At Rising this week, Workday Data Cloud opened the System of Record directly to AI agents, with no extracts, no duplicate data, and no rebuilt security. Here's how its four layers give agents live data, shared business context, and governed actions they can be trusted with.
In this article we discuss:
At Rising this week, we announced Workday Data Cloud, a new architecture that opens Workday’s System of Record directly to AI agents. Teams no longer have to extract data, duplicate it, or rebuild security. Agents work from the same live data, business logic, and permissions that already run the business.
Data Cloud is built in four layers, each solving a different part of the problem:
Open Data Core: Unifies operational and historical data into a single foundation of over 240 pre-curated business objects. Zero-copy federation extends this data to third-party platforms, ensuring native Workday security and access controls are enforced without moving a single row.
Live Data Query: Replaces slow batch extracts with sub-second, zero-ETL queries routed simultaneously across live transactional state and the Agent Lake. This gives both human analysts and AI agents instant, unified access to current operational data..
Context Layer: Gives agents a deterministic, governed understanding of your data. By pre-packaging your codified business logic, it eliminates prompt guesswork and ensures complex metrics are calculated identically by every agent, every time.
Agent Ready Tools for Data: Equips agents to take lawful action. It provides standardized data primitives that route every execution through the Workday Agent Gateway, automatically blocking and escalating any action that exceeds policy guardrails to a human-in-the-loop.
Agent Lake and Live Data Query are generally available this week. The Context Layer's semantic models and Agent-Ready Tools for Data are in Early Access. Here's how we're rebuilding each layer and what’s new.
For decades, enterprise architecture has been split between operational systems for live transactions and analytical lakes for historical AI. Between them sat a massive, slow, and expensive ETL pipeline. Now we're shattering that wall, bringing operational and analytical data together in a single core.
Now we're shattering that wall, bringing operational and analytical data together in a single core.
Workday’s transactional system is a natively object-oriented Memory Graph. Our Transactional Memory Graph encapsulates data, calculated logic, multi-instance relationships, and effective-dated history. That matters because every foundation model on the market is remarkably fluent in language but illiterate in hierarchy and time. Our graph already understands both, because that's how it's always run your business.
On the analytical side, we are hydrating over 240 pre-curated, context-rich Business Objects (including Worker, Position, Cost Center, Purchase Order, and General Ledger Account) directly into open Apache Iceberg format in our Workday Agent Lake. Teams can federate this canonical truth directly to Google BigQuery, Salesforce Data 360, and Snowflake, which are all generally available, as well as Databricks or Amazon Sagemaker, which are in early access. This is enabled with zero copy bi-driectional federation from the System of Record.
“Workday is part of a three legged stool in our enterprise, and Data Cloud extends its value by enabling report-based integration at scale. It feels like magic because we can focus on insights and outcomes instead of worrying about compute"
David Cavanaugh
manager, data solutions
By delivering curated, canonical business objects, we provide high quality grounding for agents. Every month, we process over 500 trillion rows of data across HR and finance, and we don't stop at Workday's borders. By integrating more than 229 billion rows of external data in our Lake, we unify your organization’s entire business landscape into one queryable truth.
None of that matters, though, if security has to be rebuilt from scratch the moment the data leaves Workday. Most data lakes force exactly that, a second permission system, disconnected from the one that already governs the same data inside Workday. We do the opposite.
Workday's Contextual Security over Agent Lake scopes access based on a person's actual, specific relationship to the data: a regional HR director sees their region; a compensation partner sees their pay grade, across every region. Same object, same question—two different, equally correct answers, because the security is checking the relationship, not the role.
And now, we're extending that exact model, the one customers have already configured, to analytical and agentic workloads on the lake.
Data is useless if it’s too slow. Now, we’re eliminating slow data extracts with the General Availability of Live Data Query, the ultimate bridge. It dynamically routes every single query across our live operational state and the broad Agent Lake simultaneously—sub-second, with zero ETL. This solves a broad range of use cases including AI and BI.
Live Data Query is a distributed execution layer powered by an open-source Trino engine. It isolates the execution load completely, so a massive BI dashboard or analytical agent never competes with your core transactional throughput.
Endpoints expose standard JDBC, native BI connectors, and Model Context Protocol (MCP) interfaces, facilitating zero-ETL access to unified Workday data directly within existing enterprise architectures. This open-standard framework enforces unified security policies at the query layer while bypassing traditional batch data extraction and replication pipelines.
