AWS re:Invent 2025 - How AI Agents are Changing the Way Revenue Teams Work (STP110)

Summary of AWS re:Invent 2025 Presentation: "How AI Agents are Changing the Way Revenue Teams Work"

Introduction

• The presentation was given by Shriram, the co-founder and CTO of Rocks, a company building revenue agents for the global 2000 enterprises.
• Rocks' mission is to "secure and grow the world's revenue" by redefining revenue systems from the ground up using AI-native architectures.

The Evolution of Revenue Systems

• In the on-premises era, Siebel was a powerful but rigid and expensive CRM system.
• The advent of cloud computing led to the rise of Salesforce and HubSpot as the system of record, while point tools like Clearbit and Outreach became the system of action.
• With the emergence of cloud 2.0 and data warehouses like Snowflake, data moved from CRMs into the warehouse to be integrated with product usage data and dashboards.
• This historical shift has led to revenue work becoming increasingly fragmented, manual, and inefficient, making revenue teams ripe for AI augmentation.

Challenges with Current Revenue Workflows

• Data is highly fragmented across multiple systems, including CRM, support, and product usage data, as well as unstructured data like call recordings, emails, and documents.
• Workflows are manual and inconsistent, with reps spending only 20% of their time actually selling, while the remaining 80% is spent on administrative tasks.

Rocks' AI-Native Revenue Operating System

• Rocks has built a fundamentally different AI architecture to unify context and action across the entire revenue lifecycle.
• The architecture consists of four key layers:
  1. Data Layer: Structured and unstructured data from various sources
  2. Context Layer: A knowledge graph that extracts entities and relationships from data to provide context for the agents
  3. Agent Swarm: Turnkey enterprise agents that run the entire revenue lifecycle, from lead to close, renewals, and churn prevention
  4. Delivery Channels: Agents are delivered across multiple channels like iOS, Slack, and web to meet sellers where they work

Operationalizing the Agent Swarm

• For each account, there is a swarm of AI agents orchestrated by Rocks, responsible for driving the entire revenue lifecycle.
• The agents work together to convert insights into actionable steps, automating tasks like research, email drafting, and deal risk analysis.
• Rocks' platform is built entirely on AWS, with a clean separation between the customer's data plane and Rocks' data plane, ensuring reliability, security, and global reach.
• The key to Rocks' approach is the multi-agent orchestration, where each workflow is broken down into a sequence of steps, and the right agent is chosen for each step.

Adoption and Impact

• Rocks has already achieved significant scale, processing 400 billion tokens per month, making them part of the "Trillion Token Club" alongside only 20 other organizations.
• Rocks has helped customers like Ramp, the fastest-growing SaaS company, achieve a 15% increase in six-figure deals, 90% reduction in meeting prep time, and 20% more meetings booked.
• The key to Rocks' success is their ability to deliver consistent, scalable execution across the entire revenue organization, making the best reps better and the average reps better than average.

Conclusion

• Shriram believes that just as coding agents 10x'd productivity for engineering teams, revenue agents will 10x the productivity of enterprise revenue teams by providing every account with an always-on agent that operates with full context.