How Cadence Works
The engine under the hood.
Most tools show you a dashboard. Here's what's happening behind it.
The Five Cycles
Always running. Always learning.
Every 30 min
Discovery
Researches conversations across 11 platforms. Identifies people worth knowing, posts worth joining, and opportunities your competitors are missing.
- Searches keyword and hashtag feeds every 30 minutes
- Monitors topic conversations across all connected platforms
- Identifies new people worth knowing in your space
- Feeds relevant posts to the Engagement cycle
Every 15 min
Engagement
Reads every mention, scores relevance, drafts replies using your voice and your full relationship history with each person. Detects buying intent in real time.
- Scores every mention and discovered post 0-10 for relevance
- Drafts replies using your voice profile + knowledge base + relationship history
- Detects buying intent and includes your CTA when appropriate
- Recommends likes, reposts, and follows with confidence scores
Every 6 hours
Content
Creates original posts that have never been said before. Searches 6 layers of memory to avoid repetition. Rotates through 10 creative angles per topic. Checks what your other brands are posting to prevent overlap.
- Checks trending topics in your content pillar space
- Searches memory to avoid repeating angles or topics
- Coordinates with sibling brands via the Mesh
- Scores predicted performance before proposing
Daily
Growth
Analyzes your entire relationship graph. Identifies which connections are warming up, which are cooling off, and who you should prioritize this week.
- Analyzes your full relationship graph daily
- Identifies which relationships are warming vs cooling
- Generates weekly top-10 engagement recommendations
- Suggests follows based on relevance and reciprocity signals
Daily
Research
Finds and evaluates articles from RSS feeds, PubMed, and industry sources. Drafts posts about high-relevance findings. Only proposes content that matches your pillars.
- Fetches articles from RSS, PubMed, and other configured sources
- Evaluates each for relevance to your content pillars
- Drafts posts for high-relevance articles
- Checks memory to avoid covering topics recently posted about
Six Layers of Long-Term Intelligence
Most AI tools forget everything between sessions.
Cadence doesn't.
01
Content Memory
Every post, reply, rejection, and edit — remembered with topic, creative angle, and performance. Your team never repeats itself, even months apart.
Your content stays fresh without you tracking what's been said.
02
Negative Memory
Your rejections, your edits, and posts that underperformed. Three kinds of 'don't do that again' that compound over time.
The same mistake never happens twice.
03
Relationship Memory
Full history with every person — interaction count, first and last dates, trajectory (warming/stable/cooling), weekly narrative summaries.
When replying to someone, your team knows your entire shared history.
04
Performance Memory
Weekly, monthly, and quarterly summaries. Which topics get engagement, which platforms outperform, trend direction.
Strategy informed by data, not guesswork.
05
Audience Patterns
Behavioral rules extracted from your data — format fatigue detection, timing patterns, engagement triggers.
Your team stops doing what your audience is tired of.
06
The Dream Journal
Every Sunday, the most powerful AI reads across all memory layers and generates 3-5 strategic insights that connect patterns you can't see.
Weekly intelligence that no human could synthesize manually.
“Every time you post clinical data followed by a personal story within 48 hours, the personal story gets 3x more engagement. You haven't posted a patient story in 12 days.”
— Example Dream Journal insight
Cross-Brand Intelligence
The Mesh
Multiple brands, one coordinated presence. The Mesh ensures your brands never step on each other — automatically, with no setup required.
Engagement Collision Detection
Before one brand replies to someone, Cadence checks if another brand already engaged recently. If so, it escalates for your decision — not both brands piling on.
Scheduling Stagger
Brands don’t post on the same platform at the same time. Automatic 4-hour gap between posts from sibling brands. You see the stagger in your approval queue.
Content Dedup
Before publishing, Cadence checks if another brand has similar content queued. Too similar? Escalated with a side-by-side comparison and similarity score.
10 Code-Level Safety Checks
The Brand Guardian
Pure code. No AI. Cannot be bypassed, tricked, or softened. Every action your team proposes passes through these checks before it reaches your queue.
Platform capability
Can this platform do this action type?
Permission allowlist
Is this action type enabled for this account?
Banned words
Does the content contain any word from your banned list?
Banned topics
Does the content touch any topic you've marked off-limits?
Required disclaimers
Are required disclaimers present for regulated content?
Character limits
Does the content exceed the platform's maximum length?
Duplicate detection
Has a similar action been taken recently?
Blacklist
Is the target account on your blacklist?
Topic drift
Does the content match at least one of your content pillars?
CTA frequency
Is the booking link being overused? Max 1 per person, 2 posts per day.
The AI proposes. The Guardian gates. You decide.
Buying Intent Detection
Turn conversations into appointments.
When someone says “how much does this cost?” or “can I book an appointment?”, your team recognizes the buying signal and naturally weaves in your booking link.
Without CTA
Someone asks about your services. Your team replies with helpful information in your brand voice. Good engagement — but no next step.
With CTA
Same helpful reply, ending naturally with: “If you'd like to learn more, you can book a free consultation here.” The booking link is tracked with UTM parameters for attribution.
Built-in guardrails
- ~30% of posts — not every post includes the CTA
- Max 1 CTA per person in 24 hours
- Max 2 CTA posts per day
- Context-aware — won’t pitch in condolence threads or complaints
- Every link gets UTM tracking for attribution
Your Level of Control
Three modes. You choose how much to trust.
Weeks 1-4
Full
Every action goes to your approval queue. Nothing happens without your OK.
Best for: Starting out. Building trust with your team.
Month 2
Smart
Low-risk actions (likes, follows) auto-execute. Replies and posts still need your approval.
Best for: After you trust your team's judgment on simple actions.
Month 3+
Auto
Everything auto-executes. The Guardian is the only gate. You review results in your morning briefing.
Best for: When your team has proven they sound like you.