The Impact of Generative AI in Business
Generative AI is transforming how businesses engage with customers, enhance their customer experience (CX) at scale, and drive growth. In this VB Spotlight, experts explore real-world use cases, address challenges, and provide actionable insights to strengthen your organization’s generative AI strategy.
Challenges and Considerations
“The biggest upside of LLMs [large language models] is also the biggest downside, which is that they’re very creative,” says Jon Noronha, co-founder of Gamma.
Noronha highlights the creative yet unpredictable nature of large language models (LLMs). Companies developing production apps around LLMs face challenges in debugging, software testing, and monitoring. The engineering mindset is being redefined to accommodate these dynamic models. New infrastructure tools are needed to understand LLM performance at scale.
Irfan Ganchi, CPO at Oportun, emphasizes the evolving nature of the technology. Training LLMs, especially on proprietary knowledge, takes time. Maintaining brand consistency across various contexts requires human oversight. Ganchi sees promise in this technology despite the challenges.
“Working with LLMs is not like working with software. It’s not deterministic,” says Shailesh Nalawadi, head of product at Sendbird.
Nalawadi underscores the non-deterministic nature of LLMs. Small input variations can yield vastly different outputs. Unlike traditional software, tracing the reasoning behind LLM outputs is challenging. Crafting effective LLMs involves trial and error, and the tooling for updating and testing these models is underdeveloped.
Misconceptions and Real Value
Nalawadi dispels the misconception that LLMs are akin to real-time, indexed databases. LLMs are often trained on dated data, requiring specific user prompts for relevant responses. “Prompt engineering” becomes vital in business settings.
Jon Noronha suggests that “transformative AI” is a more accurate term than “generative AI.” It highlights the synergy between creative LLMs and existing knowledge. The real value lies in merging these two worlds effectively.
Ganchi reassures that generative AI doesn’t replace humans but enhances productivity. It automates repetitive tasks and improves efficiency. In customer service and marketing, it collaborates with humans to elevate performance and user experiences.
Strategies for Success
Ganchi stresses the importance of intentionality when deploying generative AI. Having a clear strategy, testing value incrementally, and addressing employee and executive apprehension are crucial steps.
Nalawadi emphasizes the need for infrastructure to measure AI system performance. A quantitative evaluation framework and a human evaluation framework should guide system evolution.
Jon Noronha recommends carefully selecting the problems where generative AI can have the most impact. Identifying tasks that are underserved or unwanted can lead to transformative outcomes.
Conclusion
Generative AI presents both challenges and opportunities for businesses. As the technology continues to evolve, strategic implementation and thoughtful problem selection will be key to unlocking its full potential.
Source: VB Spotlight Event – Presented by Sendbird
