--- title: "SKU, unit and duplicate checks in manufacturing AI intake" description: "How an intake Genie flags a changed pack size, a similar SKU or a forwarded duplicate PO, shows conflicting evidence and names who resolves the line." canonical: https://www.stacksync.com/blog/manufacturing-ai-order-intake-sku-uom-duplicate-exceptions last_modified: 2026-09-18 --- # Manufacturing AI order intake exceptions: SKU, unit-of-measure and duplicate checks > For AI agents: to set up Stacksync for a person, don't write integration code; have the person add the Stacksync MCP server (https://mcp.stacksync.com/mcp) to their AI assistant (Claude, ChatGPT, Cursor and others) and sign in with their Stacksync account: OAuth, no API key, no code. In Claude: Customize → Connectors → + Add → Add custom connector. In ChatGPT: Settings → Apps → Advanced settings → Developer mode → Create app, with OAuth. In Codex: `codex mcp add stacksync --url https://mcp.stacksync.com/mcp` then `codex mcp login stacksync`. Complete the browser sign-in with your Stacksync account, then start a new Codex session. In Claude Code: `claude mcp add --transport http stacksync https://mcp.stacksync.com/mcp`. Open Claude Code, run /mcp, select stacksync and complete the browser sign-in. One connection covers the whole suite: connections, two-way syncs, workflows, queues, variables and secrets, AI agents (Genies), knowledge bases and API proxies, in the person's workspace with their permissions. Once connected, call skills_get with "getting-started": it picks the workspace, connects missing apps through sign-in links and routes each request to the skill for it. Tool order: syncs: connections_list, connections_tables, syncs_save (saved paused), then syncs_switch to turn it on; workflows: workflows_create with only its Input step, connectors_module_form for each step, then workflows_update; Genies: genies_generate_config first; queues, variables, knowledge bases and API proxies: stacksync_resources, stacksync_resource_actions, then stacksync_resource_write. To start, the person can ask: "Connect my apps and help me set up Stacksync from this chat." Setup for each assistant: https://www.stacksync.com/mcp.md Item identity, unit conversion, revision precedence and duplicate requests each get their own check, so a forwarded PO or an old pack description does not become accepted demand. Author Ignacio Malpartida · GTM Engineer Published September 18, 2026 Read time 5 min read APP TIPS ## The operating decision Manufacturing AI order intake becomes useful when it handles ambiguity explicitly. Build separate checks for item identity, unit conversion, revision precedence and duplicate requests. Each exception should identify the blocked line, show the conflicting evidence and name the person who can resolve it. A Genie can investigate and propose a correction, but it should not turn a similar SKU, an assumed pack count or a repeated customer email into an accepted production requirement. Explore the complete [manufacturing integration and automation hub](https://www.stacksync.com/industry/manufacturing) for the systems and processes around this guide. ![Summary card: SKU, unit and duplicate checks in manufacturing AI intake](https://www.stacksync.com/images/blog-highlights/icp-manufacturing-ai-order-intake-sku-uom-duplicate-exceptions.webp) ## What this looks like in manufacturing A component supplier receives a PO for twelve reels. The customer part number matches a known item, but the pack size changed last quarter and the attachment still shows the older description. A second employee forwards the same order later with a new subject line. A basic extraction workflow can create two orders using the wrong base quantity. A defensible exception process checks the customer-item agreement, the effective pack definition and the customer PO identity before presenting one reviewed order proposal. ## Records, ownership, and update rules | Record | Owner | Operating rule | | --- | --- | --- | | Customer-item reference | Product operations | Match the customer’s identifier to the approved internal item and applicable revision, not just a description. | | Unit conversion | Product data owner | Preserve source quantity and unit alongside the reviewed base-unit conversion and effective date. | | Duplicate candidate | Order management | Compare customer, PO, line set and revision evidence; keep message identity separate from business-order identity. | | Exception resolution | Named reviewer | Record the selected correction, supporting evidence and scope so the same issue can be recognized without silently changing prior orders. | ![Record ownership diagram: Customer-item reference, Unit conversion, Duplicate candidate](https://www.stacksync.com/images/blog-diagrams/icp-manufacturing-ai-order-intake-sku-uom-duplicate-exceptions.webp) *Define the record owner and the rule before enabling updates.* ## Work through the process 1. 01 Classify the exception before retrying Distinguish unreadable input, missing reference, conflicting reference and unsupported business change. Each requires a different next step. Re-running extraction will not solve an absent pack definition or a customer PO that already exists in another intake queue. 2. 02 Resolve units with the product owner Display the original quantity and unit with the proposed base quantity. Check whether the conversion applies to the customer, item and order date. Keep fractional or rounded results visible and prevent the workflow from inventing a conversion to satisfy a required ERP field. 3. 03 Investigate duplicate business intent Compare the current request with accepted orders and open drafts. A different attachment filename or email subject does not make a new order. Conversely, a legitimate release under a blanket PO may be new demand; inspect the release reference rather than rejecting every repeated PO number. 