do a basic data entry task, like extracting and classifying driver’s license data, that traditional AI can do easily.” Business cases for GenAI are not necessarily easy to find, and GenAI won’t always deliver benefits worth the cost.
  3. We don’t know what we don’t know. The long-term implications of GenAI, including costs and the effects of regulation, are still unknown. Tse et al. compare our current moment to the late 90s: “While companies back then may have seen the need for setting up websites, few could clearly see the specific roles that the wider internet would play as an integral part of omnichannel strategies, let alone across devices and as phone apps.”

Our own conversations with users and customers, along with the volume of GenAI discussion happening across our Stack Exchange network, reveal a similar dynamic: tons of enthusiasm for the potential impact of AI projects, but considerable hesitancy around what kind of GenAI project would realistically support business goals and what practical steps to take to get started.

Expectations that don’t match up with reality are another issue. Organizations often start out with misguided expectations of their new GenAI project: everything from how long it will take to how much it will cost to how much value, if any, it will really deliver. Software and technology companies are moving rapidly—sometimes too rapidly—eager to keep pace with the competition without taking the time to think critically about their business goals and the best way to support them with GenAI.

The Harvard Business Review article has some insightful suggestions for making your GenAI adoption easier and more successful overall. Drawing from that article along with research and recommendations from Gartner on best practices for AI adoption, we’ve put together a quick checklist of qualities that successful GenAI projects have in common. Successful AI projects…

  • Start with understanding the business problems you’re trying to solve. Let that guide your tool selection. Don’t get distracted by novelty; focus on performance.
  • Remember that GenAI alone isn’t the solution. To make GenAI genuinely valuable and useful for your organization, you’ll need to understand how other technologies, like vector databases, come into play. (Here’s a great introduction to the topic.)
  • Keep humans in the loop. AI technologies are impressive, certainly, but they’re only as powerful and effective as the humans involved. According to the HBR article, “Humans play a critical role in guiding GenAI toward business goals, managing interactions within IT systems, designing the actions required for data going to and coming out of AI models as well as mitigating hallucinations—the made-up or outright false information produced by GenAI—that remains a major problem of GenAI today.”
  • Rely on trustworthy, traceable data. Establishing a clear trail from the source of the data to end-users is key to ensuring the reliability of the GenAI’s output. This is why the HBR article urges companies to “ensure that data lineage is a prominent feature in both their technology stacks as well as processes and workflow.” This gives companies and, crucially, their users the confidence that their GenAI is using the most complete, accurate, and up-to-date data available.

To learn more about how to guide an AI project from inspiration to production, check out our practical recommendations for adoption success or explore the Gartner report on AI adoption.

Login with your stackoverflow.com account to take part in the discussion.