SAP agreed on October 6 to acquire TechWolf, a Belgian company that maintains a continuously updated graph of how work is actually structured inside an organization. Terms are undisclosed and the deal hasn't closed. The strategic logic is readable anyway.
Read this one alongside a second acquisition. A week earlier, Salesforce signed to acquire Listen Labs, which is a different kind of context asset built for a different kind of agent. The second piece below covers that deal and draws the comparison. This one covers what SAP is buying.
What Joule knows now. SAP's Joule agents in SuccessFactors currently reason from HR system records: performance goals, learning paths, career development plans, succession candidates. Several of the more ambitious capabilities, including recruiting, workforce planning, and skills governance, are still in early-adopter or roadmap status. Joule knows what SuccessFactors contains. It has no independent view of what a software engineer does all day, as shown by their Jira commits, or of how that engineer's real capabilities compare to what the external labor market pays for similar work. TechWolf supplies that missing layer.
What TechWolf built. The product pulls from two data environments that most HR platforms keep separate: the system of record (SuccessFactors, Workday, Oracle) and the systems where work happens (Jira, Confluence, ServiceNow, Salesforce, Google Drive, Asana). From the combined evidence, TechWolf infers employee skill profiles, maps tasks to roles, scores people against open positions, identifies skill gaps, and benchmarks internal roles against a labor-market data lake covering 1.5 billion job postings over nine years.
The data model has three layers: a shared vocabulary of canonical skill names, a centrally maintained ontology (35,000+ skills with adjacency and transfer relationships), and a customer-controlled taxonomy that maps those skills into the organization's own language. The API exposes structured entities (employees, jobs, tasks, vacancies, courses, org units) and returns match scores and gap analyses. An agent can query the graph directly: who in this org has skills adjacent to this open role, what's the gap, what learning would close it.
Pay attention to one design choice. TechWolf's skill attributions are inferred, not declared. The system proposes profiles from work evidence, and employees and managers validate, reject, or modify them through Teams or Slack. The graph is a running inference with a human correction loop attached, and that correction surface is how the system earns organizational trust over time.
SAP already knew this asset. TechWolf was on SAP Store as of January 2025, feeding skills data into SuccessFactors' Talent Intelligence Hub. SAP participated in TechWolf's $42.75M Series B in June 2024 alongside Workday Ventures and ServiceNow Ventures. What SAP is doing now is converting a proven integration and an existing stake into full ownership.
It has done this before. Over the past decade, Qualtrics, Signavio, and WalkMe all followed the same logic: buy a specialized product that generates data SAP's core platform doesn't produce, then integrate that data as a grounding layer for SAP's own workflows. The recurring risk is absorption, where the acquired product gets folded into SAP's platform in ways that limit its independent value. Qualtrics was eventually spun back out. Where TechWolf lands depends on how tightly SAP couples it to SuccessFactors versus keeping it as a multi-platform data layer.
Competitive blocking or capability acquisition. TechWolf wasn't exclusive to SAP. Workday announced in January 2025 that it was deploying TechWolf across its own workforce and listing it in Workday Marketplace. TechWolf also advertised Oracle integration. SAP says TechWolf will remain available to non-SAP customers after closing. Nothing public establishes API parity, pricing, roadmap priority, or data-access guarantees for those deployments.
Track what happens to TechWolf's non-SAP integrations in the first two quarters. That will tell you whether SAP bought a capability or bought a blockade.
If Workday and Oracle lose access or get degraded service, SAP has also pulled a skills graph away from competitors who were already relying on it. I can't tell from public terms which way SAP will go; treat either outcome as speculative for now. The workforce-AI landscape shifts regardless. If SAP keeps TechWolf open, it becomes the rare acquirer running a neutral data layer inside competitors' platforms. If it doesn't, Workday and Oracle face a build-or-buy decision on skills intelligence right when their own agent strategies need that grounding. Meanwhile, every company that built on TechWolf as an integration partner now buys from a competitor.
Where the graph goes wrong. An organizational skills graph fails in a specific way: it mischaracterizes what people do. Skills inferred from Jira tickets and Confluence pages are probabilistic. In the Belgian government deployment, employees reviewed their inferred profiles directly, which is the right correction mechanism. Once Joule starts making succession or workforce-planning recommendations from TechWolf's graph, though, the accuracy and freshness of that validation loop becomes the trust surface for the whole system. An agent that recommends a succession candidate on stale skill data is worse than one that says it doesn't have enough information to recommend anyone.
