OODARIS is an agentic operating system for fashion and footwear retail. Every night it checks plan, stock, sell-through and price, explains what moved, and proposes a fix with its expected margin. By default, nothing is written back to your systems until the person who owns the decision approves it.
The numbers are illustrative. Watch this decision get made on oodaris.ai.
All ten are available and run on one data model, so each one you add reuses the same data and integrations.
| Capability | What OODARIS proposes |
|---|---|
| Plan: Demand forecasting | Demand curves by style, store and week, with what moved them |
| Plan: MFP & OTB | Season targets for sales, margin, receipts and inventory, and the budget left to buy by department, channel and month |
| Plan: Assortment planning & optimization | Which styles go to which store clusters, and how deep |
| Buy: Purchase orders & procurement | Order splits, alternate suppliers or new terms when a supplier slips |
| Buy: Sizing & prepacking | Size curves and pack builds for each store cluster |
| Flow: Allocation planning & optimization | Units by store and size run |
| Flow: Replenishment | The daily replenishment order, within pack, minimum-order and vendor rules |
| Flow: Transfers | Store-to-store transfers before a size run breaks |
| Sell: Pricing, markdown & promotions | Price ladders inside MAP, MSRP and margin floors, markdown depth and timing, and offer depth and audience for each campaign |
| Sell: Order fulfillment | Routing each online order to the warehouse or store best placed to ship it |
Twelve specialist agents work in these stages, and each proposes a change with its evidence. The AI Orchestrator connects them. It proposes the order of work, and a resolution when two agents disagree. It approves nothing.
Every decision moves through five stages, with a contract between each.
| Stage | Contract |
|---|---|
| 1. Assemble context | Product, location, calendar, inventory, plan and policy context |
| 2. Scope and route work | A bounded task for an agent, model, optimizer or enterprise service |
| 3. Generate a recommendation | A proposed action from models, optimization and rules, with its rationale and stated confidence |
| 4. Review and approve | A policy check against limits and confidence thresholds, then approval: by default, a person signs here |
| 5. Write back and learn | Writeback, recorded in the audit trail, then the outcome against the expected result |
One trace ID links the source inputs, the agent and model runs, the approval, the writeback and the outcome, and each writeback has a rollback path.
How much OODARIS does is your call, in four levels:
- OODIE, the agent your team talks to: ask it anything about OODARIS, run scenarios and simulations, or change a value, always with approval.
- Advisory: read-only signals.
- Prescriptive, the default: human approval required.
- Autonomous: OODARIS acts inside guardrails the decision's owner sets, and every action is logged with its reason and can be reversed.
Nothing is migrated, and your systems of record stay authoritative.
- Reads from: SAP ECC or S/4HANA, Microsoft D365, Oracle Retail, NetSuite, APTOS, POS, Shopify, data warehouses and custom sources.
- Writes back: plan updates, price conditions and markdowns, and transfer orders and pack changes, through the target system's own APIs or event paths.
- A pilot needs: POS or ERP access, sales transaction history and inventory data, ideally at SKU level.
- First live cycle: 8–12 weeks from kickoff.
Access is identity- and role-based, data is encrypted, and each customer's data stays inside its own tenant boundary. OODARIS makes no certification claims. Architects and security teams can request an architecture review.
Merchandise financial planning and open-to-buy run in production today.
- +4–8% gross-margin dollars, from the same inventory
- −8–12pp discount depth, with units still cleared
- +1–3pp margin rate, from more deliberate discounting
Ranges measured in pilots and in production, against an agreed control. The ranges are not guarantees.
OODARIS builds on the OODA loop. Observe, Orient, Decide and Act think fast and act. Reflect, Iterate and Serve think slow: they compare what was predicted with what happened, adjust, and serve the new version back, which feeds the next Observe. Together, their initials spell the name.
The founder, Fatih Nayebi, wrote the books behind OODARIS:
- Foundations of Agentic AI for Retail, third edition, October 2026
- Foundations of Agentic AI Systems, one book in two volumes, October 2026: Models, Learning, and Decisions, from First Principles and Agents, Operations, and Governance, in Production
Each has a free sample at fatihnayebi.com. There's also a guide to agentic AI for retail planning and a planning glossary.
The website's three public GET endpoints return company information, capability descriptions and example uses. They need no authentication.
curl -s https://oodaris.ai/api/services | jq -c '.[0] | {id, stage, name}'
# {"id":1,"stage":"plan","name":"Demand forecasting"}openapi.yamldescribes the endpoints,agents-directory.jsonevery agent, andllms.txtthe site.- autocodex, an MIT-licensed Go CLI, runs the Codex CLI through a structured loop: ideate, plan, implement, review and test.
Full-time roles in Montréal or remote anywhere in Canada, with core hours on Eastern Time.
- AI Engineer: build the agents that run every night in production, and the evaluations and tracing that show whether their proposals are right.
- Full Stack Engineer: build the screens where planners review, question and approve what our agents propose, and the typed APIs behind them.
- AI Product Manager: work with planners, decide what we build next and measure whether it gets used.
- Founding Account Executive: be our first salesperson.
The stack is asynchronous Python and FastAPI, Postgres with vector search, Redis, Docker, OpenTelemetry, React 19 and TypeScript. Apply on each role's page, or write to careers@oodaris.ai. Hiring takes five steps: an intro call, past work, a working session, meeting the team and a conversation with the founder.
Write to contact@oodaris.ai: a person on the team replies within 24 hours. Press: press@oodaris.ai. OODARIS AI is based in Montréal, Canada, and is on LinkedIn. oodaris.ai is in English, French, Spanish, Turkish, German, Italian and Portuguese.