SAP is buying TechWolf to tell its AI agents what your job is made of. Nobody will say for how much.

What SAP agreed to buy

On 6 October 2026, SAP SE and TechWolf announced an agreement for SAP to acquire the Ghent company outright. Terms were not disclosed. The deal is expected to close in the fourth quarter of 2026, subject to the usual conditions including regulatory approval, and J.P. Morgan acted as exclusive financial advisor to TechWolf.

TechWolf sells what it calls a context graph for work. Per SAP's announcement, it models an organization at three levels: the work itself, down to the tasks inside a job; the skills people have and actually apply; and the external labor market. It reads that out of the HR systems a customer already runs, then maps it against their strategy. The pitch is that whoever decides on hiring, reskilling and redeployment works from evidence rather than job titles.

The customer list is real and checkable: SAP names HSBC, Atlas Copco Group, GSK, Ericsson, AMD, MetLife, PayPal and Booking.com. TechWolf says its open-source models have been downloaded more than two million times. The company will stay a separate entity under co-founder and chief executive Andreas De Neve in Ghent, with offices in London, New York and San Francisco, and its platform will remain available to non-SAP customers.

Read the SAP quote again, because it is about cost

Most coverage of this deal led on human resources. The quote SAP chose to publish is about something else entirely.

"This grounding layer perfectly matches our strategy," said Manoj Swaminathan, president and chief product officer for SAP Autonomous Suite. "It makes token usage more efficient, lowers the cost of deploying workforce agents and will make Joule more intelligent in AI-driven HR scenarios such as skills-based hiring, workforce planning and role redesign."

Token usage. Cost of deploying agents. That is not the language of a talent-management purchase, it is the language of an inference bill.

The honest translation: SAP's agents, branded Joule, have to answer questions about work, and an agent with no structured model of the organization has to be fed enormous amounts of raw context to guess at an answer, expensively and unreliably. A curated graph lets it look things up instead. This is the retrieval problem we explained in how retrieval-augmented generation works, and the economics are the ones in how AI pricing and tokens work. SAP did not buy an HR feature. It bought the lookup table its agents were missing.

What a context graph for work actually contains

Worth being plain, because the phrase hides a lot.

A system that resolves work "down to the tasks inside a job" is a system that knows your role as a list of components. That is genuinely useful if you are trying to move internally and your title undersells you, and the strongest argument for this category is that titles are terrible data and skills graphs are better. It is also the input a company needs to answer which tasks are now automatable and which roles can therefore be redesigned.

SAP uses the phrase "role redesign" in its own announcement, alongside skills-based hiring and workforce planning. De Neve was blunter about the demand driving it: "Organizations everywhere are trying to figure out how AI is reshaping work, and what their workforce needs to look like because of it." He said TechWolf has spent eight years building "the evidence layer to answer those questions."

Nothing here is sinister, and TechWolf's compliance posture beats most of the category, with EU and US data residency and ISO/IEC 42001 and 27001 certifications that SAP ties to EU AI Act governance. But anyone whose employer runs SuccessFactors should understand what is being assembled: a continuously updated, task-level map of what they do, built for leaders making deployment decisions.

SAP, Workday and ServiceNow all funded this

Here is the detail that makes this a market story, not a product one.

TechWolf has raised more than $50 million in total, per Tech.eu, from Felix Capital, Notion Capital and 20VC alongside corporate investors SAP, ServiceNow and Workday. Techzine reports a $42.75 million round in 2024 in which SAP participated. So SAP has bought a company it had already backed, and in doing so has taken a supplier off the board that two of its largest competitors in enterprise software had also funded.

The entanglement runs deeper than the cap table. TechWolf's own website lists Workday among its customer logos, next to HSBC, Ericsson, GSK, Booking.com and Adobe. SAP has bought a vendor that counts its closest HR software rival as both an investor and a customer.

TechWolf says this is the largest acquisition of a venture-backed software company in Belgian history. That claim comes from the acquired company, not an independent registry, so treat it as a company statement.

The commitment that the platform stays available to non-SAP customers is the one to watch. It is standard language at announcement and it is also standard for it to quietly stop being true two product cycles later. SAP has attached a condition to it: "subject to closing and required consultation."

What is missing from the announcement

No price. No revenue figure. No customer count. And one sentence doing real work with nothing behind it: SAP says joint customers "are seeing measurable results from the partnership," then measures none of them. For a grounding layer sold on cutting agent costs, one before-and-after number would have been the most persuasive thing in the release.

The products do not exist yet either. SAP says it and TechWolf will "jointly co-innovate a concrete set of new AI-powered workforce and skills optimization products to be designed and developed post-close." Designed and developed after a deal that has not closed.

One last note, included in the release itself and worth repeating without snark: SAP states the press release "was created with the assistance of artificial intelligence and has been reviewed by the editorial team."

Our take

This is a sound acquisition and a widely misread one. The tell is SAP's own quote: the scarce thing in enterprise AI is not a model, it is structured context an agent can query cheaply, and a skills graph built from eight years of customer systems is hard to replicate. Buying it from your own portfolio, out from under Workday and ServiceNow, is a good day's work. Whether it pays off is unevaluable, because SAP disclosed no price, no results and no products, and the only quantified claim in the release is a download count. SAP HR customers see nothing change before Q4 2026 at the earliest. If you work somewhere running SuccessFactors, the thing to notice is not the deal but what is being modeled: your job, broken into tasks, kept current.