Platform architecture

The cannabis industry's AI, end-to-end.

Three layers working together: an intelligence layer that understands your market, an execution layer that generates and dispatches compliant campaigns, and a network layer that makes every dispensary smarter with every week that passes.

Layer 1 - Intelligence Layer 2 - Execution Layer 3 - Network moat

Three layers. One pipeline. Fully autonomous.

Most marketing software is a collection of disconnected tools. Canovate is a single orchestrated system where each layer feeds the next - from raw market signals all the way to a dispatched campaign with a delivery receipt.

1

Intelligence layer

Ingests your market before every cycle. Real-time menu scraping pulls live inventory. The scorecard inspector reads your public web presence. The chat orchestrator classifies intent and routes queries to the right agent. A knowledge graph stores competitor signals, keyword opportunities, and local event hooks - all refreshed weekly.

Chat orchestrator Scorecard inspector Menu scraper Competitor tracker Market signal feed
2

Execution layer

Eight specialized agents run in a fixed pipeline order every week. Each agent reads outputs from the prior agent, so the writer agent has the strategy thesis, competitor context, and local event hooks before drafting a single word. Every artifact passes through the compliance linter before reaching the approval queue. Dispatchers handle sending to SMS, email, Twitter, and Google Business Profile.

8 specialized agents 27-state compliance linter SMS dispatcher Email dispatcher Twitter v2 API Google Business Profile
3

Network intelligence layer

Anonymized, aggregated signals from every campaign across every dispensary on the platform feed back into the system. Best send windows by market. SMS length vs click-through rate curves. Cadence vs revenue correlation. The more dispensaries run Canovate, the more precise these benchmarks become - and every dispensary benefits automatically.

Cross-dispensary benchmarks Send-window optimization CTR by channel State-level compliance signals

Eight agents. One pipeline. One week.

Each agent owns a specific job. They run in sequence so downstream agents have full context from upstream ones. No agent writes blind.

Strategy agent
Sets the weekly thesis and channel priority ranking
Pulls from: menu products, competitor deltas, market signals, prior campaign performance
Outputs: weekly thesis, featured SKUs, audience segment, channel rank
Research agent
Surfaces competitor moves, keyword gaps, and market shifts
Pulls from: competitor URLs, scrape deltas, price change events, keyword signals
Outputs: competitor delta report, keyword opportunities, market shift flags
Scout agent
Finds local events, partnership angles, and trending hashtags
Pulls from: local event feed (14-day window), city context, strategy thesis
Outputs: 3 event hooks, 2 partnership angles, hashtag set
SEO / GEO agent
Generates blog brief, two Google Business Profile posts, and schema markup
Pulls from: keyword opportunities, local event hooks, dispensary location data
Outputs: blog brief, 2 GBP post drafts, JSON-LD schema
Writer agent
Drafts weekly email and SMS campaigns
Pulls from: strategy thesis, featured SKUs, brand voice, compliance state rules
Outputs: email subject + body, SMS copy (compliant, opt-out included)
Social agent
Composes tweet, Instagram caption, and carousel concept
Pulls from: scout hooks, featured SKUs, brand hashtags, audience segment
Outputs: tweet (280 chars), IG caption, carousel slide copy
Reviews agent
Drafts GBP review response templates and flags risk reviews
Pulls from: recent review data, brand voice, compliance rules
Outputs: response drafts per review, risk flags for owner attention
Marketing agent
Writes the weekly plan and Friday recap email (runs last)
Pulls from: all agent outputs, performance data, campaign dispatch status
Outputs: weekly marketing plan, Friday recap email (400 words max)

27 states covered. Blocks fire before dispatch.

Cannabis advertising rules vary by state. The compliance linter runs on every artifact before it reaches the approval queue - checking medical claims, intoxication language, audience targeting rules, and channel-specific restrictions. Hard blocks stop the artifact. Warnings surface to the owner. Disclaimers auto-inject when required by state law.

27
US states covered (adult-use + medical)
DC
Plus DC I-71 gray market rules
4
Violation categories: block, warn, auto-inject, federal
Weekly
Rule pack update cadence as regulations change

Adult-use states

AK AZ CA CO CT DE IL MA MD ME MI MN MO MT NJ NM NV NY OH OR RI VA VT WA DC

Medical-only states

FL GA NH OK PA TX NC

Example rules in action

CA - block: no health claims on cannabis products. "This flower reduces anxiety" is flagged and the artifact is held.
CO - warn: intoxication language in social copy ("get high", "stoniest deals") surfaces a warning for owner review.
NY - auto-inject: "For use only by adults 21 and older" disclaimer is appended automatically to every customer-facing artifact.
SMS - block: all outbound texts auto-include opt-out language. Missing opt-out = hard block before send.
Federal - warn: crossing state lines, DEA scheduling references, or FDA drug claims in any artifact triggers an escalation flag.

The moat gets deeper every week.

"Every campaign sent through Canovate makes every dispensary smarter."

Performance signals from every sent campaign - open rates, click-throughs, SMS reply rates, GBP engagement - feed back into anonymized benchmarks. No dispensary's data is ever exposed. But every dispensary benefits from the aggregate. A shop in Denver gets smarter from what works in Portland. A Tuesday 4pm send window that consistently outperforms in MI shows up as a recommendation for a new shop in the same market.

Anonymized benchmark example

Best SMS send window (West) Tue 2-4pm PT
Avg SMS CTR (120-160 chars) 8.3%
Email open rate - flower promo 34.1%
GBP post clicks - with CTA +62% vs no CTA
Campaign-to-revenue correlation Bloom: +18% week-over-week
Built on
A Anthropic Claude (foundation model)
Pg PostgreSQL (Neon serverless)
Rnd Render (cloud infrastructure)
@ Transactional email infrastructure
SMS Carrier SMS infrastructure
Live intelligence layer

See real market data in action

Competitor landscapes, live pricing benchmarks, and compliance rules for 10 US metros — publicly available, no login required.

See all briefs →
Live Denver market brief → Live Los Angeles market brief → Live New York market brief →

Ready to see it work on your dispensary?

Start with a free audit. No credit card required. Your first campaign runs within 24 hours of onboarding.