How open ERP APIs and data speed up agentic AI integration, deployment

ERP systems are rich stores of the data that AI needs to learn over time, understand business contexts and conduct transactions on behalf of users. But giving AI access to that data is hardly a simple matter.

Vendors are struggling not just with how to make their ERP data available to generative and agentic AI, but what kinds of controls to put on that availability. For example, SAP was roundly criticized in April for issuing a policy requiring third-party AI agents that use APIs to access data in SAP systems to do so in architectures approved by SAP.

Now, ERP vendors are jostling to show their data is more open than the next guy's. One of them is Infor, a major player known for the specialized "micro-vertical" ERP products in its CloudSuite brand, and for its history of acquiring dozens of on-premises ERP vendors.

In this episode of Enterprise Apps Unpacked, Infor's Max Fisher compares his company's approach to AI data with SAP's and explains how Infor's underlying architecture supports deploying AI in ERP.

How open is open?

Fisher is director of product strategy for Infor Velocity Suite, the vendor's package of AI agents, apps, tools and consulting services for deploying AI in CloudSuite applications.

While reluctant to judge SAP, except to call its API policy and "walled off" data access a strategic decision, Fisher touted the benefits of openness.

"The more open we can be, the better experience it is for the end customer, whether they're able to deliver it themselves and develop on that platform and innovate quickly, or our partners are, or we are." He said being able to connect to third parties on-premises, as well as to multi-tenant SaaS products, helps build a developer ecosystem.

Fisher said third parties can access 6,000 published APIs through Infor's API gateway. Customers with a cloud data warehouse from Databricks or Snowflake -- with which SAP has significant partnerships -- can integrate with Infor's platform. CloudSuite applications publish their data to Infor's data lake. From there, it can be streamed in real time to the third-party warehouses. Infor will soon release its own data warehouse, he said.

Fisher gave an example of a healthcare customer whose Microsoft Copilot assistant communicates with Infor agents to help users decide how to distribute cash to optimize early payment discounts. In some cases, agents can execute the payments. The math would take hours to calculate in a spreadsheet, he said.

"Having that agent-to-agent ability for them to talk to each other is key, and that's how we keep it open," Fisher said. "We don't want to see customers take another system and try to build agents on top of our ecosystem because we've already built them. Let your agents talk to ours."

Other topics discussed in the podcast include the following:

• How on-premises ERP systems can access AI in the cloud.
• Whether Model Context Protocol is an adequate standard for multi-agent orchestration.
• How data in Infor's industry-specific ERP systems provides context for AI.

David Essex is a senior editor who creates in-depth content on enterprise applications, emerging technology and market trends for TechTarget's CIO group.