Citi^® Gen AI Summit 2024 Takeaways: Citi Ventures’ Perspective on Gen AI

For the second session the 2024 Citi^® Gen AI Summit, I was pleased to give a talk outlining Citi Ventures' Gen AI investing thesis. I also invited Gaurav Oberoi, CEO of Lexion — a Citi Ventures portfolio company recently acquired by DocuSign — to discuss the learnings he gleaned from building a category-leading AI company from the ground up and navigating the often arduous journey to a successful exit. Read on for the key takeaways from that session.

Citi Ventures is looking to invest in four areas of Gen AI

We at Citi Ventures are particularly interested in solutions that address the Gen AI ecosystem’s primary challenges and/or areas of growth. Those include:

1. Cost. It’s no secret that LLM use for training and inferencing can become expensive quickly, so we’re keen on solutions that can help make Gen AI cheaper and more accessible, such as model routing optimization.
2. Performance. Gen AI applications, which by nature are nondeterministic, still have problems providing predictable and accurate outputs — as seen with hallucinations. To help solve for that, we’re spending a lot of time looking at retrieval-augmented generation (RAG) solutions, embedding models, different types of evaluation metrics and more.
3. Security. At Citi, no application will go into production if it presents a data exfiltration risk. As the bank works to operationalize Gen AI across its many businesses and functions, we’re investing heavily in cybersecurity guardrails and solutions that can block prompt injection and other attacks on Gen AI apps and models.
4. Applications. Making smart, strategic AI investments is core to Citi’s Transformation, long-term plan and mission to empower growth and progress around the world. So we’re actively investing in application-layer Gen AI solutions in search and knowledge management, customer engagement, software development and more.

Key insights from Lexion’s CEO on building an AI company

I then invited Lexion CEO Gaurav Oberoi to share his insights about a range of AI- and startup-related topics in light of Lexion’s recent acquisition by Docusign.

• On the power of LLMs:
  As an AI-driven contract review automation and workflow company, Lexion saw LLMs’ potential fairly early on, discovering LLMs could augment their existing technology stack by closing gaps where some of their proprietary models would underperform. Gaurav advised founders to consider the value Gen AI can unlock at the application layer — in Lexion’s case, it allowed the company to offer its customers new services like automating email content creation and approvals processes.
• On scaling an AI company:
  As the Gen AI hype cycle starts to wane, Gaurav stressed that AI startups need to get back to business basics and focus on driving real ROI for their customers. While consumers and enterprises alike are still fascinated by AI and its potential, in the end only one thing will ensure high customer retention: delivering results.
• On partnering with strategic investors:
  Gaurav imparted that founders raising funds should consider not only the capital they’ll receive from investors, but also how the investor can support their company’s mission.
  
  For startups trying to sell to large enterprises, taking strategic capital from corporate venture teams makes a lot of sense — they can help founders build relationships within the enterprise and can often offer marketing support and insights into the target sector as well. Through Citi Ventures, Lexion met key business leaders within Citi and received guidance on how to navigate a large corporation — knowledge he now takes with him into his new role at Docusign

Are you a Gen AI innovator looking to connect? Please email me at vibhor.rastogi@citi.com