3 min read

Aarthi Srinivasan has led product innovation in Personalization and AI at Walmart, Target, and now Amazon. In the role as Product Executive, Program Leader at Amazon Rufus (Gen AI Assistant), Srinivasan is focused on building and scaling Rufus, a generative AI-powered expert shopping assistant. She also advises startups on AI strategy, mentors founders, and angel invests in emerging tech ventures. “There’s so much innovation happening right now, and it’s up to businesses to harness that momentum to create solutions that matter,” Srinivasan says. “To do that, we need to empower teams to be customer-focused and data-driven, ensuring we’re always solving real problems.” 

Srinivasan talks about how companies can navigate the crossroads of technical innovation and business priorities – and how that is driving the next generation of AI innovation. 

Bridges and Barriers

So much AI innovation is happening in the startup community, where the startups are linked to an open innovation model that brings them closer to established organizations. “By fostering an incubator-style relationship with startups, businesses can position themselves to engage with the next big innovation at its earliest stages,” Srinivasan says. It’s a win-win because startups gain access to enterprise technology for deploying their products while enterprises gain access to support disruptive startups early in their development.  

This can also translate into ground-up adoption, where the excitement and momentum generated by developers and engineers quickly percolate to the highest levels of an organization. “When developers lead the charge in adopting new technologies, executives start to see the potential and are eager to catalyze that momentum,” explains Srinivasan. “This means everyone needs to get familiar with AI tools, and they need to understand that AI won’t replace them, but will augment everything we do and make it better, faster, more productive.” 

Experimentation is key to understanding which platforms and tools will be most effective. “The hyperscalers, such as AWS, Google Cloud and Azure, are offering easy-to-use platforms to experiment with Generative AI. This experimentation is less complex than a full IT transformation, which makes it accessible to businesses of all sizes,” Srinivasan says. “You can start small, test, and iterate towards a more robust solution.” 

Srinivasan emphasizes that the experimentation process is not a one-size-fits-all solution. “For example, when working with agentic architectures, companies can create specialized expert models,” she says. “This approach lets businesses customize AI solutions to their specific needs, whether that’s coding, content generation, search, or reasoning and decision-making.” 

For businesses looking to build highly focused systems, this “small expert model” approach makes it easier to experiment with AI in a cost-effective and manageable way. For instance, if you want to build a recommendation engine for a clothing retailer, your model will be highly specialized and tailored specifically for that task. 

While large-scale models are invaluable for deep reasoning and processing complex tasks, Srinivasan believes the future of AI will involve a combination of base models and specialized expert models. “Think of it like a textbook. It contains all the knowledge, but we need to train the model with examples like adding practice problems in the book, rewarding the system when it provides the right answers, just like a teacher would,” she says. This training and human intervention process helps ensure that AI behaves in a way that’s both practical and ethical. 

AI in Society

“As a product executive with deep experience in AI, I focus on helping business stakeholders adopt powerful new technology in a way that aligns with the company’s vision, mission, and strategy – while staying customer-centric and data-driven.

A key part of this process is identifying and onboarding mission-aligned startups and innovators early. This is done through accelerator programs, outreach to university research labs, and engagement with investors, VCs, and private equity firms that back the next big players.

By empowering leaders and fostering a culture of experimentation and learning, companies can drive innovation and build products and services that achieve rapid growth and strong customer adoption.”

Aarthi Srinivasan, Product Executive, Program Leader at Amazon Rufus (Gen AI Assistant)

As AI becomes more integrated into business models, companies must understand the layers that exist within AI development. From training and fine-tuning AI models to the specialized applications built on top of them, businesses are recognizing the power of both general and expert-level AI models.  

For example, AI applications for mental health, education or financial analysis all require specialized models built on top of foundational AI systems. “These models can be deployed across multiple data centers globally, allowing for a more flexible, scalable approach to solving real-world problems,” she adds. 

Disclaimer: The views expressed in this interview are her own from her experiences and not that of her employer.