Just this week, SAP published a new API policy that's already generating significant pushback from customers, partners, and the broader SAP community. And one thing in the policy is hard to miss: it explicitly singles out AI. SAP now prohibits API use for "interaction or integration with (semi-) autonomous or generative AI systems that plan, select, or execute sequences of API calls," as well as "scraping, harvesting, or systematic and/or large-scale data extraction or replication," except through SAP-controlled architectures and pathways.

In other words, if you want to use AI agents to access SAP data directly via APIs, SAP wants that to happen on SAP’s terms. 

The reaction was swift enough that SAP has already updated the document and issued an FAQ to address some of the concerns. CEO Christian Klein, on last week's earnings call, reassured customers and partners that SAP wants an open platform and that no one should worry. But the language in the policy itself hasn't materially softened. 

At a moment when every enterprise is racing to figure out how to use AI in a cost-effective and performant way, it’s the AI clause that should give every data leader pause. 

[CTA_MODULE]

What this means for Fivetran customers

First, the good news for our customers: Fivetran's SAP connectors are not affected. You can keep moving your SAP data wherever you need it, without disruption. It's also worth being clear about what this policy actually changes today. This is a policy, not a contract, so for now, enforcement is limited. 

The main thing SAP is acting on is one specific interface (ODP RFC), which they'll start blocking in July, with some exceptions running through the end of the year. So the immediate impact for most SAP customers may be minimal. But the long-term effects on data infrastructure, and especially AI strategy, are what’s worth looking at, particularly given SAP can continue to change and restrict access to your data on their platform.

The bigger story is the direction SAP is heading, and that's the part to watch.

Why the direction matters

The AI restriction is the most pressing part of the policy, but it’s part of something bigger. The broader stance is that integration of SAP data with non-SAP systems should flow through SAP-controlled paths. 

That's a position that works well for SAP, but it puts an important question in front of every data team: do you want your integration architecture, and the AI strategy that depends on it, defined by your ERP vendor's commercial roadmap or by your own?

The trade-offs tend to become clear only after you’re committed:

• Cost stacking: Customers often end up paying for multiple SAP data products, and when an external compute engine joins the architecture, the costs compound quickly.
• Restricted cloud flexibility: Even when your environment runs on AWS or Azure, you're limited to the services your vendor has certified, not the full native catalog you've already vetted and are paying for.
• Architectural lock-in: Your business logic, security model, and semantic layers stay tied to the vendor. Moving primary analytics elsewhere later means rebuilding that work from scratch.
• AI on someone else’s terms: If your AI agents can only access your data through vendor-defined pathways, the pace and direction of your AI strategy gets tied to the pace and direction of theirs. 

This is why an Open Data Infrastructure matters

Our position at Fivetran is straightforward: your data belongs to you, not to your vendor. And the AI tools you choose to build on top of it should be your decision too. That's the premise behind Open Data Infrastructure (ODI): an architecture where storage, compute, and integration are decoupled, and open formats keep your options open.

In practice, ODI means:

• You own your data. It lands in a lake or warehouse you fully control. No vendor IP claims on your data products.
• Decoupled storage and compute. Store once, centrally. Use whichever compute engine best fits the workload.
• Open formats, real portability. Iceberg and Delta Lake open table formats let you use the best BI, ML, and AI tools available today, without waiting for any vendor to certify them.
• No artificial blockers. Data flows where your business needs it, including to the AI systems you choose.

The companies that will move fastest on AI over the next year will be the ones whose data is portable, well-governed, and free of vendor restrictions on where it can go or how it can be used.

Final thoughts

For Fivetran customers running SAP today: nothing changes. Your pipelines and models run, your data lands where you want it, and we'll continue to advocate on your behalf as this policy evolves.

The architecture you have today was built for the work you were doing 5 years ago. The work you'll be doing a year from now — agents reasoning across systems, models trained on data you didn't know you'd need, decisions made faster than any human can support — will ask more of it. Maybe the architecture you have is ready for that. Maybe it isn't. That’s the question worth answering before someone else answers it for you.

[CTA_MODULE]