In an exclusive interview with ET, New York-based Raghavan said his conversations with companies are increasingly focusing on smaller, fit-for-purpose models as they can be more efficient.
About AI use cases
The first use case is general customer service, which is usually the simplest value proposition. It’s one of those areas that has moved from pre-GenAI to GenAI, where the value proposition remains the same, but now the new technology allows for faster scaling.
The second use case is code and modernization or, more broadly, developer productivity. Any company that has been around for the last 10 to 15 years probably has a large software footprint. This means you have legacy software that you are always looking to modernize while maintaining your domain and technology. The role of AI in modernizing your software, applications and code base is enormous.
The third use case is digital work, specifically administrative automation. This includes human resources (HR) onboarding, talent recruitment, supply chain procurement, and financial processes. These are the main areas where, globally, we are starting to see traction, which is why we are prioritizing our technology investments.
Discover the stories of your interest
About business use
You’re hearing everyone double down on the idea that you don’t need the biggest model that can handle a billion tasks for every use case. Increasingly, the conversation is shifting toward smaller, fit-for-purpose models: small language models (SLMs), which may have 8 billion parameters but are trained on large volumes of data. If I don’t need the same model to translate COBOL to Java, write Shakespearean poetry, and generate images, I’ll opt for “fit for purpose” models. Many companies are now looking at profitability and saying, “Give me the smallest model and the cheapest infrastructure.”In cost savings
The difference between running the largest, most powerful GPU versus a more affordable one can result in cost savings ranging from 2x to 10x. At IBM, we have focused on applying AI in HR, where we have dramatically increased the value that AI delivers. Today, more than 80% of employee inquiries are handled through self-service, saving between 12,000 and 50,000 hours of manual work. Furthermore, the quality of service is better than what a human being could offer.
About the implementation of AI
In the short term, much of the value proposition for companies revolves around efficiency and automation. Successful customers do not engage in pilot testing or proof of concepts (PoCs) scattered across multiple lines of business. If you do, you may end up with 100 PoC and nothing to scale. Instead, they focus on a specific area, whether defined by a product, business unit, or process, and try to test and scale it. Customers who follow this strategy tend to see a faster return on investment than those who run many PoCs.
About the challenges
We see between 3 and 4 common challenges, regardless of geography. Cost, data preparation, skills and trust, and governance are the four common themes that customers are increasingly focusing on.
Disclaimer:
The information contained in this post is for general information purposes only. We make no representations or warranties of any kind, express or implied, about the completeness, accuracy, reliability, suitability or availability with respect to the website or the information, products, services, or related graphics contained on the post for any purpose.
We respect the intellectual property rights of content creators. If you are the owner of any material featured on our website and have concerns about its use, please contact us. We are committed to addressing any copyright issues promptly and will remove any material within 2 days of receiving a request from the rightful owner.