THE PROBLEM (AND WHY NOW)
Procurement is being asked to manage more risk with fewer people and tighter budgets. Price swings collide with slow approvals, expiring supplier quotes, and data scattered across tools. The result: missed hedging windows, re-sourcing under pressure, and margin leakage that finance sees only after month-end. Recent CPO surveys show risk management and resilience have moved to the top of the agenda; teams need earlier signals, clearer explanations, and faster decisions — volatility didn’t end; slack did. - Sunflower oil +39% (Q1 ’25) - Aluminum premiums +250% (Oct ’25) - Cocoa +85% (Apr ’25)
WHAT WE BUILT FIRST (BEFORE AGENTS)
To validate the need, we shipped four apps that teams use today:
1) Global OSINT Engine Configurable and fully parametric. Define query templates, select countries and languages, and capture local-language signals (policy, logistics, trade) across 80+ languages and 230+ countries. It often surfaces indicators before market feeds.
2) Predictive Core Uses market, macro, and sentiment features (500+ signals) to generate price forecasts up to 18 months. On commodities currently released in our platform, backtests show 90–95% accuracy, outperforming the tools our pilot manufacturers use today. (We can share evaluation details—splits, metrics, baselines—in the comments.)
3) Spot & Forward Prices Combines quoted and non-quoted spot prices with forward curves from major exchanges to give a full time view and a simple way to compare forecasts to market structure.
4) Multimodal What-If Runs instant simulations to quantify exposure and show how price shifts flow into unit costs and margins (useful for make/buy timing and hedging).
Together, these apps give procurement comparable views (signals ↔ forecasts ↔ forward curves ↔ P&L impact) so decisions aren’t made in the dark or too late.
WHAT WE'RE ADDING ON
We’re adding an agent that uses these apps together, so teams can work conversationally instead of hunting through dashboards.
Ask:
- “Why did copper rise this week?” - “What happens to our cost base if gas is +10%?” - “Simulate Chinese export limits on copper.”
The agent runs OSINT → Forecasts → Spot/Forward → What-If, cross-checks the results, and returns a sourced explanation in plain language (in the app or via Teams/Slack). The goal isn’t to replace judgment but to shorten the path to a defensible decision.
WHERE WE ARE
Early access with manufacturers in the EU and US. Next milestones: proactive alerts, deeper ERP/P2P integrations, and public notebooks for long-horizon forecast evaluation.
WHAT WE’D LIKE FEEDBACK ON
- Better ways to evaluate long-horizon forecast quality (beyond MAPE/hit-rate). - How to design explanations users can trust (data traces, competing hypotheses, ablations). - First integration points you’d want (ERP, P2P, Teams/Slack workflows). - Any additional valuable feedback is more than welcome.
LINK
Website: https://www.inaya.ai
Thanks for reading — happy to go deep on the technical details.