Summary
- SEIDOR's Autonomous Enterprise Factory brings together more than 50 specialists and is supporting ten projects involving SAP business systems.
- The consultancy is developing reusable AI agents for procurement, liquidity management, production planning, and other operational functions.
- Customers may need to modernise their ERP environments and improve data quality before allowing agents to make or execute consequential decisions.
Spanish technology consultancy SEIDOR has established a global operation to develop artificial intelligence agents for enterprise resource planning systems, with ten customer projects already under way and more than 50 specialists involved.
The SEIDOR Autonomous Enterprise Factory will develop applications capable of analysing business information, recommending operational decisions, and carrying out authorised tasks within SAP environments. Its initial work covers finance, procurement, logistics, and production, drawing on the company’s experience implementing enterprise software across different industries.
Based in Barcelona, SEIDOR has worked with SAP since 1996 and now employs more than 5,000 SAP specialists across its international operations. The new unit brings together software architects, functional consultants, applied AI specialists, and experts in data integration to develop agents that can be adapted to customers’ existing systems.
The initiative follows SAP’s introduction of its Autonomous Enterprise strategy at Sapphire 2026, which envisages business applications taking greater responsibility for interpreting operational information and coordinating tasks. SEIDOR’s projects are being developed through SAP’s Business AI Partner-Led Adoption Program, supported by a global €100 million fund for partner-led AI development.
Agents enter established business processes
SEIDOR is developing applications intended to identify opportunities to improve liquidity by examining receivables, payments, and inventory. Other agents compare supplier quotations and purchasing conditions or analyse potential production risks associated with materials, capacity, and delivery schedules.
These functions involve decisions already embedded within enterprise software, but their execution has traditionally depended on employees interpreting information and transferring it between different processes. An AI agent connected to the underlying applications could reduce some of that work by gathering relevant records, evaluating alternatives, and preparing a recommended action.
For procurement, this might involve comparing supplier terms against business requirements rather than relying solely on the lowest quoted price. In production planning, a similar approach could identify material or capacity constraints before they affect scheduled work, provided the information available to the system is accurate and sufficiently current.
SEIDOR has already developed Intelligent Quote Sourcing, an application designed to analyse supplier quotations and recommend alternatives against defined business criteria. The solution won an SAP AI Agents Challenge in Italy, while the consultancy has 13 applications published on SAP Store.
Those existing applications provide a development base for the Factory, although their presence in SAP’s marketplace does not establish the extent to which customers have deployed them or the financial benefits achieved. SEIDOR has not disclosed production performance measurements for the ten projects currently under way.
Operational autonomy depends on the underlying systems
Allowing software to recommend a purchasing decision is different from authorising it to place an order, commit expenditure, or alter production arrangements. Organisations must determine which activities an agent can complete independently and where an employee remains responsible for approval.
Those boundaries become particularly important when automated processes operate across departments. A procurement decision may affect inventory levels, cash flow, production schedules, and contractual obligations, while the information governing those activities may be stored in separate applications.
Enterprise systems also carry the consequences of years of adaptation. Businesses frequently maintain customised ERP components, specialist integrations, and data structures developed around established operating practices. Introducing agents into these environments requires more than connecting an AI model to a database.
SEIDOR acknowledges that customers may need to modernise ERP components, improve data quality, and review legacy customisations before adopting more autonomous workflows. Preparing those environments is included within the Factory’s remit, alongside agent development and implementation.
The work will vary according to the technical maturity of each customer. A business with well-maintained operational records and standardised processes may be able to introduce decision support relatively quickly, while another organisation could need substantial data reconciliation and systems development before automated recommendations become dependable.
Permissions and auditability introduce further requirements. Employees need to establish what information an agent can access, which actions it may initiate, and how its decisions will be recorded. Where an agent’s recommendation is based on incomplete information, the system must also have an appropriate mechanism for identifying uncertainty or referring the decision for review.
These arrangements influence the economics of implementation because the cost extends beyond software development and model usage. Integration, security testing, staff training, process redesign, and continuing maintenance all contribute to the resources required to operate an automated workflow.
Enterprise software partners expand their role
SEIDOR’s approach reflects a broader change in the services offered by enterprise software consultancies. Having previously concentrated on deploying and maintaining complex applications, implementation partners are developing proprietary automation tools that operate within the systems they already understand.
Reusable agents may reduce the amount of development required for activities common to several customers, although individual implementations will still need to reflect local business rules, data structures, and approval procedures. The commercial value of that approach will depend on how much functionality can be transferred between organisations without extensive redesign.
The Factory is intended to serve midmarket businesses as well as larger enterprises, where application estates and integration requirements can be more complicated. SEIDOR’s presence in 45 countries provides a potential delivery network, but the company has not identified the customers involved in the first ten projects or confirmed the geographical distribution of those engagements.
Recent enterprise technology developments underline the importance of these underlying systems. AstraZeneca’s move to SAP cloud infrastructure, covered by Techopia earlier this month, is a separate project but demonstrates the continuing investment involved in modernising the software on which operational applications depend.
SEIDOR now plans to extend its catalogue of agents using the experience gained from existing implementations. The immediate test is whether its first ten projects can progress from development into dependable operational use, with customers able to demonstrate benefits against the cost and complexity of integrating AI into established business processes.












