Workday seeks to buy Honeydew to give AI agents consistent business definitions
The planned integration would turn business definitions into controlled queries on live data. Workday promises greater accuracy and lower costs, but the acquisition remains pending.
Listen to this story
The audio brief
Story brief
3 key pointsThe planned acquisition would add Honeydew’s semantic layer to Workday Data Cloud, translating business definitions into deterministic database queries instead of leaving models to interpret metrics through prompts or custom code. Workday says definitions could vary by requester and context, with permissions checked before queries run against live data. The companies have not disclosed financial terms, and the integration is not yet available: Workday expects the deal to close in its fiscal fourth quarter of 2027, subject to customary conditions.
- 01
Honeydew’s semantic compiler is intended to translate business rules into fixed queries before an AI model receives the results.
- 02
Workday says the system would apply definitions to live Data Cloud information, rather than rely on warehouse extracts whose freshness depends on update timing.
- 03
Workday says unauthorized or invalid requests would be stopped before a database query runs; this would complement its existing agent controls.
Workday wants its AI agents and business dashboards to agree on what the numbers mean. On November 18, 2026, it announced its intent to acquire Honeydew at Workday Rising EMEA in Barcelona. The planned integration would give tools shared business definitions, rather than leave each system to interpret a question on its own.
The same question, different calculations
Consider a headcount request. A dashboard might count full-time employees, a chatbot might include contractors, and an agent might also count approved but unfilled jobs. Those answers reflect different definitions, not necessarily different arithmetic. The problem is deciding which definition belongs to the question before calculating the result.
Honeydew supplies a semantic layer: software that connects business terms to the rules used to query data. Teams can define a metric for reuse across business intelligence tools and AI agents. That makes the definition shared infrastructure rather than something each tool must separately recreate.
Shared does not mean every request must use one universal definition. Workday says Honeydew can capture multiple meanings and select the applicable one based on who is asking, the use case and the surrounding context. The aim is consistency for the relevant situation, rather than forcing every headcount question into the same calculation.
Business rules outside the prompt
Workday contrasts that approach with packing business rules into AI prompts and custom integration code. It says data teams can spend months adding data dictionaries, definitions and rules for joining database tables to model requests. Changing models or platforms can mean rebuilding that work; moving a custom finance metric to another analytics tool can require manual reconstruction, without guaranteeing that the resulting numbers match.
The planned alternative is a semantic compiler, which translates those definitions into deterministic queries: database instructions built to follow fixed rules. Workday intends to combine Honeydew’s compiler with Workday Data Cloud, moving business logic out of prompts and compiling it before a request reaches the model.
In Workday’s description, the model specifies what information it needs, while the platform builds the query within the requester’s permissions. The company says customers could get more accurate and efficient answers, use fewer tokens—the units models process—and pay less. Those are expected benefits of the planned system, not demonstrated results from the combined platform.
An AI model shouldn’t be improvising the calculation behind a payroll number.
Gabe Monroy, Workday chief technology officer
Live data, with checks before execution
The integration would also change where those definitions operate. Honeydew describes its existing approach as SQL-based and centered on the data warehouse. Workday plans to apply definitions directly to live data in Workday Data Cloud, rather than rely on an extracted copy whose freshness depends on its last update.
Permissions would accompany the definitions. Workday says invalid or unauthorized requests would be stopped before a query runs, rather than flagged afterward. Its existing agent controls cover identity, verification, access and behavior; Honeydew would add a check at the point where a business question becomes a database query.
The proposed addition targets Workday’s platform for human resources, finance and information technology. Workday says more than 11,500 organizations use its software worldwide, including more than 65% of the Fortune 500.
Financial terms were not disclosed. Workday expects the acquisition to close in its fiscal fourth quarter of 2027, ending January 31, 2027, subject to customary closing conditions. The purchase and Data Cloud integration remain plans, not a completed transaction or an announcement that the combined system is available.
Sources
- tbreak.comWorkday to buy Honeydew, a semantic layer for AI agents
- securities.ioWorkday Plans Honeydew Acquisition to Standardize AI Data Queries
Reader comments
Newest comments first. Replies stay oldest first.