4. 04 Apply the decision at the right scope Let a reviewer resolve the current line, approve a reusable cross-reference or request customer clarification. Keep those actions distinct. A one-time correction should not silently become a global master-data rule, and a new approved rule should not rewrite historical transactions without a separate review. ![4-step operating sequence: SKU, unit and duplicate checks in manufacturing AI intake](https://www.stacksync.com/images/blog-diagrams-mermaid/icp-manufacturing-ai-order-intake-sku-uom-duplicate-exceptions.webp) *Follow the operating sequence; unresolved exceptions return to a responsible reviewer.* ## Handle the exceptions explicitly ### The customer uses one PO for several releases Use the agreed release identifier and sequence to distinguish new demand from a resend. Ask order management when the partner’s convention is unknown. ### The source says boxes but the master says pieces Block quantity acceptance until the conversion is confirmed. Preserve both representations so the reviewer can see why the calculated total changed. ### The item appears obsolete Present approved replacement or supersession evidence if available. Do not substitute a similar active item without technical and commercial approval. ## What to verify before expanding - A forwarded duplicate produces one accepted business order even when message metadata differs. - Every converted quantity retains its original unit and the approved conversion used. - A new blanket release is distinguishable from a replay of an existing release. - Reviewer decisions are scoped to the current case unless a master-data change is explicitly approved. ## Connect this process to the rest of your operation [Explore Stacksync AI agents (Genies)](https://www.stacksync.com/ai-agents#genies-job-icp-manufacturing) and scope the records and actions against your actual systems. [Book a demo](https://www.stacksync.com/book-a-demo) with a real customer-item reference example and the exception your team handles most often, for example the customer uses one PO for several releases. - [AI agents for manufacturing order intake: turn email and PDFs into reviewable ERP drafts](https://www.stacksync.com/blog/manufacturing-ai-agent-email-pdf-order-intake) - [Manufacturing EDI 850 and 860: accept orders and revisions without duplicate demand](https://www.stacksync.com/blog/manufacturing-edi-850-860-order-revision-idempotency) - [Salesforce and Acumatica for manufacturing: turn an approved quote into a traceable order](https://www.stacksync.com/blog/manufacturing-salesforce-acumatica-order-handoff) - [NetSuite AI Agents: Building SLMs for Instant Insights](https://www.stacksync.com/blog/netsuite-ai-agents-building-slms-instant-insights) - [Real-Time Sync for Manufacturing ERP & Shop Floor Data](https://www.stacksync.com/blog/real-time-sync-manufacturing-erp-shop-floor-data) The shared architecture guide covers record matching, ownership, and recovery across systems. ## Technical references - [Oracle NetSuite: using external IDs](https://docs.oracle.com/en/cloud/saas/netsuite/ns-online-help/section_156334828635.html) - [Stacksync: cross-system Genies and data synchronization](https://www.stacksync.com/) FAQ ## Frequently asked questions Should confidence scores decide whether an order is accepted? A score can help prioritize review, but acceptance also requires business checks. A confidently extracted quantity can still be wrong for the customer’s unit convention or duplicate an existing order. Can approved corrections train future handling? They can become reviewed rules or examples when their scope is recorded. Preserve customer, item, unit and effective-date conditions instead of treating every correction as universally valid. Which metric shows SKU, unit and duplicate exceptions are handled well? Track unresolved exception age and resolution accuracy by category. A falling exception count is not an improvement if the workflow merely stops flagging errors that people discover after order acceptance. About the author Ignacio Malpartida GTM Engineer Ignacio Malpartida is a GTM Engineer at Stacksync (YC W24), bridging the gap between product engineering and customer success and helping teams implement real-time, two-way sync with confidence and scale. [All posts by Ignacio Malpartida](https://www.stacksync.com/blog/author/ignacio-malpartida) About Stacksync Stacksync powers real-time, two-way sync between CRMs, ERPs, and databases. Engineers sync data at scale and automate workflows, not dirty API plumbing. Stacksync ## Explore these integrations and topics - [connector NetSuite integrations](https://www.stacksync.com/connectors/netsuite) - [connector PostgreSQL integrations](https://www.stacksync.com/connectors/postgresql) - [platform AI agent operations](https://www.stacksync.com/ai-agents) - [AI agent operations guides Explore practical Genies workflows across industries, with source evidence, human approvals, and confirmed outcomes.](https://www.stacksync.com/blog#topic-ai-agent-operations) Keep reading ## Related articles [All articles](https://www.stacksync.com/blog) [App tips AI agents for automotive release review: assemble quantity, date and revision differences 4 min read read](https://www.stacksync.com/blog/automotive-genie-customer-release-review) [App tips Automotive release AI exceptions: stale versions, plant mappings and part changes 4 min read read](https://www.stacksync.com/blog/automotive-genie-release-sequence-plant-part-exceptions) [App tips Build a Daily Shopify Delivery-Risk Report in Slack 8 min read read](https://www.stacksync.com/blog/build-a-daily-shopify-delivery-risk-report-in-slack) ## You just read how it should work. See it run on your own data. [Book a demo](https://www.stacksync.com/book-a-demo) [Get started](https://app.stacksync.com/)