AI Enterprise - Databricks & Glean | BG2 Guest Interview

Speaker 1 asserts that we have reached Artificial General Intelligence (AGI) and views Large Language Models (LLMs) as commodities. They categorize the AI landscape into three groups: the Super Intelligence Quest Camp, which raises concerns; the Research Camp, which is often overlooked; and the Value Creation Camp, focused on generating economic value efficiently. Speaker 2 highlights significant consumer adoption of AI tools like ChatGPT, while noting a stark divide in enterprise AI deployment, with 95% of AI projects failing. Speaker 1 attributes this high failure rate to a lack of experimentation rather than technological shortcomings, emphasizing the need for successful use cases that demonstrate tangible benefits, particularly to CFOs.

Successful AI implementation requires careful engineering and a strong team. An example from the Royal Bank of Canada illustrates how AI agents significantly reduce the time needed to analyze earnings reports. Other potential use cases include processing large volumes of documents and SEC reports. In healthcare, Merck's TEDI model aids drug discovery by predicting missing genomes, while 7-Eleven utilizes AI agents to automate marketing, allowing for efficient audience segmentation and campaign creation.

Speaker 1 notes that LLMs have become interchangeable across providers, stressing the importance of leveraging unique company data and understanding specific business processes. Many companies are developing non-specialized AI solutions, leading to ineffective applications. A solid data strategy is essential for successful AI implementation, ensuring that data is organized and ready for use.

The conversation also touches on the challenges of automating internal business processes, with Speaker 1 advocating for AI agents to help employees set and document weekly priorities. They compare the current excitement around AI to past enthusiasm for Robotic Process Automation (RPA), questioning how AI differs from previous automation cycles. Speaker 2 believes AI's unique capabilities, such as emotional understanding and autonomous functioning, set it apart from RPA, which was rule-based and lacked learning capabilities.

The dialogue addresses the trade-off between breadth and determinism in AI, emphasizing the need for CIOs to allocate budgets effectively in a rapidly evolving market. Concerns are raised about significant capital expenditure in AI, particularly regarding Nvidia, with implications that current spending levels may not be sustainable. Arvind stresses the importance of focusing on product development and value addition rather than merely spending, arguing that AI represents a distinct product category with potential revenue shifts from the services industry.

The discussion identifies three camps regarding the future of AI: one pursuing superintelligence, another questioning current learning approaches, and a third asserting that AGI may already exist. Speaker 1 highlights impressive results achieved by AI in challenging tasks, suggesting current capabilities can automate tasks and generate economic value.

The conversation also emphasizes the intelligence layer as crucial for enterprise value, with unique data being more valuable than just AI technology. Concerns about data security and governance are raised, emphasizing the need for proper measures to prevent misuse. Glean is mentioned as a valuable tool for improving organizational efficiency amidst coordination challenges.

Speaker 1 questions the longevity of companies like Databricks and Glean, drawing parallels to the tech landscape of 1998. They believe that while new valuable applications will emerge, existing companies are unlikely to become obsolete.

Arvind addresses the evolution of human interaction with technology, predicting a shift from manual data entry to more natural interfaces, such as voice commands. He identifies current challenges in automated data entry and suggests that Zoom could play a significant role by capturing conversations and extracting key information.

The integration of AI in meetings is discussed, with Glean's practice of recording meetings to capture valuable insights noted. The conversation highlights the heavy automation in marketing and back-office functions, as well as the transition from Excel to machine learning-based forecasting in finance.

Speaker 2 shares experiences with AI agents that enhance daily task organization and meeting preparations, reflecting a shift in instinct from relying on others to utilizing AI for data analysis. The discussion emphasizes AI's potential for effective collaboration, noting that while AI enhances output quality, it does not replace human effort.

In rapid-fire predictions, Ali and Arvind express optimism about revenue growth for OpenAI and Anthropic, attributing it to the continued success of ChatGPT and Gemini. They agree on the existence of an AI bubble, identifying three camps: those pursuing superintelligence, sober researchers focused on valuable contributions, and value-driven organizations cautious with spending.

Looking ahead, Ali predicts a decline in keyboard usage in favor of speech interaction, while expressing skepticism about the hype surrounding coding and automating customer service. Arvind is excited about proactive AI products that deeply understand user needs. Ali envisions Glean as a personal companion for every employee, focusing on confidentiality and personalized support, with future functionalities aimed at proactively assisting users with their tasks.