Thoughtful AI

As we begin a super charged 2024 (super charged Indian economy, trillion dollar GenAI industry, electric/self driven cars, advances in cancer cure etc.), I wanted to share a few perspectives specifically on the direction & potential of AI / GenAI in 2024 in our industry.

Firstly, its really important to take a step back and ask the “Why AI or GenAI” question, so as to not fall into the “FOMO on AI” trap. The fundamental business levers are still the same. Are we stepping up the ante on Profitability (Cost), Productivity (Time) or Creativity (Value). As a startup mentor & investor, I see so many startups these days introducing themselves as “We are a ML or AI or LLM” company. My advise: please don’t. That’s the high school equivalent of identifying oneself as a trigonometry company or calculus company . Keeping the focus on business fundamentals & value creation at all times, helps maintain the perspective that technology is a means to an end, not an end in itself. The end being a business, customer or social need / impact.

Secondly, while the interest in and the capability of LLMs and GenAI are reaching stratospheric levels, so are costs and complexity of the models. In general, we have found it helpful to take a step back and categorize business goals starting with low hanging fruits, and take a phased approach as opposed to taking a “To a Hammer everything looks like a nail” approach when it comes to LLMs/GenAI. We have found that Enterprise use-cases lend themselves to lower risk, higher tangible rewards in the first phase of deploying Gen AI solutions. For e.g., Conversational AI for Customer Support, Employee/HR support or code generation for legacy or newer programming languages. Those learnings build confidence, a repeatable model and a strong ability to take risks with consumer use-cases involving hundreds of millions of users/shoppers/customers.

Finally, commercializing AI isn’t easy. Using AI for internal use-cases (enterprise or consumer) is one-thing. Commercializing AI as a business/product/service is challenging. OpenAI, Azure AI, GoogleAI are currently only a handful of commercially available AI solutions because of the inherent high complexity, cost & scale of data storage, privacy, compute, sciences, and the evolving data sciences & algorithms such as GPT4.x.

For e.g, take our Rakuten SixthSense Cognitive AI solution for AI-OPS, which offers Forecasting, Anomaly Detection, Correlation and Causation of critical Data Center Events, Application & Infrastructure Performance at massive scale.

As we learned with GTM of Rakuten SixthSense, it is critical to maintain a sharp focus on use-cases and evolve the solution along with customers, while training the models on a narrower set of use-cases and data sets to manage costs like GPU, Storage, Ingestion & Compute costs which can quickly overwhelm the business model. Being really smart & creative about what, when & where to store/ingest data & compute insights (at consumption end or creation end) can help unlock value while keeping costs in check.

For companies that are considering broad-based AI deployment across a variety of businesses, identifying a dozen or so broad business & consumer use-cases that can benefit from AI and mapping them to specific AI solutions can help with the “build-buy-use” decision of the underlying AI offerings.

As a internet-first conglomerate with 75+ internet companies under our umbrella (from Ecommerce to Fintech to Media, Communications, Mobile & Sports), our AI creation-adoption-scaling strategy has evolved over the years through a lot of experimentation, risk management, real world business impact and feedback from users. We have thousands of data scientists and AI experts working across dozens of business problems trained over hundreds of Petabytes of user, business & systems data - whether its using LLMs for Semantic Search or Cognitive AI for Supply-Demand Forecasting for Ecommerce, IT-Ops or Fraud Detection or Customer/Product Graphs for Recommendations.

To summarize, AI is well positioned between the bookends of necessity? & novelty. Being strategic and thoughtful about outcomes, a focused GTM and selecting the best solution for the problem, helps maximize ROI. Will plan to share some of our key learnings / frameworks for choosing the most effective AI solutions for a diverse business portfolio in my next post.

Until then, Happy New Year & hopefully Happy LLMs ??