As corporate boardrooms across the globe get swept up in AI mania, executives are feeling the pressure to integrate artificial intelligence into the structure of their companies. And while there are ample stockpiles of buzzwords and slick demos for them to turn to, these leaders are largely navigating uncharted territory without a playbook for how to press forward.
In a wide-ranging conversation at the Fortune Brainstorm AI conference on Tuesday, executives from several pioneering AI companies shared their insights about implementing AI, along with guidance on feasible objectives and successful strategies. Perhaps just as important, the executives—from Adobe, Workday, Credo AI, and Snowflake—discussed some of the most frequent mistakes they’ve witnessed among companies scrambling to bring AI into their organizations.
According to Navrina Singh, the founder and CEO of Credo AI, the successful companies are the ones that involve people from data science, risk, compliance, and policy, so that everyone understands what’s at stake.
“AI truly is a techno-social problem,” Singh said. “When you have something which is going to be that economic engine, that societal change engine, you really need to make sure that the oversight deficit between technical and business stakeholders is managed appropriately.”
Each organization has its own tolerance for risk, she added, so each needs to figure out what it’s okay with and show it’s responsible. Some companies mess this up, but the ones that get it right are doing well, Singh said.
Workday’s vice president of AI, Kathy Pham, stressed the importance of aligning AI efforts with customer needs. “I think a big mistake would be to only build fun models without knowing or thinking about the real practical uses that can benefit the users, the customers,” she said.
That means sticking to business basics like going out “into the field” to understand the real needs of customers, whether they be truckers or health care workers.
Companies who don’t consider their customers in their zeal to jump on the AI bandwagon do so at their own peril, explained Adobe VP of generative AI Alexandru Costin. In the content creation field, Costin said he’s seen companies move too fast, leading to confusion and backlash from paying customers. Another mistake Costin often sees is companies underestimating the importance of high-quality, unique data for building successful AI models.
Implementing AI responsibly and earning trust is critical given the impact the technology will have. “We think AI disruption is going to be bigger than any previous disruption in human history,” Costin said.
Sridhar Ramaswamy, senior VP of AI at Snowflake, explained that it’s crucial for everyone, especially business leaders, to grasp what AI is and what it is not. AI is like a powerful language tool, but it’s not a magic solution that automatically understands complex tasks or predicts outcomes. Some people mistakenly expect AI to provide quick fixes, like boosting profits by a specific percentage, but in reality, the hard work of understanding and predicting complex situations still requires human input. Recognizing these limitations helps set realistic expectations and prevents overhyping AI as a cure-all technology.
“Understanding what it can and cannot do, I think, will help a lot, first of all, in not overpromising with yet another technology that we are like, ‘Oh, this is going to save the world,'” Ramaswamy said. “I think this lets us calibrate ourselves to what can be delivered in reasonable periods of time.”