Business analysts can use standard JDBC connectors to run live Power BI or Tableau dashboards directly against up-to-the-second Workday transactional state under native security permissions, enabling real-time insights while eliminating brittle ETL pipelines and overnight batch delays.
Live Data Query enables sub-second, zero-ETL queries across live transactions and the Agent Lake.
Our early adopters have already put this to work: more than 1,645 live queries run across Databricks, Snowflake, Power BI, and Tableau in the first three months—two to three times faster than legacy RaaS extraction, and successfully executing queries that used to fail outright under old guardrails.
Teams often try to fix AI hallucinations by pointing models at raw query logs or cramming complex enterprise relationships into a prompt. Both approaches are set up to fail. As frontier AI researchers recently concluded, the bottleneck for analytics agents isn't access to historical data—it’s structural understanding.
The Workday Context Layer codifies 20 years of enterprise domain expertise to elevate AI from probabilistic guesswork to deterministic precision. By providing a pre-defined, deterministic model for LLMs and agents to query via MCP endpoints, the Context Layer intelligently prunes prompt context—slashing enterprise AI token consumption and eliminating hallucinated metrics.
The Workday Context Layer codifies 20 years of enterprise domain expertise to elevate AI from probabilistic guesswork to deterministic precision.
Security tells an agent what it’s allowed to see; semantics tell it what things actually mean. That's why we're opening Early Access to pre-configured Financial and HCM Semantic Models. These aren't auto-generated schemas. They package the decades of KPIs, dashboard logic, and codified business rules that already live inside the Transactional Memory Graph. Covering objects such as Headcount, Cost Centers, and Voluntary Turnover, they let agents inherit that knowledge instead of reverse-engineering it from raw tables.
At Rising this week, we made out-of-the-box Delivered Semantic Models covering common HCM and Finance objects available for Workday agent development. We extract the semantics directly from the business logic, relationships, and calculated fields already encoded in Workday's own object model—the same twenty years of domain expertise already governing how Business Objects like Worker, Position, and Cost Center relate and calculate.
Because delivered models use the same governed platform customers use to extend them, organizations can easily build customized versions. These tenanted models inherit existing security and business logic automatically.
Every agent, whether built by Workday, a customer, or a third party, now draws on one consistent, audit-ready definition of metrics like turnover or headcount.
For example, the question “what’s our regrettable attrition rate for people managers in EMEA this quarter” will now produce the same answer, every time, from any agent, in any tool. That’s because the definition of “regrettable attrition,” the org hierarchy, and the effective-dated quarter boundary already live in the semantic model, not in a prompt somebody wrote last Tuesday.
That consistency is the difference between AI you can demo and AI you can govern.
Data and context tell an agent what's true. They don't tell it what it's allowed to do—or who answers for it when something goes wrong. That's the problem this layer solves.
At Rising, we introduced the Early Availability of Workday Agent-Ready Tools for Data. Exposed via standard Model Context Protocol (MCP), these tools equip agents with:
Every agentic tool call is routed through the Workday Agent Gateway, which enforces runtime authorization, rate limiting, and business policy boundaries. Simultaneously, every agent identity, skill set, and execution log is registered within the Workday Agent System of Record (ASOR) using Agent Passport, a governance and security tool to test, verify, and continuously monitor the agents before and during deployment.
Data and context tell an agent what's true. They don't tell it what it's allowed to do—or who answers for it when something goes wrong. That's the problem this layer solves.
This means agents execute under explicit user delegation. If an autonomous agent attempts an action that exceeds designated guardrails, such as spending limits or regional policy rules, the platform automatically routes the request to a human-in-the-loop approval workflow or executes stateful transaction rollback protection.
This same governed foundation carries over to partners as well. And, a partner-built application isn't a side door into your data; it's held to the same governance as everything else running against the System of Record.
Explore the Workday Data Cloud User Guide to see how it opens the System of Record to AI agents.
More Reading
At Workday Rising, Intermountain Health’s Mike Harmer shared how Workday agents and a unified data foundation are giving caregivers time back, lowering costs, and improving patient care.
Hear from the leaders behind transformations across HR, finance, and IT about what each award means to them.
The opportunity for agentic HR, Finance and IT is so much more than efficiency—it’s about unlocking business growth and reinvention.
