Summary

• SAP is targeting 1,000 partner-developed AI agents ahead of TechEd after partners created more than 680 before Sapphire in May.
• More than 120 partner agents are published on SAP Discovery Center, with over 300 forecast by year-end; listings are not equivalent to live deployments.
• SAP's existing €100m investment programme, AI Agent Hub, and partner validation requirements support the move towards customer implementation.

SAP is pushing its technology partners to move beyond producing large numbers of artificial intelligence agents and demonstrate that the software is being deployed successfully in customer businesses, as its ecosystem approaches a target of 1,000 partner-built agents.

The German enterprise software group says partners created more than 680 agents on SAP Business AI Platform before its Sapphire conference in May, exceeding an initial target of between 100 and 200. It has since set a new goal of 1,000 ahead of SAP TechEd, although the latest programme places greater emphasis on operational use and measurable results.

More than 120 partner-developed agents are published on SAP Discovery Center, with the company forecasting that the total will exceed 300 by the end of 2026. These are different measures of activity: an agent that has been developed is not necessarily listed for customers, while a listing does not establish successful deployment.

SAP is supporting the initiative through a €100 million partner investment fund announced earlier this year, alongside agent management infrastructure and additional validation requirements. The latest announcement concerns progress towards customer adoption rather than a new launch of the funding programme.

Building software is only the first stage

Enterprise AI agents are intended to perform work across business systems, using information and tools to complete tasks rather than simply responding to questions. Their potential applications extend into procurement, finance, human resources, maintenance, and other processes where employees currently move between different applications.

Developers can create agents using increasingly accessible tools, but building a working demonstration does not necessarily mean the software is suitable for daily business operations.

An agent might retrieve purchase order information in a test environment, for example, while struggling with the supplier-specific contracts, approval limits, or incomplete records encountered in an established procurement system.

Deployment requires integration with the customer’s software, permissions appropriate to the work, testing against realistic information, and arrangements for maintaining the agent as business requirements change.

SAP’s initial agent competition demonstrated strong participation from partners, but the number produced provided limited evidence of customer value. Karl Fahrbach, SAP’s chief partner officer, is now arguing for partner performance to be assessed using results such as operating efficiency, cost reductions, and revenue improvements.

Those outcomes require measurement after implementation. A business may deploy an agent that reduces one administrative step without reducing the total time needed to complete the wider process, particularly if employees must check its decisions or correct information transferred between applications.

Similarly, the number of agents in operation cannot establish their combined value when different systems perform overlapping functions or introduce additional support requirements.

Managing agents across the application estate

SAP is bringing partner-developed agents into SAP AI Agent Hub, where customers can discover, select, provision, and manage them alongside agents supplied directly by SAP.

The company says the hub already supports agent governance across hundreds of businesses and is available without charge to customers managing up to 500 AI agents.

Those figures describe the management platform rather than the number of organisations that have deployed the newly developed partner agents. They nevertheless indicate how SAP intends to address the operational challenge created when multiple developers supply automation into the same enterprise environment.

Individual agents may require access to finance records, supplier information, employee data, or operational applications. Managing those connections becomes harder when different teams introduce software independently, particularly where the systems are supplied by separate partners.

A common management environment can provide visibility over available agents and the permissions and lifecycle arrangements through which they operate. However, the extent of oversight depends on the integrations available and how customers configure the underlying services.

SAP has also introduced a partner validation programme covering development requirements, certification, and a review process that agents must complete before qualifying for distribution through the programme.

Validation can establish standards for software entering the ecosystem, but it does not replace customer testing. An agent that satisfies platform requirements may still be unsuitable for a particular organisation’s data quality, approval procedures, or regulatory obligations.

The consequences of errors can also vary considerably between applications. A routine document retrieval task presents different risks from an automated action affecting supplier payments or regulated manufacturing records.

Partners need specialist operational knowledge

SAP’s strategy increasingly depends on partners adapting AI to the specific processes used within different industries, drawing on the existing implementation work carried out by software consultancies and specialist developers.

Enterprise applications may perform similar functions across sectors while operating under very different constraints. Procurement in automotive manufacturing, for example, can involve technical specifications and supplier dependencies that do not arise in the same way within professional services.

Developers familiar with those differences may be better positioned to design agents that work within established procedures, although industry knowledge alone does not guarantee reliable software.

SAP is seeking participation from AI-native businesses alongside the systems integrators already working with its enterprise customers. The €100 million investment fund announced at Sapphire is intended to support partners that build and successfully deploy agents.

The company reports that nearly 1,000 customers have expressed interest in the programme. As of 1 October, more than 500 customers had been confirmed and over 400 partners had enrolled.

Those figures demonstrate participation in the initiative, not that the fund has already distributed €100 million or that every confirmed customer has an operating partner agent.

The financial structure may encourage developers to focus on implementation, but its effectiveness will depend on the criteria used to determine successful deployment and whether reported customer benefits are independently measurable.

Development tools lower the initial barrier

SAP is also providing access to Joule Studio for qualifying customers in its Early Adopter Care programme through the end of 2026. More than two dozen partners are contributing to that programme.

Joule Studio supports the development of applications and agents through intent-based tools, allowing users to describe desired functionality in natural language. This can reduce the amount of conventional programming required to create certain components.

Such tools may make initial development faster, although they do not remove the engineering required to validate behaviour, establish permissions, or connect applications to reliable enterprise data.

SAP plans to provide practical development activities at TechEd, including an Agent Lab in which partners can experiment with the platform and its latest software capabilities.

The wider partner network includes more than 200,000 cloud-certified consultants and over 38,000 business AI-certified professionals, according to SAP. Certification totals indicate training activity across the ecosystem rather than completed customer projects or verified AI-related productivity gains.

As the number of available agents increases, organisations will need to decide which capabilities justify deployment and how to prevent the resulting software environment becoming harder to administer.

That places greater importance on interoperability, agent ownership, support arrangements, and the ability to measure performance within complete business processes.

Customer evidence becomes the commercial test

SAP’s move towards implementation reflects the difference between expanding a software catalogue and establishing a sustainable enterprise application business.

Developers may be able to produce agents quickly, but customers still face the cost of integration, governance, process changes, and ongoing maintenance. Those costs influence whether an automation project produces a financial return.

The company already has an extensive customer base using its enterprise software, giving partners access to organisations with established operational requirements and substantial application estates.

That opportunity also comes with demanding expectations concerning reliability and accountability. AI agents connected to important business records cannot be evaluated solely on their ability to complete a convincing demonstration.

SAP TechEd, scheduled for 27–29 October in Berlin, will provide an early assessment of progress towards the expanded development target. The more durable measures will be the number of agents operating reliably in customer environments and the benefits those customers can demonstrate after implementation.

For now, SAP has reported rapid partner development, a growing catalogue, and significant programme participation. Whether that activity becomes commercially useful automation will depend on the work undertaken after the agents have been built.