Workday announced on 18 November that it intends to acquire Honeydew, a semantic layer for AI and business intelligence. The company says Honeydew gives AI agents and BI tools the same governed business definitions, so an answer stays consistent whichever tool or model asks the question. The transaction is expected to close in the fourth quarter of Workday’s fiscal year 2027, which ends on 31 January 2027, subject to customary closing conditions.

What Workday announced at Rising EMEA

The announcement came from Workday Rising EMEA in Barcelona, in a press release. Workday describes itself there as the agentic platform for HR, finance and IT. It says Honeydew’s semantic compiler, which turns business definitions into governed queries, will work together with Workday Data Cloud so that agents get the business context they need to act quickly and consistently.

The release uses a headcount example to explain the problem. A dashboard, a chat experience and an AI agent can each answer the question “what’s our headcount?” differently. One counts only full-time employees, another includes contractors, and a third counts approved-but-unfilled roles as already staffed. Workday’s point is that all three numbers can be defensible and still disagree. Honeydew, according to the release, captures multiple definitions and pulls the right one based on the person or agent asking, the use case and the situational context.

Advertisement

HRTech Your brand belongs here. Reach the decision-makers who read HRTech every day. Premium placements across the site and newsletter. Advertise with us

How Workday says the semantic layer would work

Workday says that without a native semantic layer, connecting an AI agent to enterprise data takes a large amount of manual work. Data teams spend months writing custom code to intercept requests to large language models, hardcode business rules into data pipelines and pack data dictionaries and join rules into every prompt, a practice the release calls “schema-stuffing”. It adds that a custom finance metric can take weeks to get right and often has to be rebuilt by hand in a tool such as Tableau.

With Honeydew, Workday says it will move business logic out of the prompt and compile it into deterministic queries before a request reaches the model. In the release’s words, the AI gets the question and the right path to the answer, rather than the whole handbook. Workday expects customers to see greater accuracy and efficiency, fewer tokens and lower costs, with definitions intact across whichever tools or models they choose.

Gabe Monroy, chief technology officer at Workday, put the payroll case directly. “An AI model shouldn’t be improvising the calculation behind a payroll number,” he said. “With Honeydew, the model says what it needs and Workday builds the exact query, the same way every time, within each person’s or agent’s permissions.”

David Krakov, CEO of Honeydew, framed the product around definitions. “Getting AI to write database queries was never the hard part,” he said. “Getting it to understand the true definitions behind your company’s data is.”

Live data and permissions

Two other claims in the release matter to anyone who owns people data. First, Workday says most semantic layers sit on data already extracted into a warehouse or lake, so answers are only as current as the last refresh. Honeydew would instead apply a customer’s definitions directly to live data in Workday Data Cloud. Second, Workday says its permissions would travel with the definitions, so each person or agent sees only what they are allowed to see, with no extracts to manage. Per the release, invalid or unauthorized requests would be blocked before any query runs.

Yasmeen Ahmad, senior vice president and general manager of Workday Data Cloud, said that “the greatest bottleneck in enterprise AI today isn’t reasoning, it’s context.” Workday describes its agents as governed in identity, verification, access and behavior, and says Honeydew would add governance at the moment a question becomes a query.

Newsletter

Get the week's best tech coverage.

Free. Read by thousands of HR, tech, and business leaders.

Where this fits with Workday’s recent HR announcements

The deal arrives alongside other agent and payroll moves from Workday this autumn. In October the company announced Autopilot agents for hiring and payroll with customer-set limits, and at the same Rising EMEA event it said it would bring native payroll to seven more markets. We also compared four vendors’ cases for trusting AI agents with pay data. A semantic layer addresses a neighbouring question, which is whether every agent calculates a figure from the same definition.

In our reading, the announcement matters most to HR teams that already report the same metric from more than one place. Headcount, attrition, time to fill and cost per hire often differ between the HRIS report, the finance view and the dashboard a business partner built. Workday’s release says a shared governed definition is the fix for agents. The same logic applies to the people who read the reports today.

What it means for the HR leader

The release is an announcement of intent. The transaction has not closed, and Workday’s statements about accuracy, token savings and lower costs are expectations the company has stated, with no customer data attached. Honeydew’s capabilities would reach customers only after the deal closes and Workday integrates the product into Workday Data Cloud, and the release gives no product date.

HR leaders who run Workday can use the time before then. Write down the definitions your team uses for the five or six metrics leadership asks about most, and note who owns each one. Check where the HRIS, finance and your BI tool currently disagree and why. When Workday gives detail on availability, ask for three things in writing: which definitions you can store and how conflicts between them are resolved, how permissions carry over to agents and BI tools, and what runs against live data and what still runs against extracts. Teams on other platforms can ask their own vendors the same questions.

Source: Workday, press release, 18 November 2026