# Building GTM AI Agents: Lessons from Deploying to 6,000 Users — Sait Izmit, Snowflake

## Executive summary

This talk details the process of building and scaling an internal Go-To-Market (GTM) AI assistant using Snowflake Co-work. The speaker emphasizes that successful deployment hinges less on technological capability and more on strategic execution: prioritizing quality over coverage, managing user trust, and planning for continuous architectural evolution. The system has processed over one million questions for 6,000 users, evolving from simple agent instructions to a complex architecture involving semantic views, skills, and MCP connections.

## Key takeaways

- Quality Over Coverage: Focusing on high accuracy (e.g., 95% correct) for a smaller set of critical questions is more effective than attempting to cover all possible data points with lower accuracy, as the first few interactions build user trust.
- Phased Rollout and Activation: AI tools must follow a controlled launch process: Pilot (proving accuracy), Beta (e.g., 10% of users, tracking >70% retention rate), and General Availability (GA). The biggest failure point is often activation and change management, not the technology itself.
- Addressing the Collapsing Wow Factor: After initial novelty wears off, the roadmap must evolve beyond simple Q&A. The progression should move from 'Talk to your data' (democratization) to 'Automate workflows' (integrations/MCPs), then to 'Team building skills,' and finally 'Hyper-personalization.'

## Technical details

- System Scale and Architecture: The agent now manages a large system spanning 15 semantic views, 85 tables, and 3,000 columns of data. It utilizes multiple MCP connections and nearly 20 skills.
- Architectural Evolution: The system architecture has progressed significantly: starting with simple agent instructions (initially managed via Google Docs), evolving to include CI/CD, evaluation infrastructure (unit tests, routing tests), and eventually incorporating Skills and MCPs to handle complex business processes.
- Feedback Loops and Logging: Implementing robust logging is crucial for identifying feature gaps in real-time. LLMs can classify logs, allowing teams to track question categories and subcategories, enabling the automated generation of sales enablement documents (e.g., battle cards) by connecting to sources like Confluence or Jira.

## Practical implications

- When building AI agents, prioritize the user experience and accuracy (quality) over comprehensive data coverage.
- Treat change management and activation as core components of the project lifecycle; allocate significant time to demos, sponsorship, and training post-launch.
- Plan for continuous re-architecting rather than seeking a single 'perfect' architecture. Embrace flexibility and iterative development using modern CI/CD practices.
- Establish detailed logging mechanisms early on to create real-time feedback loops that guide feature development.

## Topics

Generative AI, AI Agents, Snowflake, Change Management, Workflow Automation, Data Democratization, Snowflake Co-work, Cortex Analyst / Cortex Search

Source: https://www.youtube.com/watch?v=DrTdD-ttjCY