Enterprise use case · Shopify + Odoo + OpenAI
AI-assisted stock correction
Physical counts, ERP quantities and storefront availability disagree, while teams lack a consistent way to decide which differences need attention first. Category managers receive prioritized stock discrepancies and specific adjustment proposals for approval.
Systems involved
What each system contributes
39 actions
- Inventory
- Products
64 actions
- Products13
- Sales Orders12
25 actions
- Chat Completions8
Physical-count and inventory-adjustment operations are not in the listed Odoo action groups, so the design checks the installed inventory interfaces and uses authorized API or Rules + Code steps where required. The model step runs against the AI provider you choose, including Anthropic, Google Gemini or a model you host yourself. Anything outside a pre-built action runs through an API, protocol or database call, or as Rules + Code in Python, Bash, Go, Java, .NET, PHP, PowerShell or JavaScript/TypeScript, with the same approvals and audit.
Architecture
One governed run
How it starts: An approved physical-count batch is available for comparison with the corresponding Odoo and Shopify inventory context.
Step by step
How the process runs
- 01
Align the count scope
Read the authorized count through the configured Odoo interface and match its item, location and timestamp to the relevant system records.
- 02
Calculate the difference
Rules compare counted and recorded quantities while accounting for the agreed treatment of reservations and movements after the count.
- 03
Rank and explain
The selected model evaluates the discrepancy context, proposes review priorities and recommends a possible adjustment or recount with its supporting evidence.
- 04
Review the proposal
The category manager inspects the count and movement context, corrects assumptions and selects the permitted action.
- 05
Apply the approved correction
A separate Odoo action or API step records the authorized inventory adjustment before the agreed availability value is published to Shopify.
Control
Human decision and exception path
Human decision
The named category or inventory owner approves each adjustment scope before stock records change; any financial posting follows the required finance authority as well. A model-ranked discrepancy cannot alter inventory by itself.
Exception path
- Stale counts or unresolved in-transit movements hold the proposal for reconciliation or recount.
- Unmatched variants, units or locations stop adjustment of the affected item.
- Partial adjustment or publication failures retain separate outcomes so storefront updates do not imply a completed ERP correction.
Evidence
What gets audited
- Count batch, item, location and inventory snapshot references.
- Quantity comparison and movement-treatment rules.
- Model priority, proposed correction and owner approval.
- Odoo adjustment and Shopify publication responses.
Nexus runs on premises, in private or sovereign cloud, or air-gapped. Credentials stay in the vault, roles decide who can approve, and the execution record stays inside your deployment boundary.
Sovereignty and deploymentRelated
Patterns, connectors and next steps
- Shopify-Odoo integration
Integration pattern · Shopify sales reach Odoo with mapped variants, stock locations and the order detail fulfillment needs.
- Shopify connector
39 pre-built actions
- Odoo connector
64 pre-built actions
- OpenAI connector
25 pre-built actions
- Enterprise automation
Governed processes across the systems you already run.
- Order to cash across CRM, ERP and billing
Another enterprise use case.
FAQ
Questions about this use case
- Does Nexus replace any of these systems?
- No. Each system keeps its own records. Nexus runs the process between them and records each decision.
- How does the run start?
- An approved physical-count batch is available for comparison with the corresponding Odoo and Shopify inventory context.
- What if a step is not covered by a pre-built action?
- Physical-count and inventory-adjustment operations are not in the listed Odoo action groups, so the design checks the installed inventory interfaces and uses authorized API or Rules + Code steps where required. The model step runs against the AI provider you choose, including Anthropic, Google Gemini or a model you host yourself. Anything outside a pre-built action runs through an API, protocol or database call, or as Rules + Code in Python, Bash, Go, Java, .NET, PHP, PowerShell or JavaScript/TypeScript, with the same approvals and audit.
- Does the model decide the physical quantity?
- No; an authorized count supplies that evidence, while the model prioritizes investigation and proposes how to handle the discrepancy.
Use case review
Discuss this use case
Bring your version of the process. We will map the systems, approvals and evidence it needs.