A retail company is planning an AI agent that recommends personalized promotions to store associates during customer interactions. The grounding data includes a product catalog updated nightly, a loyalty database synced every 15 minutes, and a real-time inventory feed. During requirements analysis, the architect discovers that the promotions logic depends heavily on current stock levels, but the inventory feed frequently lags by several hours during peak hours due to batch processing. Which grounding data quality dimension is MOST at risk and should drive the architect's remediation priority?