On July 6, SAP announced the completion of its acquisition of Dremio, a data lakehouse company. The acquisition amount was not disclosed, but the significance of this news lies not in the price—the data foundation of ERP is undergoing a transformation: from "recording business" to "enabling AI to understand business." On the same day, SAP also published an internal article titled "Fall in Love with the Problem, Not the Solution," discussing why AI projects should not start from a technical perspective. The next day (July 7), SAP officially announced new capabilities for People Intelligence on Business Data Cloud.
Within three days, acquisitions, concepts, and products were launched in succession. The signal SAP is sending to the market is very straightforward: the data foundation of ERP is no longer just a place to store data; it is becoming the infrastructure that enables AI to run.
What exactly did Dremio add
What does Dremio do? Simply put, it is a data lakehouse platform that allows enterprises to read data scattered across different systems and databases using a single query engine, without needing to move data or perform format conversions. It natively supports open table formats like Apache Iceberg — meaning data won't be locked in if you switch vendors.
SAP's official press release put it diplomatically: acquiring Dremio "accelerates agentic AI" and enables customers to "combine SAP and non-SAP data in real time without data movement or transformation." In plain terms: SAP's AI agents (Joule and those 200+ agents) previously mainly ran on SAP's own data, but now they can directly cross systems to understand the entire enterprise data landscape.
Three capabilities Dremio brings to SAP Business Data Cloud:
1. Open Data Format (Apache Iceberg) – Data is Not Locked
2. Unified Query Engine – SQL Acceleration Across Data Lakes and Databases
3. Enterprise-level data governance - lineage tracking and access control
A friend who has been implementing ERP for over a decade told me after reading this news: "In the past, when working on SAP projects, the most headache-inducing part was data integration with non-SAP systems—every time it was custom development, every time it was endless bickering. If Dremio's capability can truly be realized, it would turn 'connecting data' from a project-level task into a platform-level one."
This statement is a bit idealistic, but the direction is correct.
People Intelligence: SAP is using it first
The People Intelligence, released on July 7, is a product developed by SAP's internal IT department as "Customer Zero." SAP's People Analytics team had previously been working on something called MyTeam Dashboard—a small HR dashboard that displays each manager's team data, including birthdays, compensation, performance, and so on. This dashboard is the most frequently used report within SAP, but the problem lies here: because it is so popular and demands are too high, the team is always rejecting requests and always dealing with backlogs.
Their solution is to shift the focus from the "consumption layer" (dashboard) to the "data layer": instead of creating customized reports one by one, they build Data Products on SAP Business Data Cloud, allowing business departments to conduct their own analysis based on these data products. People Intelligence is a pre-built catalog of these data products—just the Workforce Composition Insights module alone comes with 69 pre-built data products, encoding hundreds of join logics. SAP's official blog states, "This level of complexity cannot even be reliably replicated by AI-assisted modeling tools at present."
I think this judgment is quite practical. Many CIOs have told me that they tried to let AI directly analyze HR data, and the AI gave very polished answers, but the data was wrong—because the underlying data model hadn't been governed properly. SAP's approach is to govern the data first, then let AI run on top of it—the order is correct.
More critically, SAP has clearly stated that in November 2026, it will release the Joule Assistant and Joule Agent based on People Intelligence. AI agents will be able to directly query employee data, compensation data, and skills data on well-governed data products. In the words of SAP's internal IT leader Oliver Huth: "Investing in a data product strategy is the first step toward making AI trustworthy."
Three news stories linked together
These three SAP news items—the completion of the Dremio acquisition, the launch of People Intelligence, and the article "Fall in Love with the Problem"—when viewed together, tell the same story: the AI competition among ERP vendors has shifted its focus from the model layer to the data foundation.
Let me translate what this means. From 2025 to the first half of 2026, all ERP vendors are boasting about how many AI Agents they have and which large models they've integrated. But the most pressing bottleneck right now is not model capability—there are plenty of models, and they're getting cheaper. The bottleneck is: Is your ERP data organized? Can your non-SAP data and SAP data be connected? Can your AI Agent access governed, trustworthy data?
SAP's answer was "no" — so it spent months buying Dremio, building Business Data Cloud, and launching People Intelligence. This three-layer structure is clear:
| Level | ability | Corresponding action |
|---|---|---|
| Data Access | Real-time integration of SAP and non-SAP data with zero data migration | Dremio acquisition completed (July 6) |
| Data Governance | Pre-built data product layer, data lineage + access control | People Intelligence goes live (July 7) |
| AI execution | Joule Assistant/Agent, natural language query governance data | November release |
In contrast, Odoo takes a different approach — it uses the MCP protocol to enable AI to directly interact with ERP data, following an "open interface" path. Odoo 20 natively integrates an MCP Server, requiring no middleware, allowing its 2 million paying users to directly connect AI workflows via MCP. One builds a data foundation from within the platform, the other opens data capabilities at the interface level — different paths, but the same direction: enabling AI to understand ERP data.
Practical Impact on CIOs
What does the SAP+Dremio combination mean for existing SAP customers? Let me highlight the three most practical aspects.
Data integration costs will decrease. Previously, moving data between SAP and non-SAP systems required either ETL or custom development. If Dremio's "zero data migration" query engine is properly implemented, this repetitive development work can be reduced. Of course, the prerequisite is that your organization's data governance is in place — this is not just a technical issue.
The data preparation cycle for AI projects will be shortened. People Intelligence comes with 69 pre-built HR data products. If this model expands to finance, supply chain, and manufacturing (which is highly likely), enterprise AI data governance will shift from "starting from scratch" to "starting from 80%."
The retirement of OpenAI's GPT (GPT-4.5 removed from ChatGPT on June 27) and SAP's acquisition of Dremio occurring in the same week is no coincidence. Models may be replaced, retired, or have their prices raised—but your enterprise data will not be replaced. The value of a data foundation becomes more prominent as the number of models increases.
Finally, let me share a point from SAP's article "Fall in Love with the Problem" — which I believe was the most valuable sentence over the three days:
"The biggest risk of an AI project is not that the technology fails, but that the solution you fall in love with is not solving the right problem at all."
After the data foundation is built and the AI Agent is connected, what problems does your enterprise actually need to solve? Thinking about this in reverse might be the most worthwhile thing to spend time on right now.
