Most of the market is building for workers who do not exist. Tools for ephemeral automation, non-human seats on the org chart, agents without a home, all of it presumes the interesting problem lives in the machine layer. Rippling is leaning hard in the opposite direction, doubling down on the human layer and treating the technology that makes HR, IT, and finance possible as a single connected surface, something closer to an employee graph.
That bet, and the engineering philosophy behind it, comes from Albert Strasheim, CTO at Rippling, who has led the company's engineering organization through a period of massive scale. His argument is that an agent which knows who reports to whom, who has access to what, and the true shape of an organization arrives with far more context and a far better ability to deliver real work. Building that human-centric foundation, he contends, is the jet fuel for the agents of tomorrow.
The employee graph becomes the foundation agentic AI needs
HCM and payroll look, at first blush, like the most boring back-office software imaginable. Old-school, unglamorous, a system of record for getting people paid. But that framing misses what payroll actually is. Everyone in a company is incentivized to keep it current. Every new hire has to be added so they get paid. Every departure has to be recorded so the payments stop. The result is a live, self-correcting picture of the organization. "It's this hook into the organization that is up to the minute accurate," Strasheim says.
Once that hook exists, nearly every other business application gets a little bit better. IT and access management improves because the system already knows who joined, who left, who reports to whom, which department someone sits in, and which subsets of data a given role should touch. Payroll stops being an endpoint and becomes a trigger for security, provisioning, and role-based permissions across the whole org. Agents inherit all of that context automatically. They understand the reporting structure and the access boundaries without any additional engineering, because the graph already encodes them.
The strategic move is to recognize that the boring system was never the destination. "It's almost like the Trojan horse that gets you access to all of this foundational organizational data," Strasheim says. What started as a system of record for pay turns into the entry point for building far more ambitious, organization-aware software in adjacent domains.
Unifying HR, IT, and finance demands one maximally ambitious platform
The through-line for all of it predates AI, and it starts with scope rather than technology. "You really need a maximally ambitious vision," Strasheim says, crediting CEO Parker Conrad as the original vision maximalist behind Rippling's compound startup idea, the notion of building business software across the entire company stack rather than picking a single lane.
The vision has to come first. Only after it is set does the question of how to build it make sense, and more often than not the answer resolves into a platform play, a fundamental set of building blocks rather than a scramble for point solutions. Rippling's early team invested heavily in those primitives, and that investment is what the company keeps carrying forward. "You need a big vision, and then you need a big platform to do all of the things, and probably not just in one product area, but in a number of adjacent product areas," Strasheim says.
Going broad this way produces compounding returns. Use cases fall out of the platform, solutions fall out of it, and happy customers follow from building across a wide surface rather than a narrow one. It directly challenges the conventional wisdom that a company should focus and do just one thing well. That era, in this reading, is ending for now, replaced by a go broad, go fast, go big posture that AI has only made easier to sustain.
Breaking down engineering silos unlocks true cross-functional teams
The trouble with a broad platform is that no single team can ship anything interesting from inside its own boundaries. Building almost anything worthwhile now means pulling a few people or a few components from across three, four, or five teams at once. "There's never gonna be one team that has all of the puzzle pieces or all of the building blocks under their control," Strasheim says. Reorganizing to concentrate all the pieces under one roof does not work either, since a single team responsible for everything is not a team, it is a bottleneck.
The alternative is to stay agile about how teams get assembled. Rather than rigid reorgs, Rippling's engineering leaders, Strasheim among them, have jumped in and line-managed cross-cutting efforts, deliberately blending platform engineers with product engineers. He calls the result a smoothie, a bit of a technology smoothie and a bit of a team smoothie, and then something ships. Pairing an AI-fluent engineer who has figured out harness engineering with colleagues still learning it spreads new practices organically, and each person carries those practices to the next project.
The forcing function that makes it click is an impossible goal. When Rippling set out to expand its data connectors, the mandate was blunt: a task that used to take weeks per connector now had to produce ten of them in two weeks. The team spent a couple of weeks building harnesses to write specs, generate code, and validate output, and everyone who touched the effort came out thinking differently about how software gets built.
Rippling data cloud turns business data into a living platform
The next phase pushes the graph outward, ingesting, processing, and computing over far wider sets of business data. App Studio, which shipped roughly a year ago, let customers define custom objects and functions and build canvases on top of them. Since then the platform has added data connectors that pull third-party data into custom objects or into large-scale lake objects, plus a unified catalog where native Rippling records and custom objects live side by side. When an agent has to write queries across every business system, "having access to a catalog is critically important," Strasheim says. A SQL-based transformations product lets teams pre-materialize custom views, and a heavily expanded reporting layer lets AI build dashboards outright. "It's dashboards all the way down," he says, describing agents that read the catalog, grab sample data, and assemble tailored insights that replace the old point-and-click grind.
Data alone, though, is not enough to survive as a platform. Everyone in the industry will fight to import everyone else's data, bilateral data treaties will emerge, and customers will rightly argue over whose data it really is. Winning that fight requires becoming several systems at once. A durable platform has to be a system of triggers that sees real-world events first, a system of record that stores them, a system of work where humans and agents actually operate, a system of compute where agents run close to the data, and a system of action that produces real side effects, a payment made, a tax form filed, a laptop shipped. String triggers to data to compute to action with real work happening on top, and the platform sticks around. "If you're merely a database with some forms on top and you're not a key part of those workflows, I think you would struggle in the long term," Strasheim says.
That framing positions Data Cloud as connective tissue rather than a warehouse, a place where agents query, compute, and act on business data in real time while humans engage with it in natural language. It also names the next hard problem honestly. As these non-deterministic workflows multiply, keeping them reliable becomes an observability and maintenance challenge the industry has barely begun to solve, and evals are only the beginning of the answer. The companies that treat the employee graph as a living operating layer, and that build the testing discipline to keep it trustworthy, are the ones positioned to turn organizational data into durable, agent-ready leverage.
To hear more of Albert Strasheim's insights on the employee graph as a foundation for agentic AI, breaking down engineering silos, and the Rippling Data Cloud, listen to the full episode on the Dev Interrupted podcast.
