As Generative AI rapidly transitions from experimentation to production, industry leaders gathered at the MachineCon GCC Summit 2024 event to share real-world examples and insights on leveraging it.
The panel, moderated by Shashank Garg, CEO of Infocepts, included executives from Broadridge, Rakuten India, Grant Thornton, AlphaSense, and Schneider Electric.
It included Sheenam Ohrie (Managing Director at Broadridge), Anirban Nandi (Head of AI Products & Analytics at Rakuten India), Kalpana Balasubramanian (CEO and Chief Thinker at Grant Thornton), Amod Deshpande (Country Managing Director, India at AlphaSense), and Madhu Hosadurga (Global Vice President, Enterprise AI at Schneider Electric).
GenAI Adoption in Highly Regulated Industries
Broadridge’s Sheenam Ohrie kicked off the conversation with real-world examples of how their organisations are leveraging Generative AI.
Ohrie explained that Broadridge, a highly regulated tech organization that provides a platform for investor communications and capital markets, is using it for customer-facing applications such as Bond GBD, an interactive chatbot that helps investors understand the vast landscape of available bonds.
They also use it internally for a chatbot called Broad GBD, which is used daily by 2,500 to 3,000 associates. Additionally, Broadridge has developed Obstitutely, a tool that enables transparency and smoother operations for transaction settlements, and Distribution GBD, which provides insights to wealth managers.
Ohrie emphasized the importance of focusing on non-functional requirements (NFRs) before starting any innovative projects.
“The most important thing when we start off on anything which is innovative is to first understand how we’re protecting the PII data, how we’re protecting any kind of leakage of data to the external world, and the third is cybersecurity,” she said. “We really focus on these three aspects from an NFR perspective before we start off anything.”
Nandi highlighted Rakuten’s aim to become an AI-empowered company by 2024. They have developed their own Japanese LLM and are applying AI in three categories: internal applications for increasing productivity, customer-facing applications, and partnerships with companies like OpenAI and Anthropic. Rakuten has built an in-house framework called Rakuten AI, which has over 20,000 active users and 7,000 daily active users.
Balasubramanian mentioned that Grant Thornton, a digital consulting firm specializing in new tech, is actively using Generative AI to assist in content production, training, legal risk management, and contract management. They see the most value in customer-facing applications where AI can significantly increase customer value or reduce time.
Deshpande shared that AlphaSense, a financial market research product, has been integrating Generative AI since 2017-2018. They are launching a new product that uses a state-of-the-art Generative AI stack to provide personalized summaries and suggestions to users based on their behaviour and interests.
“The entire search model is changing to a push model,” Deshpande explained. “Before you know it, you will be getting a reading summarization and suggestions that this is what you should be looking at right now.”
Hosadurga discussed how Schneider Electric, with 160,000 employees worldwide, uses AI for chatbots, knowledge management, last-mile automation, and content generation. They have automated a significant portion of their digital asset management for product catalogs and images using Generative AI, which previously required agencies.
“Most of our images get into multiple platforms like the e-commerce shop, ourselves, the partners, the marketplaces,” Hosadurga said. “So it’s a very difficult landscape out there to manage these digital assets consistently across thousands of platforms. So far we were using agencies to do this job. Now, thanks to Gen-AI, a good amount of that work has been automated.”
Data Security and Ethical Considerations
The panelists also addressed data security and ethical considerations. Hosadurga explained that Schneider Electric has blocked public AI platforms and instead uses an enterprise version with a one-way architecture to ensure confidentiality. They enrich pre-trained models with internal information using a retrieval-augmented generation (RAG) architecture.
Nandi highlighted the need for observability and hallucination measurements in Generative AI models. Rakuten has developed a product called Gen-I that detects security threats in prompts and provides observability for responses.
“Internally, a business can make a decision, I don’t even want to send the prompt to OpenAI or even our own LLM model,” Nandi said. “Some businesses choose to take the risk and send prompts to AI models, but they want to be notified when a user sends a prompt to verify it is appropriate. Once the prompt is sent and a response is received, observability is crucial to detect any inaccuracies or hallucinations in the generated output.”
This has significantly improved their internal AI applications and given them the confidence to develop customer-facing applications.
Solving Real Business Pain Points
The panel further focused on the importance of solving real business pain points with Generative AI. Garg gave an example of using AI to create hyper-personalized product descriptions for e-commerce websites, significantly increasing customer value and reducing the time required for manual copywriting.
“If you combine the power of data and behavioral profiles that we already create for our clients, potential clients in the digital world, and then use Gen AI to on the fly create hyper-personalized product descriptions, what’s called copywriting,” Garg explained. “So, going from 7 copywriters writing manually product descriptions for 100,000 products, you go to millions of descriptions on the fly using the power of Gen AI.”
Nandi emphasized the need to focus on solving business pain points rather than just generating more content. “Imagine a situation where you have customer service, and you call up, and it gets escalated to the next agent, and the next manager, next one.
Why does the wait time increase? Because from the first person, when it goes to the second person in the call center line, somebody is actually going through the transcript of what was discussed. Can Gen AI actually summarize that?”
As the panel concluded, the speakers agreed that Generative AI is rapidly moving from experimentation to production, with enterprises across various industries finding innovative ways to leverage the technology. However, they stressed the importance of implementing proper security measures, ethical guidelines, and observability to ensure responsible and safe adoption.
The post Generative AI Moves from Hype to Enterprise Adoption appeared first on Analytics India Magazine.