Case Study: How a Pharmacy Chain Cut Near-Expiry Medicines by 38% with an AI Agent

1. Pain – Imbalanced stock and rising near-expiry medicines

Minh, who manages a 35-store pharmacy chain in Ho Chi Minh City and nearby provinces, faced a familiar contradiction: one store in Binh Duong ran out of fast-moving medicine while another store nearby held excess stock approaching expiry.

Every week, the warehouse team had to pull data from the point-of-sale system, Excel files and internal chat groups to compare quantities, expiry dates, sales velocity and incoming shipments. The process was still run by people manually. One delayed file, an absent employee or a computer power failure could make the entire report unreliable.

The biggest risk was not losing a few boxes of medicine. It was missing the window to transfer near-expiry stock, forcing the chain to destroy it while capital remained tied up in inventory and high-demand stores lost sales because products were unavailable.

2. Agitate – Vanity KPIs cannot protect margins

Previously, the management team spent almost two days on every inventory reporting cycle. By the time the report was completed, the data could already be outdated. Employees kept patching the process, reconciling spreadsheets and explaining discrepancies instead of focusing on quality control and product portfolio decisions.

This is half-baked optimization: reports exist, but timely action does not. If the chain continues with short-term fixes, technical debt, vanity KPIs and avoidable losses from expired stock will accumulate. In a simulated case study involving 35 pharmacies, the value of near-expiry medicines requiring disposal fell by 38% after three months, equal to an estimated annual saving of VND 420 million. Local stockout incidents also fell by approximately 25%.

3. Solve – Deploy the AI Agent in 3 controlled steps

Step 1 – Consolidate reliable data. HimiTek connects sales, inventory, expiry and inbound shipment data from every branch. The Agent standardizes records by SKU, batch and location, then ranks products into three groups: near-expiry risk, stockout risk and monitoring required.

Step 2 – Forecast and recommend transfers. The AI analyzes sales velocity, seasonality and local demand to recommend moving medicine from stores with excess stock to stores that can sell it before expiry. Regional managers receive prioritized alerts and pre-filled transfer proposals for review and approval.

Step 3 – Block unsafe execution. The workflow follows the rule: AI recommends, people approve. OpenClaw Gatekeeper controls tool access and commands, while rate limiting, automatic API key rotation and hard budget caps, such as 5 USD per month for each virtual key, reduce the risk of runaway loops. The Reasoner is separated from the Actuator so prompt injection cannot take over the entire server.

def recommend_transfer(stock, sales_30d, days_to_expiry):
    daily_rate = max(sales_30d / 30, 0.1)
    projected_days = stock / daily_rate
    risk = days_to_expiry < projected_days + 14
    return {'action': 'review_transfer' if risk else 'monitor',
            'projected_days': round(projected_days, 1)}

# The AI only recommends; a manager must approve before creating a transfer order

• Validate input data and enforce role-based access.
• Set alert thresholds for expiry, minimum stock and sales velocity.
• Track monthly: disposal value, reporting time, stockout rate and approval rate for recommendations.

After deployment, inventory reporting time fell from two days to 30 minutes per cycle. The warehouse team spent less time entering and reconciling data and more time on quality control and assortment management.

4. CTA – Start with a process that can be measured in money

A pharmacy chain does not need to design a massive AI project on day one. Start with inventory, near-expiry medicines and inter-store transfers. HimiTek can help set up a pilot for selected branches, define measurable KPIs and prove savings before the workflow is expanded.