Every Vendor Has an AI Story Now
Walk into any ERP conference in 2026 and the messaging is uniform: AI is embedded everywhere, transforming everything, available now. SAP has its Joule copilot. Oracle has its AI agents. Microsoft has Copilot woven through every module of Dynamics 365. NetSuite has its AI-powered forecasting and anomaly detection. Even smaller vendors who have not shipped a meaningful product update in three years have rebranded their roadmaps as "AI-first."
The marketing is loud, confident, and almost entirely disconnected from what is actually in production and delivering value today. Our job in this article is to give manufacturers and distributors an honest assessment of which generative AI capabilities are real, working, and worth paying for — and which are aspirational features being sold as present-tense deliverables.
What Is Actually Working Today
The generative AI capabilities in enterprise ERP that are delivering genuine, measurable value today fall into three categories: natural language querying, automated document processing, and intelligent exception handling.
The generative AI capabilities in enterprise ERP that are delivering genuine, measurable value today fall into three categories: natural language querying, automated document processing, and intelligent exception handling.
Natural language querying — asking your ERP questions in plain English and receiving accurate answers — is the most mature generative AI capability in the enterprise software market. Microsoft Copilot for Dynamics 365 Finance and Operations is the most widely deployed example: users can ask questions like "what is our current inventory position for item 4471 across all warehouses" or "show me open purchase orders from suppliers with delivery dates in the next 14 days" and receive accurate, contextual responses. This is not science fiction — it is in production at thousands of organizations.
The business value is real but concentrated. The users who benefit most are managers and executives who need data but do not want to learn query tools or navigate ERP report hierarchies. For power users who already know how to access the data they need, the value is modest. For occasional users or leaders who currently depend on someone else to pull reports, the value is substantial.
Automated document processing — using AI to extract and process information from purchase orders, invoices, shipping documents, and customer communications — is the second area of genuine current capability. AI document extraction tools can process unstructured documents with accuracy rates above 95% for well-formatted inputs, dramatically reducing the manual data entry burden in accounts payable, procurement, and order management. This capability is available through native ERP features and standalone tools that integrate with most major ERP platforms.
Intelligent exception handling — AI-powered identification of anomalies, discrepancies, and risks in operational data — is the third area of current value. Examples include AI that flags purchase orders where pricing deviates significantly from historical rates, inventory levels that are trending toward stockout faster than replenishment can cover, or customer orders with characteristics that correlate with downstream payment risk. These are not predictive analytics in the traditional sense — they are AI systems that continuously monitor operational data and surface exceptions that warrant human attention.
What Is Being Oversold
The generative AI capabilities that are being most aggressively marketed but are least mature in production fall into two categories: autonomous agents and AI-generated business insights.
Autonomous agents — AI systems that take actions in your ERP without human approval, such as automatically generating purchase orders, releasing production orders, or adjusting pricing — are technically feasible but organizationally premature for most manufacturers. The challenge is not the AI capability. It is the governance frameworks, exception handling protocols, and audit trail requirements needed to deploy autonomous systems in business-critical workflows. Organizations that have deployed autonomous agents successfully have invested heavily in the governance infrastructure around them — work that is rarely visible in vendor demos.
AI-generated business insights — narrative summaries, strategic recommendations, and forward-looking analysis generated by ERP AI — are the most hyped and least reliable category. The models are fluent and authoritative-sounding. They are also prone to confident hallucinations, particularly when reasoning about multi-dimensional business data with complex interdependencies. We have seen AI-generated ERP reports that were grammatically perfect and factually incorrect in ways that required domain expertise to detect.
The Right Framework for Evaluating AI Claims
When evaluating ERP vendor AI claims, we recommend applying three questions. First: is this feature in production at reference customers who are willing to discuss it candidly? Not on a roadmap, not in beta — in production, with measurable results. Second: what is the governance model when the AI is wrong? Every AI system produces errors. The question is whether the vendor has designed error detection and correction workflows rather than assuming the AI will always be right. Third: what does the implementation actually involve? AI features that require significant data preparation, model training, and integration work to deliver value should be evaluated based on the fully-loaded implementation cost, not just the feature checklist.
The manufacturers who are getting the most value from ERP AI today are not the ones who bought the most AI features. They are the ones who identified the two or three AI capabilities most relevant to their specific operational pain points, implemented them rigorously, and built the internal processes to sustain them.
Found this useful? Share it.
Share on LinkedIn