Rippling's Parker Conrad: core business systems become AI harnesses
Key points
Key takeaways from a Grit (Kleiner Perkins) "Leaders" interview with Rippling founder and CEO Parker Conrad (November 2026):
Core systems become LLM harnesses. Conrad expects core business systems — HCM, finance, sales — to look in 5–10 years like "giant harnesses and orchestration systems around LLMs": a specialised domain harness, domain-specific tools and work orchestration, with the LLM as the intelligence at the centre.
Permissions are the AI unlock. The HCM model of row-level access — employees see their own records, managers their reports', HRBPs their client groups — is exactly the permission model AI agents need when querying sensitive data like compensation and performance; Conrad argues Snowflake and Databricks cannot implement it.
Airlocks to the real world. Rippling's edge is breadth of "airlocks": moving money, filing taxes, updating insurance enrolments and shipping computers become specialised tools on top of the harness, so the LLM can act, not just answer.
Software written inside core systems. Conrad expects most future software to be built inside platforms like Rippling: shared capabilities built once for 100,000 customers go through normal code review, while bespoke single-customer apps are built by coding agents inside the system, interoperable with everything else.
"Go and see" at scale. Rippling's first principle is executives getting to the ground truth; AI makes it practical for a CEO to read every pull request or inspect every sales call rather than rely on filtered management layers.
Better decisions, faster. With permissions and data in place, executives gain far more context — catching problems far more quickly than when information is filtered up through multiple layers of management.
Read more: YouTube — Grit (Kleiner Perkins)
