SAP just closed its ERP to third-party AI – here’s what we’re building instead
SAP recently restricted third-party AI agents from its platform, locking ERP customers into its own AI tool. Here’s what the policy actually restricts, why it hits smaller businesses hardest, and how Katana keeps AI inventory management open for every customer.
Riikka Söderlund

A significant change happened in the ERP world this week. SAP, the ERP behind hundreds of thousands of businesses, updated its API Terms of Use this week to prohibit third-party AI agents from interacting with its platform — a change with direct consequences for how customers can build and run AI inventory management workflows.
The change didn’t come with a big press release, but the implications are worth paying attention to.
| SAP API Policy — Exact language |
|---|
| “Except through and within the limits of SAP-endorsed architectures, data services, or service-specific pathways expressly identified and intended for such purposes, SAP prohibits API use for: (a) interaction or integration with (semi-)autonomous or generative AI systems that plan, select, or execute sequences of API calls, and (b) scraping, harvesting, or systematic and/or large-scale data extraction or replication.” |
In plain terms: if you want to connect a third-party AI agent or any
autonomous tool that reads your SAP data, reasons over it, and takes action, you can’t.
SAP has unilaterally decided that the AI systems talking to your operational data must be theirs. Specifically, Joule, SAP’s own AI assistant.
What SAP’s new policy restricts
SAP’s official reasoning leans on security and compliance. Autonomous AI agents interacting with APIs at scale introduce risks that SAP can’t control. That concern is legitimate, and it’s not entirely wrong. Poorly-governed agents hammering enterprise APIs can cause genuine problems.
The restriction goes further than that concern warrants, though. Rather than targeting rogue or unapproved agents specifically, it blocks all third-party AI agent integration. Whether this reads as a security measure or a competitive moat depends on your perspective, but the outcome is the same either way.
SAP has invested heavily in Joule, its native AI copilot. Opening their API to GPT-based agents, Claude, or any autonomous workflow tool from the thousands of ISVs in their ecosystem would compete directly with that investment, and this policy protects it. That’s a rational business decision, just not a customer-first one.
The large legacy ERPs cannot replicate an open, extensible strategy without cannibalising themselves. Their entire value proposition is predicated on feature breadth. That’s an asymmetric opportunity for everyone they’ve locked in.
Why small and mid-sized businesses feel this most
Enterprise businesses with seven-figure SAP contracts have options. They have in-
house developers, vendor management teams, and enough purchasing power to
negotiate carve-outs. They’ll find a way around this, or simply absorb Joule and move
on.
SMBs don’t have that cushion. For a growing manufacturer, a multi-channel brand, or
an inventory-intensive D2C business running on SAP Business One or SAP Business
ByDesign, this policy creates a very specific and painful reality:
- AI tools you chose because they fit your workflow are now locked out.
- Replenishment workflows built on Claude or similar agents, pulling from SAP data, are in violation.
- Third-party demand forecasting tools connected via API are in the same position.
- Integration partners who spent months building AI-enabled connectors are back to square one — and the problem extends beyond new builds. Existing integrations are at risk too.
Worse, SMBs are the segment with the most to gain from AI agents. They don’t have teams of analysts running queries and building dashboards. They have one or two people wearing six hats, trying to make faster decisions with incomplete information. AI agents help a small team work like a bigger one, and this policy removes that option for the businesses that need it most.
Why open ERP ecosystems have the advantage
The data in your ERP is a record of your business — your products, suppliers, customers, operations. You generated it, and you should be able to use it with whatever tools make sense for how you work. When a software vendor decides which AI is allowed to touch that data, it changes the nature of the relationship in a way most businesses didn’t agree to when they signed up.
There’s also a strategic case worth making here. The business software companies that have grown most durably tend to be the ones that made it easy to build on top of them. Shopify’s power comes as much from 10,000 apps as it does from its own checkout. Salesforce published APIs and let partners build; Atlassian and HubSpot built moats through their ecosystems. Restricting who can integrate is a short-term play that tends to slow a platform down over time as the most innovative builders move toward more open ground.
SAP’s ecosystem is built on ISVs, implementation partners, and builders who have spent years developing integrations on SAP APIs. This policy signals to all of them that the terms can change unilaterally in ways that break what they’ve built — and that’s a difficult signal to ignore when deciding where to invest. The most innovative builders will start hedging toward platforms that don’t do that to them.
SAP vs. Katana: how the two API approaches compare
| Policy area | SAP | Katana |
|---|---|---|
| Third-party AI agents | Prohibited | Fully permitted |
| API access tier | Enterprise contracts only; tiered restrictions apply | Included in every plan, including the free plan |
| Who controls AI integration | SAP (Joule only for autonomous workflows) | You. Bring whatever AI you want. |
| Building on the platform | Restricted ecosystem; terms can change unilaterally | Open: your developers, your partners, and customers themselves can build on Katana |
| Data philosophy | Vendor controls access layer | Your data, your rules |
Riikka Söderlund
How Katana approaches AI inventory management
We built Katana’s API to be open from the start. The future of software is an open system that lets you connect the tools you already trust.
Our API is available to every customer, including the free plan. Want to connect Claude, GPT, Gemini, or a custom agent your developer built on a Tuesday afternoon to your inventory data? Go ahead.
We’re also building toward Katana becoming a platform — one where solution engineers, implementation partners, and eventually customers themselves can extend it to fit their workflows without waiting for us to build every feature they need. That means custom fields, community-built solutions, and an extensibility layer that lets the platform grow beyond what we build ourselves. The goal is a system of record you can build on top of.
Your operational data — inventory positions, supplier lead times, sales orders across every channel — sits in one place, accessible to the AI tools you choose to work with. The right setup means those tools can handle AI inventory management automatically: adjusting reorder points, running replenishment, and surfacing decisions without someone manually triggering each step.
SAP is betting that walls protect their relevance. We’re betting that openness creates it.
Table of contents
Get inventory trends, news, and tips every month
Get visibility over your sales and stock
Wave goodbye to uncertainty with Katana Cloud Inventory — AI-powered for total inventory control