# Willy Chavarria Used AI Agents to Restock BFCM Inventory Mid-Sale Without Human Intervention *Agentic commerce let the fashion brand adjust pricing and allocation in real time as orders flowed through Black Friday.* By **Jenny Huang Goodman MPA MSc MHSA, Principal** — The Stash Edge, Hako Shikin LLC. Published 2026-10-20. Canonical: https://www.pops4.com/stash/articles/willy-chavarria-2026-10-20t18-1 Subject: Willy Chavarria Tags: agentic commerce, bfcm, inventory management, dynamic pricing, ai automation, willy chavarria --- Willy Chavarria deployed agentic commerce during Black Friday Cyber Monday 2024, allowing AI agents to manage inventory allocation and pricing adjustments without manual input as orders arrived, according to Modern Retail. The fashion brand handed over live decision-making to software that monitored demand signals and shifted stock between channels while the sale ran. The system tracked real-time order velocity by SKU and geography, then reallocated inventory from slower channels to faster ones and adjusted pricing on items approaching stockout or languishing unsold. When a size or colorway spiked in one region, the agent moved units from warehouses serving slower regions. When an item hit a threshold sell-through rate, the agent raised the price incrementally to preserve margin on remaining stock. The brand set guardrails — minimum margins, maximum discounts, prioritized fulfillment centers — but the agent executed the moves autonomously. This worked because BFCM compresses weeks of demand into hours, creating a data density that overwhelms human merchandisers. Traditional playbooks rely on pre-event planning: set the discount, load the inventory, hope the forecast holds. Agentic commerce inverts that. The system observes actual behavior, compares it to the plan, and corrects in the window when correction still matters. A human sees the stockout report Monday morning. The agent sees the trend Saturday afternoon and reroutes a pallet. The underlying mechanism is closed-loop feedback. The agent receives a continuous stream: orders placed, orders fulfilled, cart adds, cart abandons, inventory positions by node. It compares current state to target state — sell-through rate, margin, channel mix — and fires the cheapest move that closes the gap. Move twenty units from Dallas to Brooklyn. Drop the discount on Item A from 25% to 20%. Raise Item B from 15% off to 18% off to clear the last fifty pieces. Each move updates the data stream, and the loop continues. For a small physical-product brand, the same logic runs on simpler infrastructure. Start with a single high-turnover SKU during a known demand spike — a drop, a promotion, a seasonal peak. Use Shopify or WooCommerce inventory APIs to pull live stock levels and order counts every fifteen minutes. Set two thresholds: if inventory falls below 25% of starting level and orders are still arriving at peak rate, raise the price 10%. If inventory sits above 75% of starting level six hours into the event, add 5% to the discount. Write the rules in a spreadsheet, then use Zapier or Make to connect the API to a pricing app like Bold or Prisync. Cost: domain knowledge, **$50/month** in app fees, and the willingness to let the rule run unsupervised for the duration of the event. The first time, monitor it live. The second time, set it and staff the fulfillment line instead. Test the system on a product with known demand elasticity — something you have sold during prior events, where you have clean data on how price changes moved volume. Avoid your hero SKU the first time. Use a solid secondary item where a misstep costs margin but not the customer relationship. Log every automated decision and compare it to what you would have done manually. After the event, diff the results: did the agent preserve more margin, move more units, or reduce waste compared to your static plan? If yes, expand the ruleset. If no, tighten the thresholds and retest on the next spike. The broader pattern is decision frequency. Brands that win peak events treat pricing and allocation as live variables, not set-it-and-forget-it parameters. Agentic commerce is the industrial version of that discipline, but the discipline itself scales down. A one-person brand cannot run a thirty-node warehouse network, but it can check inventory twice a day during a four-day sale and adjust the discount each time. The question is whether you write the rule once and let it fire, or whether you check your phone between packing boxes and hope you remember. ## The takeaway Run live pricing rules during demand spikes so you adjust before stockouts or waste happen, not after. --- ## Publisher **Hako Shikin LLC** — Virginia Beach, Virginia. Founded 1997. ASI 217876 · DUNS 18-204-6339. Principal and author: **Jenny Huang Goodman MPA MSc MHSA**. - Author: https://www.huanggoodman.com/about - LLM context: https://www.pops4.com/stash/llms.txt - MCP endpoint, for AI agents: https://mcp.pops4.com/mcp - Client dashboard: https://dashboard.pops4.com/ - Catalogue: 70,000+ products, 200+ brands