What ERPs now ship as AI features for fashion, and what is still marketing
ERP vendors now announce AI agents in every release. A sober look at which features are generally available, what they do in order management and how fashion companies should test the claims.
KEY TAKEAWAYS Summary by the editors
- Microsoft's Sales Order Agent for Business Central is generally available: it reads customer emails, creates quotes or orders and routes them for configurable human review.
- SAP's Q1 2026 release listed AI-assisted sales order creation from unstructured documents such as PDFs as generally available in S/4HANA Cloud Public Edition.
- Several heavily promoted agents are still in beta, such as SAP's Dispute Resolution Agent for invoice disputes in the same release.
- Gartner predicted in June 2025 that over 40 percent of agentic AI projects will be cancelled by the end of 2027 and warned of widespread agent washing by vendors.
- None of the order-related features reviewed here is specific to fashion; size, colour and seasonal order logic still has to be tested against each company's own data.
ERPs now ship a real but narrow set of AI features that matter for fashion order management: turning emails and PDF orders into draft sales orders, explaining errors, summarising data and assisting with disputes. Broader claims about autonomous agents running planning or wholesale are, for now, mostly roadmap or beta. The useful test is simple: is the feature generally available in your edition, what exactly does it do, and where does a person stay in control?
What AI features are ERP vendors actually shipping?
Two examples show what is available today. Microsoft documents a Sales Order Agent for Dynamics 365 Business Central as generally available. It monitors a designated mailbox, identifies the customer by email address, extracts requested products and quantities, can check item availability, creates sales quotes or orders, replies to the customer and creates tasks for user review. Review is configurable: by default, the first message from registered senders and all messages from unregistered senders are reviewed. The agent consumes Copilot Credits and stops processing when they run out, and custom pricing or discount logic typically requires manual review.
SAP's Q1 2026 release highlights, published in April 2026, list AI-assisted sales order creation as generally available in S/4HANA Cloud Public Edition, processing unstructured data such as PDFs to populate sales orders. The same release lists AI-assisted error explanation and AI-assisted payment advice processing as generally available, and a Dispute Resolution Agent for invoice disputes as beta. SAP also states that Joule now operates across 35 solutions with more than 2,500 skills and over 30 specialised agents.
| Feature | Vendor | Status (as published) | Relevance for fashion |
|---|---|---|---|
| Sales Order Agent (email to quote or order) | Microsoft, Business Central | Generally available | Re-orders and small accounts ordering by email |
| AI-assisted sales order creation from PDFs | SAP, S/4HANA Cloud Public Edition | Generally available | Retailer orders that arrive as documents, not EDI |
| AI-assisted error explanation | SAP | Generally available | Faster resolution of posting and integration errors |
| Dispute Resolution Agent | SAP | Beta | Invoice disputes and retailer deductions |
| AI-assisted payment advice processing | SAP | Generally available | Matching retailer payments and remittances |
Which of these matter for a fashion wholesale business?
The order intake features are the most relevant. Large department stores send orders via EDI, but many specialist retailers still order by email, spreadsheet or PDF, especially for re-orders between seasons. Converting those into draft orders saves manual entry, provided the conversion handles fashion specifics: style, colour and size matrices, size scales that differ by product group, delivery windows and prepacks. Neither vendor's documentation reviewed here describes fashion-specific handling, so this has to be tested.
Error explanation and dispute support are relevant to the integration layer. Many EDI and invoicing problems surface as cryptic ERP errors or as payment differences, and tools that explain causes in plain language can shorten the time to resolution.
Payment and remittance processing is a further practical area. Retailers often pay several invoices at once, net of deductions, and send remittance details in varied formats. Features that read payment advices and propose matches against open items can reduce manual work in accounts receivable, as long as unmatched differences are routed to someone who understands the retailer's deduction codes.
What is still marketing?
Gartner's June 2025 prediction that over 40 percent of agentic AI projects will be cancelled by the end of 2027 cites escalating costs, unclear business value and inadequate risk controls. Gartner also described agent washing, the rebranding of chatbots, robotic process automation and existing assistants as agents, and estimated that only about 130 of the thousands of agentic AI vendors are real. It noted that many use cases positioned as agentic do not require agentic implementations.
Signs that an ERP AI claim is ahead of reality include:
- No clear statement of availability (general availability, preview or beta) and the editions it applies to.
- Demonstrations on generic data rather than on multi-size, multi-season fashion orders.
- Productivity percentages without a stated baseline or method.
- Undefined pricing, for example consumption credits without an estimate per order processed.
- No description of how a person reviews, approves or reverses the agent's actions.
How should fashion companies test ERP AI features?
- Pick one process with clear volume and pain, such as email re-orders from specialist retailers.
- Check the feature's status and licensing in your edition, including consumption-based costs.
- Build a test set of real historical orders, including difficult ones: size runs, prepacks, partial cancellations and mixed delivery dates.
- Measure accuracy per line, not per order, and the time a person needs to review and correct.
- Set review rules: which customers or order types may be processed with less review, and which always need a person.
What should IT and sales leaders expect next?
Gartner also forecast that 33 percent of enterprise software applications will include agentic AI by 2028, up from less than 1 percent in 2024. ERP vendors are moving in that direction, so the question is less whether AI arrives in the ERP than whether the company's data and processes are ready for it. Clean customer, product and price master data, documented order rules and reliable integrations will decide whether these features save time or simply produce errors faster.
Governance matters as much as features. Decide who may switch an agent on, which customers it may serve, how its actions are logged and how quickly a wrong order can be reversed. These rules are easier to agree before a pilot than after the first error reaches a retail customer.
Frequently asked questions
Can an ERP AI agent take B2B orders from email?
Yes, at least one generally available feature does this: Microsoft's Sales Order Agent for Business Central reads emails, creates quotes or orders and routes them for configurable review. Fashion companies should test how it handles size and colour matrices before relying on it.
What is agent washing?
Gartner uses the term for vendors rebranding existing products, such as chatbots or robotic process automation, as AI agents without substantial agentic capabilities. It estimated in 2025 that only about 130 of thousands of agentic AI vendors are real.
Does SAP offer AI for sales orders?
SAP's Q1 2026 release highlights list AI-assisted sales order creation from unstructured documents such as PDFs as generally available in S/4HANA Cloud Public Edition. Availability in other editions should be checked with SAP.
Are ERP AI features included in the licence?
Not always. Some features consume credits or require additional licences; Microsoft's Sales Order Agent, for example, uses Copilot Credits and stops when they are depleted. Check pricing per edition before planning a rollout.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
