SaaStr 849: How We Built Our AI VP of Customer Success with SaaStr's...
The Official SaaStr PodcastFull Title
SaaStr 849: How We Built Our AI VP of Customer Success with SaaStr's CEO and CAIO
Summary
The episode details the creation and evolution of "QB," an AI VP of Customer Success built by SaaStr, demonstrating how custom AI applications can significantly enhance customer management and operational efficiency.
The hosts emphasize the iterative nature of AI development, the importance of deploying to production, and the universal applicability of such tools for B2B companies.
Key Points
- SaaStr built an AI customer success tool, QB, which began as a project management tool and evolved into an agentic system, outperforming off-the-shelf software by enabling rapid customization and enhancement based on customer needs.
- The development process highlighted that custom-built AI applications improve over time, with each iteration becoming more effective and valuable than previous versions.
- The creators stressed that the AI tools they share are universally applicable to B2B companies, not niche to their media and community business, focusing on core operational needs like managing customer relationships at scale.
- QB's development was iterative, starting with basic functionalities and expanding significantly after real-world deployment, leveraging user feedback to add new capabilities.
- The AI tool provided granular visibility into customer interactions and task completion, drastically reducing manual hours and improving efficiency compared to previous methods.
- The cost of building custom AI applications is often offset by significant savings in human input hours and operational costs, with hard costs for AI usage remaining surprisingly low for many use cases.
- Building and deploying custom AI agents requires ongoing maintenance and monitoring to address bugs, regressions, and evolving user needs, necessitating a human in the loop for oversight.
- The development of QB demonstrated the power of "agent hopping" for enhanced security, where sensitive data is stored externally and accessed via APIs rather than directly within the agent.
- The primary goal in building custom AI solutions is to achieve full customer coverage and address process gaps, rather than replacing human interaction entirely, especially for high-value clients.
Conclusion
Building custom AI agents like QB offers significant advantages in operational efficiency, customer engagement, and scalability by enabling rapid iteration and customization.
The iterative development and deployment of AI tools are crucial, requiring ongoing maintenance and human oversight to ensure effectiveness and address evolving needs.
Organizations should identify critical gaps in their customer success processes and leverage AI to automate workflows, thereby freeing up human resources for more strategic and high-value interactions.
Discussion Topics
- How can businesses effectively identify and prioritize areas for AI implementation within their customer success operations?
- What are the biggest challenges and best practices for maintaining and iterating on custom AI agents to ensure their long-term effectiveness?
- How does the balance between human interaction and AI-driven automation evolve as businesses scale their customer base and offerings?
Key Terms
- Agentic Journey
- The process of developing and implementing AI agents that can perform tasks autonomously and proactively.
- AI VP of Customer Success
- An artificial intelligence system designed to manage and enhance customer relationships, onboarding, and support.
- Low-code Platform
- Software development environments that allow users to create applications through graphical user interfaces and configuration instead of traditional computer programming.
- Agent Hopping
- A security practice where sensitive data is not stored directly within an AI agent but is accessed via external systems or APIs, reducing the risk of data breaches.
- QBR (Quarterly Business Review)
- A periodic meeting between a vendor and a customer to discuss account performance, progress, and future strategies.
- CSM (Customer Success Manager)
- A professional responsible for ensuring customers achieve their desired outcomes while using a product or service.
Timeline
SaaStr discusses their "agentic journey," starting with AI SDRs and VC tools, and then building custom agents like their AI VP of Marketing and the AI VP of Customer Success, QB.
Jason highlights that QB is the best thing they've built so far, noting it works better, has fewer bugs, and adds more value than previous iterations.
The hosts emphasize sharing universally applicable AI solutions, not niche ones, focusing on how custom agents can address common B2B challenges like managing large customer bases.
Amelia introduces QB, detailing its evolution from a sponsor project management tool to a comprehensive AI VP of Customer Success.
QB's origin story is shared: it started as a basic project management tool to replace an existing paid SaaS tool, lacking AI functionality, and evolved significantly.
A pivotal moment occurred when the AI started generating weekly sponsor emails, revealing its potential for much greater involvement with customers.
QB is described as a sophisticated onboarding automation tool, managing 13 core tasks with subtasks for each customer sponsor.
The increased visibility provided by QB is discussed, with admin views and daily email summaries offering insights into customer progress and engagement.
A comparison of human hours and inputs before and after implementing QB reveals a dramatic reduction, estimated at a 70% decrease in billable hours for agencies and production teams.
The cost-effectiveness of building QB is highlighted, with development costs being minimal compared to the significant savings in human input hours.
The AI token usage for all custom apps, including QB, is capped at $200 per month, demonstrating low hard costs for AI usage.
The evolution of QB from a customer portal to a fully agentic tool is shown through "then and now" examples, illustrating enhanced personalization and efficiency.
QB's ability to send individualized, highly customized sponsor emails in minutes is contrasted with the manual and time-consuming previous methods.
Amelia outlines the steps to reverse-engineer QB, starting with writing a detailed spec and iterating on it.
The emphasis is placed on not needing to be a prompt engineer to build AI applications; iteration and understanding the agent's capabilities are key.
The process of deploying QB involved loading the spec into a low-code platform, testing every function rigorously, and then integrating email capabilities.
Agent hopping is introduced as a security best practice, involving storing sensitive data externally (e.g., in Salesforce) rather than directly within the agent.
Decisions regarding login systems and data flow are crucial for deployment, with a preference for segmenting data to enhance security.
The critical step of deploying QB to a few pilot customers is discussed, highlighting that initial deployments will inevitably reveal issues that require fixing.
The dynamic nature of QB's development is emphasized, with new features and functionalities continuously added based on user requests and evolving needs.
Leveraging the data generated by AI agents to make them more agentic and proactive is a key strategy for enhancing their value.
The importance of maintaining AI agents daily is stressed, as regressions and bugs are common, requiring ongoing checks and updates.
QB is now considered QB1, the AI VP of Customer Success, fundamentally changing how SaaStr operates and manages customer relationships.
The hybrid approach of AI and human interaction is preferred for high-value clients, with the AI handling routine tasks and humans focusing on strategic engagement.
The evolution of QB has enabled SaaStr to be more flexible and responsive to customer preferences, meeting them on their preferred platforms like Slack or Google Meet.
The strategy is not to replace humans entirely but to augment them, with the AI handling the majority of automated tasks, especially for lower-value interactions.
The hosts challenge listeners to identify and address the most broken or gapped areas in their customer success processes by building custom AI solutions.
Plans for a follow-up session after Sastra Annual are announced, promising to share analytics and insights from further AI deployments.
Episode Details
- Podcast
- The Official SaaStr Podcast
- Episode
- SaaStr 849: How We Built Our AI VP of Customer Success with SaaStr's CEO and CAIO
- Official Link
- https://www.saastr.com/
- Published
- April 8, 2026