Epicor Prism and JobBOSS AI assist inside your ERP. But the RFQ email lands outside the system. Here is why the intake layer must exist before any AI agent inside your ERP can act.
The ERP AI promise
The pitch: Every major ERP vendor now offers AI agents. Epicor Prism combines a network of vertical AI agents with industry-focused ERP and data structures. ECI's JobBOSS AI Assistant targets job shop quoting workflows. Genius ERP has Cortex AI. The marketing language is consistent: "automate workflows," "reduce manual work," "make faster decisions."
These tools are genuinely useful once the data is structured inside the ERP. Prism Business Communications, launched in June 2026, automates supplier RFQ workflows and procurement communications. It extracts pricing and lead times from supplier emails and accelerates purchasing decisions. These are real capabilities for the procurement team managing outbound RFQs.
But here is the practical test: paste a CAD file into the Prism chat interface. It cannot read the geometry. It cannot extract the bounding box, the material, the surface finish requirements. The ERP AI agent acts on structured data. It cannot ingest unstructured RFQs that arrive with email attachments.
Where the gap sits
The boundary: The RFQ email lands in the estimator's inbox. It contains the customer email, a CAD file attachment or PDF, and sometimes a spreadsheet of quantities. That is the intake point. From there, the estimator opens the CAD file, reads the dimensions, identifies the material, estimates cycle time, checks material stock, and builds a cost model. Only then does any data enter the ERP.
The gap is the gap between the inbox and the ERP quoting module. Every ERP AI agent on the market today operates on data that is already inside the system. Prism pulls from Epicor Kinetic. JobBOSS AI reads from the JobBOSS database. Neither reads the email attachment. Neither extracts geometry from a STEP file. Neither recognises that an M6 threaded hole with a 12 mm depth requires a specific tapping cycle time that differs from an M10 at 20 mm.
This is not a limitation the vendors are hiding. It is a structural boundary. ERP systems are designed to manage structured business records. Inbound RFQs are inherently unstructured. The conversion step requires CAD knowledge, geometric reasoning, and estimating expertise that no ERP AI agent yet provides.
Epicor Prism in practice
Supplier side, not estimator side: Epicor Prism Business Communications is an AI agent for supply chain procurement professionals sending RFQs to suppliers. On June 23 2026, CIO reported that the tool "automates RFQ workflows and accelerates supplier communications, helping supply chain businesses cut cycle time."
That is the procurement side of the transaction. The supplier receiving those automated RFQs still needs a tool to read them, extract the requirements, and build a quote. Prism does not help the fabricator or machine shop on the receiving end. It helps the buyer.
The Epicor Prism limitation analysis explains why Prism agents need a quoting front-end on the receiving side. For shops evaluating Epicor, the intake gap is the practical constraint that limits how much of the quoting workflow Prism can actually automate.
JobBOSS AI limits
Inside the system: ECI's JobBOSS AI Assistant targets quoting assistance inside the JobBOSS environment. It helps with AI BOM Builder and smart search across historical jobs. These are genuinely useful for an estimator working inside JobBOSS. The limitation is that the data must already be in the system for the AI to assist.
The Practical Machinist thread on inbound RFQ handling documents that a significant percentage of RFQs never make it into the ERP the same day they arrive. When the front office is slammed, RFQ calls go to voicemail, emails sit unread, and the data entry backlog grows. By the time the AI Assistant could help, the shop has already lost 12 to 24 hours of response time advantage.
The JobBOSS and Epicor quoting module limitations article covers the operator evidence from JobBOSS Ideas and Epicor User Forum that confirms the quoting module gap.
Genius Cortex needs data
Same pattern: Genius ERP's Cortex AI follows the same architecture. It assists with data already inside Genius , quote comparison, margin analysis, production planning. Capabilities that add real value once the quote is structured. But Cortex cannot help with the initial intake step: the PDF emailed from the customer, the DXF file that needs geometry extraction, the material specification buried on page three of a scanned drawing.
Every ERP AI vendor is building the same layer: intelligence on top of structured data. None of them are building the intake layer that converts unstructured inbound RFQs into structured data in the first place. That is not a criticism of the ERP vendors. It is a recognition that the intake problem requires fundamentally different technology , CAD geometry processing, file format handling, takeoff tools , that sits outside the ERP domain.
Common pattern
Intake before AI: The pattern across all ERP AI agents is identical. The AI operates on structured records inside the ERP. The inbound RFQ arrives outside the ERP. The gap between them is a manual data entry step that no AI agent in the ERP ecosystem currently automates.
The why 18.8 billion in AI funding skipped the estimator article puts this in context. Every major 2026 funding round in manufacturing AI went to systems that operate on already-structured data. The funders did not overlook the intake layer deliberately. They simply invested in what they could see , and the inbox is invisible to institutional capital.
The 5 signs your quoting process needs an ERP front-end article provides a practical diagnostic for shops trying to assess whether the intake gap is costing them work.
The practical takeaway
What this means: The gap between your ERP AI agent and the inbound RFQ is not a bug that Epicor or ECI will fix. It is a structural limitation of ERP architecture. ERP systems manage structured records. Inbound RFQs are unstructured. No amount of AI agents layered on top of the ERP changes that fundamental boundary.
Shops that recognise this boundary stop waiting for their ERP vendor to solve the intake problem. They add a front-end that handles the unstructured-to-structured conversion, then let the ERP AI agents do their job on the data inside the system.
What a front-end changes
The intake layer: A quoting front-end sits between the estimator's inbox and the ERP. It receives the inbound email, reads the CAD attachment, extracts dimensional data and material specifications, and sends a structured estimate to the ERP quoting module. The estimator reviews geometry and sets pricing. The front-end handles the data conversion.
This is not a replacement for the ERP AI agent. It is the prerequisite. The ERP AI cannot act until the data is inside the system. The quoting front-end puts it there. The combination of intake layer plus ERP AI gives the shop automated RFQ processing without requiring the estimator to double as a data entry operator.
<a href="/download">Run Kwantflow</a> as the intake front-end that feeds your ERP AI agents.
Ways estimators can keep quote review clear:
- Epicor Prism Business Communications handles supplier communications inside the ERP but cannot read or process inbound RFQ email attachments with CAD files.
- JobBOSS AI Assistant and AI BOM Builder operate on data already inside the ERP , neither can triage incoming RFQs or extract geometry from emailed STEP files.
- The RFQ intake gap is structural: ERP AI agents act on structured records, but the inbound RFQ arrives as unstructured email with a CAD attachment.
- A quoting front-end that processes RFQs locally , reading email attachments, extracting geometry, and sending structured estimates to the ERP , fills the gap before the ERP AI ever sees the data.
- Shops that bridge the intake gap respond faster to RFQs without replacing their ERP, because the intake layer complements rather than competes with the ERP AI investment.
