LLMs in the Enterprise - Avoid the Event Horizon

For many companies, the cost and time to implement new systems is a major obstacle, slowing down ROI and impact. A key trend is for vendors to add Large Language Model (LLM) capabilities to simplify access to complex software functions, much like giving users a treasure map to navigate. Done right, this can speed up implementation, reduce training time, and allow real-time policy configuration and updates.

Enterprise computing isn’t my main focus these days, as I'm far more interested in emerging robotics and AI in general at the Robotics Hub VC . I revisited my former beat earlier this week to share some insights at Identiverse . Many thanks to Sanjay Nadimpalli , CEO of Tuebora , for the invitation. (I funded Tuebora back in the early days at Citrix Startup Accelerator*.)

Here’s a summary of the talk:

• LLMs (AI) represent a significant shift in computing, greatly affecting how we use knowledge.
• Much media coverage is exaggerated, partly because we tend to assign human traits to anything that seems intelligent. This is understandable given the impressive abilities of LLMs, like giving robots implicit knowledge about the world .
• However, LLMs are still in their infancy and have clear limitations, such as producing false data when pushed beyond their limits (known as hallucinations).
• In high-stakes environments, it's wise to stay well away from the event horizon of promised LLM capabilities. That is, start where the technology is reliable and stable.
• Enterprises using LLMs should : <Prevent costly mistakes> <Keep humans in the loop> <Use strong safeguards against AI errors, or “hallucinations”> <Manage data leakage risks>
• A key opportunity for vendors and customers is to use LLMs to simplify access to deeply buried functions, effectively handing users a treasure map to navigate complex software.
• This can speed up installations, reduce training time, and keep policies updated. This will lead to far more agile IT organizations, able to implement new initiatives more quickly, and improved productivity.

In, for example, the Identity Governance and Administration (IGA) space, this means faster setups, easier daily use, and quicker review cycles. If you’re looking for an IGA solution, I recommend giving my friends at Tuebora a call to explore the possibilities.

__

At the Robotics Hub , we take a pragmatic approach to finding and working with ground breaking companies that use emerging robotics and AI breakthroughs to solve large scale, real world challenges.

* Citrix Startup Accelerator was an pre-seed enterprise-computing startup accelerator that ran from 2010-2015 in Santa Clara, focused on emerging technologies and markets.