Rippling CTO: The Human Data Layer for AI Agents
Rippling CTO Albert Strasheim explains how unifying HR, IT, and finance data creates a reliable 'hook' for AI agents. Learn why broad platform strategies and aggressive goal-setting are driving new engineering productivity in the agentic era.
The Strategic Shift to Unified Organizational Data
Rippling’s CTO, Albert Strasheim, outlines a fundamental shift in enterprise software strategy: the unification of HR, IT, and finance data into a single, real-time organizational graph. This 'hook' into the organization provides AI agents with the precise context needed to operate effectively, moving beyond fragmented silos. By leveraging the inherent accuracy of payroll systems, Rippling creates a trustworthy foundation for broader business applications, enabling agents to understand reporting lines, access permissions, and departmental structures dynamically.
Engineering for the Agentic Era
Leading engineering in this environment requires a departure from traditional focus strategies. Strasheim advocates for a 'maximally ambitious' vision supported by robust platform primitives. The engineering organization must adopt a 'smoothie' approach, breaking down silos to assemble cross-functional teams capable of composing complex AI products. A key tactical insight is the use of impossible goals, such as building ten data connectors in two weeks, to force teams to innovate and adopt AI-driven development practices. This pressure accelerates learning and reveals new operational efficiencies.
The Future of Platform Durability
To survive in the AI era, platforms must evolve beyond being mere systems of record. Strasheim identifies a flywheel of requirements: systems of triggers (detecting real-world events), systems of work (where humans and agents operate), systems of compute (running agents close to data), and systems of action (executing real-world side effects). This comprehensive approach ensures that platforms remain integral to business workflows rather than becoming obsolete data stores. Additionally, the reliability of these systems demands a balanced evaluation strategy, combining end-to-end tests with deep subsystem inspection to manage the non-determinism inherent in AI models.
Conclusion
The convergence of unified organizational data and agentic AI represents a significant opportunity for enterprise software. Companies that build broad, reliable data foundations and empower their engineering teams with ambitious, cross-functional goals will lead the next wave of productivity. The focus must shift from isolated tools to integrated platforms that can perceive, compute, and act within the complex reality of modern organizations.
Key insights
-
Payroll and HR data serve as a real-time 'hook' into the organization, providing AI agents with accurate, up-to-date context on reporting lines and access permissions. This foundational data layer enables more reliable and context-aware agentic workflows.
Impact: Enhances AI agent accuracy and reduces hallucinations by grounding them in verified organizational structures, leading to higher trust in automated processes.
-
The traditional advice to 'focus on one thing' is becoming obsolete in the AI era. Companies must adopt broad platform strategies that build foundational primitives across adjacent product areas to support diverse AI use cases.
Impact: Allows for rapid iteration and expansion into new markets by leveraging shared infrastructure, reducing time-to-market for new AI-driven features.
-
Setting near-impossible short-term goals forces engineering teams to rethink traditional workflows and adopt AI-driven methods. This pressure accelerates skill acquisition and reveals innovative solutions that would otherwise remain undiscovered.
Impact: Drives significant productivity gains by compelling teams to optimize processes and leverage AI tools effectively, creating a culture of continuous innovation.
-
Building complex AI products requires breaking down organizational silos. Assembling cross-functional 'smoothie' teams from multiple departments allows for the composition of diverse skills and components necessary for agentic systems.
Impact: Improves collaboration and reduces bottlenecks, enabling faster development cycles and more robust AI solutions that integrate seamlessly across business functions.
-
End-to-end evaluations are insufficient for ensuring AI reliability. Teams must combine evals with deep subsystem tracing and inspection to identify inefficiencies and failures that models might mask through resilient but suboptimal behavior.
Impact: Improves the reliability and efficiency of AI systems by catching underlying issues early, reducing operational risks and maintenance costs.
Action items
-
Audit your existing HR and IT data to identify opportunities for unification. Create a single source of truth for organizational structure and access permissions to serve as a context layer for AI agents.
Impact: Provides a reliable foundation for AI applications, reducing errors and improving the accuracy of automated workflows.
-
Reassess your product strategy to identify adjacent areas where platform primitives can be leveraged. Shift from narrow focus to a broader platform approach that supports multiple use cases.
Impact: Expands market reach and increases the value of your core infrastructure, enabling faster innovation and customer acquisition.
-
Implement a goal-setting framework that includes near-impossible short-term targets for engineering teams. Use these goals to drive experimentation with AI tools and new development methodologies.
Impact: Accelerates team learning and adoption of AI-driven practices, leading to significant productivity improvements and innovative solutions.
-
Break down silos by forming cross-functional teams for AI projects. Ensure these teams include members from different departments to leverage diverse skills and components.
Impact: Enhances collaboration and reduces development bottlenecks, resulting in more robust and integrated AI products.
-
Develop a comprehensive evaluation strategy that combines end-to-end tests with deep subsystem tracing. Monitor both the overall output and the underlying components of AI systems to ensure reliability.
Impact: Improves the quality and reliability of AI systems by identifying and addressing inefficiencies and failures that may be masked by high-level metrics.
Quotes
“I think you really need a maximally ambitious vision.”
“It's almost like the Trojan horse that gets you access to, you know, all of this foundational organizational data.”
“I think the era of that, at least for a while, is ending.”