Greylock Change Agents• Jan 20, 2026• 56:00Interview

The Enterprise Brain for AI Agents with Glean and Cresta

From Greylock Change Agents

Executive Summary

Glean and Cresta, leaders in enterprise AI, detail their evolution from pre-generative AI search and assistance tools to building sophisticated, workflow-automating AI agents.
A critical theme is the tension between agent autonomy and control; both companies advocate for a 'human-in-the-loop' approach for any high-stakes task that modifies enterprise data, highlighting current AI reliability limits.
Both companies face strategic challenges from incumbent software and telephony providers who are beginning to restrict or charge for the data access essential for AI performance, creating new data moats.
The future vision for enterprise AI is a shift from a reactive, query-driven model to a proactive one, where AI companions anticipate user needs to automate significant portions of knowledge work and customer service.

12 quotes

Concerns Raised

Reliability of LLMs for high-stakes, autonomous tasks that mutate enterprise state.
The risk of 'agent sprawl' as enterprises build thousands of redundant or low-quality agents without proper governance.
Incumbent software and telephony providers creating data moats by blocking or charging for data access.
The emerging complexity of AI-to-AI conversations in customer support channels.

Opportunities Identified

Creating a horizontal 'enterprise brain' to provide unified context for all AI applications.
Shifting from reactive, user-prompted AI to proactive companions that automate significant portions of work.
Seamlessly integrating AI into existing workflows until it becomes invisible, like electricity.
Leveraging future multi-modal models that can understand on-screen context to unlock more sophisticated automation.

Key Themes

Evolution from Foundational AI to Agentic Systems

Both Glean and Cresta began by applying earlier AI models like Transformers to enhance enterprise search and human agent assistance. The advent of powerful LLMs like GPT-4 marked a pivotal moment, enabling them to transition to building 'agentic' products that can execute complex, multi-step workflows autonomously.

This trajectory illustrates the rapid productization of AI advancements, moving from tools that assist humans to systems that can perform tasks, representing a fundamental shift in enterprise software capabilities.

The Primacy of Enterprise Context

A core tenet for both companies is that agent effectiveness hinges on deep integration with a company's unique data, knowledge, and systems. Glean positions itself as a horizontal 'enterprise brain' providing this context to any agent, while Cresta's agents require rich conversational and customer data to function.

This highlights that the value of enterprise AI is not in the generic model itself, but in its ability to be safely and accurately grounded in proprietary, real-time business context, which is a key differentiator.

Balancing Autonomy and Control

The discussion reveals a crucial challenge in deploying agents: managing risk. For high-stakes actions, such as modifying a database or executing a bank transfer, both companies strongly recommend human oversight or hard-coding the core logic rather than relying on probabilistic LLMs.

This provides a pragmatic framework for enterprise AI adoption, emphasizing that a risk-based approach is necessary and that full autonomy is not yet feasible or desirable for many critical business processes.

The Shift to Proactive AI

A key future direction is moving beyond reactive agents that only respond to user commands. The vision is for proactive AI companions that anticipate needs, suggest actions, and automate work without being prompted, with Glean aiming to handle 'half of a knowledge worker's tasks'.

This signals the next paradigm in human-computer interaction, where software transitions from a passive tool to an active collaborator, promising a step-function increase in productivity and efficiency.

Navigating Data Access and Ecosystem Politics

A significant business hurdle is emerging as incumbent software and telephony providers begin to block or charge premium fees for data access. This creates data moats that challenge the 'connect-to-everything' model of horizontal AI platforms like Glean and Cresta.

This underscores that the future of AI is not just a technical challenge but also a strategic one, where data access will be a key competitive battleground influencing the architecture of the enterprise AI stack.

Get started free

Topics

AI Agents Agentic AI Enterprise AI Enterprise Search Customer Support AI Contact Center AI Generative AI LLMs Transformer Models Human-in-the-loop AI Governance Proactive AI Data Access Data Moats Workflow Automation

Processed May 4, 2026 yt-dlp + mlx-whisper + Gemini

You're reading a preview

Get started free →