The same budtenders who inspired the AI Budtender had another complaint: answering the phone while actively serving in-store customers. Callers asking about store hours, directions, website ordering, and product availability were pulling staff away from the floor during the busiest moments.
The fix seemed straightforward — until the constraints of the cannabis industry made it interesting.
Pricing can’t be given over the phone due to age verification requirements. Delivery and pickup orders go through the website. And the most common caller question — “do you have this product?” — can’t be answered by AI either, because inventory turns so fast that only a budtender physically checking the shelf can give an accurate answer.
So the system was designed around those constraints. Routine queries are handled directly. For product availability, the AI collects the caller’s name and up to 5 items, then executes a warm transfer to a budtender with full context already in hand — no re-explaining, no repeated questions.
What It Does An AI voice receptionist handles inbound calls across three dispensary locations, fielding routine inquiries and routing product availability calls to budtenders with customer name and product list already captured. A callback flow manages overflow when all budtenders are occupied.
The Stack
- Platform: Retell AI
- Model: GPT-4.1 in Rigid Mode
- Architecture: Central 888 router (“Lily”) → location-specific agents
- Automation: Make.com + Twilio (explored, ultimately not needed)
- Data: Google Sheets for call logging and callback tracking
A Key Pivot The original design included an SMS notification system — AI collects inquiry, texts budtender, budtender texts customer back. Budtender and manager feedback killed it fast: direct customer relationships are a core competitive advantage. The warm transfer approach preserves that human element while still reducing the repetitive intake burden.
The Honest ROI With the main store averaging 540 calls/month and a satellite location at 207 calls/month, the current financial ROI is modest — approximately 3.8% annually at the main. The system runs at a small monthly shortfall relative to labor savings alone.
The real case is elsewhere. After-hours coverage at no additional cost. Consistent caller experience across all locations. Budtenders freed to focus on in-store customers. And call pattern data that didn’t exist before — peak call times, most-requested products, common questions — intelligence that can inform staffing, purchasing, and promotions.
The infrastructure cost is largely fixed. As call volume grows, the unit economics improve. And the system has natural room to expand: operational analytics surfaced to management, an after-hours concierge that turns a dead-end call into a retained customer, and proactive outbound callbacks that close the loop without requiring the customer to call again. None of these require rebuilding — they’re extensions of what’s already running.
The 3.8% ROI measures what the system does today. It doesn’t account for what it becomes.
Status: Currently in testing with management — live validation coming soon.
[The juicy details are forthcoming in the Blog Section]