Workday said on November 18, 2026 that it intends to acquire Honeydew, a Tel Aviv startup whose software keeps business definitions consistent across AI agents and business intelligence tools. The announcement came at Workday Rising EMEA, the company's conference in Barcelona. Neither side disclosed the price or other terms.
The deal has not closed yet. Workday expects it to complete in the fourth quarter of its fiscal year 2027, which ends on January 31, 2027, once customary closing conditions have been met. Honeydew's co-founder and chief executive, David Krakov, wrote on the company blog that the team plans to join Workday subject to those conditions.
Workday intends to build Honeydew into Workday Data Cloud, the part of its platform where AI agents work with customers' HR and finance data. The announcement did not say what will happen to Honeydew's standalone product, its current customers or its staff after the deal closes.
What Honeydew built
Honeydew makes what the industry calls a semantic layer: a single place where a company writes down how each of its metrics is calculated, so every tool that asks for the number gets the same answer. Workday illustrated the problem with headcount. A dashboard might count only full-time staff, a chat assistant might add contractors, and an AI agent might treat approved but unfilled roles as filled, producing three different figures for one question.
Rather than forcing one definition, Honeydew stores several versions of a metric and picks the one that fits the person or agent asking, the task and the context, according to Workday. At its core is what the company calls a semantic compiler, which turns those definitions into governed database queries.
When Honeydew presented itself in Y Combinator's Winter 2023 batch, it pitched the product to data teams working in the data warehouse. It connected to Snowflake and let users apply the same logic from SQL, Tableau, Looker, Python or dbt. The founders said at the time that they had spoken with more than 300 analytics teams and that early users included unicorn companies in cybersecurity, insurance technology and cloud operations.
“Getting AI to write database queries was never the hard part. Getting it to understand the true definitions behind your company's data is. We built Honeydew to solve that problem once, rather than in every tool and every prompt. With Workday, we plan to bring that shared understanding to every agent working with a company's most important data.”
Why Workday is buying
Workday argued that linking AI agents to company data is slow and costly without a semantic layer. It said data teams now spend months writing code to intercept requests to large language models, hard-coding business rules into data pipelines and packing data dictionaries and SQL join rules into every prompt, a practice it called schema-stuffing. That work has to be redone when a team changes model or platform.
With Honeydew, Workday said, the business logic would be compiled into fixed, repeatable queries before a request reaches the AI model. The company told customers to expect better accuracy, fewer tokens and lower costs. It also said Honeydew would run against live data in Workday Data Cloud rather than against copies extracted into a warehouse, which most semantic layers rely on.
Permissions are the other part of the pitch. Workday said its access rules would travel with each definition, so a person or agent would only see data it is allowed to see, and requests that are invalid or unauthorised would be stopped before a query runs. Gabe Monroy, Workday's chief technology officer, said a model should not be making up the calculation behind a payroll figure. Yasmeen Ahmad, who runs Workday Data Cloud, said the main obstacle for enterprise AI was context rather than reasoning, and that Honeydew would give agents what she called a living semantic graph.
The founders and the funding
Honeydew was founded in 2022 and lists its location as Tel Aviv-Yafo, according to its Y Combinator profile, which gives a team size of six. Krakov described the team in his blog post as small.
Krakov has spent about 20 years in data and software. He previously founded Varada, which built an acceleration engine for data lakes and was acquired by Starburst, and before that was an early employee at XtremIO, later part of EMC. Co-founder Baruch Oxman has more than 20 years in software; he was the first employee of Implisit, an AI-for-sales company bought by Salesforce, where he went on to lead architecture and engineering for Einstein for Sales. The third co-founder, Liron Slonimsky-Shemy, was a screenwriter before founding Awear, which sold data to fashion companies on how people wear their clothes.
Honeydew went through Y Combinator but has not disclosed how much money it raised. Krakov thanked the company's investors in his post without naming them, and no other backers were named in the sources reviewed.
What it means for Workday customers
Workday describes itself as an agentic platform for HR, finance and IT and says more than 11,500 organisations use it, including over 65% of the Fortune 500. Krakov wrote that Workday holds the data where errors matter most, people and money, and contrasted a weak email subject line, which costs a few clicks, with a payroll mistake that could cost someone their rent.
Workday's release cautioned that features not yet available may change or may not be delivered, and listed the integration of Honeydew's operations and technology among the risks of the deal. Dealroom noted that other software groups have also been buying tools for AI agents, including Snowflake, which agreed in May 2026 to acquire Natoma, a provider of identity controls for agents.




