No forms. No imports. No consultants. Just ten hours of conversation, over ten business days.
Most enterprise software onboarding is designed for the software's convenience, not the team's. Cadence inverts that entirely. A senior coordinator already carries everything Cadence needs to know — in their head. The onboarding programme is a structured way to get it out of their head and into the system.
Ten conversations. Ten things the system learns. One live coordinator on Day 11.
The manager sits down. They talk. The AI listens.
Sessions can be conducted in three modes. The manager chooses what works. There's no preparation, no homework, no wrong answer. Even "I've never really thought about that" is useful signal.
Voice
Manager speaks; AI transcribes, processes, responds in real-time in the browser. Best for thinking out loud — talking is faster than typing and surfaces nuance that structured inputs miss.
Video call
Conducted via Zoom with Cadence listening as a meeting bot. Best for teams with a standing call cadence already, or where two people want to be on the line.
Chat
Text-based conversation in the Cadence interface. Best for managers who prefer to compose thoughts in writing, or for sessions that need to fit in around a noisy day.
Ten sessions. One hour each. Click a day to walk through it.
Each session has a theme and a set of questions. The format never changes: the manager talks, the AI asks follow-up questions to go deeper. By Day 11, the coordinator opens Cadence and the system already understands the business.
The AI is running eight background processes simultaneously, across all ten days.
The manager experiences a conversation. Cadence is building the Agency Brain — a structured, living model of the agency seeded from your own history: connected entities, classified emails, voice patterns, and escalation logic — that continues to evolve after go-live.
Real-time transcription
Speech → structured text · spoken language preserved with context
Entity recognition
Clients, suppliers, campaigns, processes, rules, relationships, preferences
Knowledge graph update
Connecting entities: Client A → Campaign B → Supplier C → Coordinator D
Classification engine training
Map historical emails to the 39-type taxonomy using conversational context
Voice profile building
Extract style patterns from approved sent emails — four-dimension blueprint
Escalation rule generation
Convert verbal rules into structured routing logic and overrides
SOT schema mapping
Map existing tracking fields → Campaign Hub structure for ETL migration
Supplier memory seeding
Initialise relationship records for every supplier contact found in history
Cadence reads the email history while the conversations are happening — under scoped, revocable consent.
From Day 1, Cadence connects via the Gmail API with delegated OAuth permissions scoped to the minimum required. By Day 5, the system has processed weeks or months of history and the manager reviews what it's inferred. Corrections become the first training signal in the voice diff engine.
- ✓ All emails sent from team members to supplier/talent contacts
- ✓ All supplier/talent replies received
- ✓ Thread history — full back-and-forth of each conversation
- ✓ Subject lines, metadata, classification signals
- ✕ Personal emails or internal team threads outside scope
- ✕ Cross-tenant data sharing of any kind
- ✕ Email content used to train any global model
- ✕ Data leaving the isolated tenant environment
Privacy and consent — by design
Email ingestion operates on delegated OAuth permissions, scoped to the minimum required (gmail.readonly for Gmail — more providers in the pipeline). The manager explicitly consents to scope during Day 1 setup. Ingestion can be paused or revoked at any time. No email content leaves the agency's isolated tenant environment.
What every team has, the morning Cadence goes live.
The 10 days produce seven concrete outputs. From the next email that lands in the inbox, the system is operational.
The team that finishes onboarding on Day 10 will be using a better AI on Day 90 — without anyone running another session.
Every coordinator edit feeds the voice diff engine. Every escalation override adds a data point to the routing model. Every supplier interaction extends the memory. The system learns continuously, from real work.
From Day 11, the coordinator's job changes. They stop managing email. They start managing outcomes.
Eight email shapes. Written from your last twelve months of sent mail.
Most AI email tools ship with generic templates. Cadence doesn't. During onboarding, the system reads your last twelve months of sent emails and synthesises eight templates — written in your structural pattern. Your openers. Your sign-offs. Your way of asking for a deadline. From Day 11, every draft uses these as the skeleton.
Campaign opportunity
First-touch outreach to a new manager or talent — brief teaser, soft commercial framing.
View template →Indicative talent request
Asks for a shortlist with rates ranges, before any creator names are exchanged.
View template →Rate request
Direct ask for a fee on a named creator + deliverable set.
View template →Follow-up
Time-bounded nudge that mirrors how you write your own follow-ups.
View template →Negotiation counter
Reframes the ask with your characteristic concession ladder.
View template →Contract send
Cover note for legal — the exact phrasing you've used before.
View template →Content approval
V1 review, edit-request and sign-off — all in one shape.
View template →Confidential brief
Cover note for any brief carrying a confidentiality clause — calibrated to first-contact level.
View template →If your inbox is empty (new coordinator, first day), Cadence falls back to a baseline set you can edit on Day 8. Either way, by Day 11, the templates are yours.
Ninety minutes. One conversation. A working AI by the same afternoon.
The 10-day journey is the way Cadence learns you properly. The Fast Path is for when "properly" isn't an option. One ninety-minute session. Six phases. Background email ingestion runs while you talk. By the end of the session, Cadence has a working model of your voice, a seeded relationship matrix for your top contacts, a populated campaign hub, and a baseline classification calibration. You walk away with a product that drafts emails the same afternoon — not at Day-11 accuracy, but at Day-1 accuracy, which is enough to start.
The honest trade-off A Fast Path coordinator on Day 1 sits at roughly the same accuracy as a 10-day coordinator on Day 1. The difference shows up over the next month — the 10-day coordinator improves faster because more signal was captured. Both models converge by Day 90. The Fast Path is for people who can't wait ten days to start. It's not the better choice. It's the available choice.
By minute 90, you walk away with:
- 01 A voice profile encoded across four axes
- 02 A seeded relationship matrix with 12 dimensions across 20+ contacts
- 03 A populated campaign hub with brief summaries and navigation indexes
- 04 A baseline classification calibrated to your inbox patterns
- 05 Configured routing thresholds and hard-stop guardrails
- 06 A first draft preview you've already approved or edited
- 07 A scheduled morning briefing for the next day
- 08 The full product available immediately
Day-by-day accuracy both paths converge by Day 90
| Day 1 | Day 10 | Day 30 | Day 90 | |
|---|---|---|---|---|
| 10-day journeyrecommended | 68–72% | 78–82% | 88–91% | 93–96% |
| Fast Path90 minutes | 42–48% | 56–62% | 74–80% | 86–90% |
The Fast Path coordinator catches up because they're using the product daily — every edit feeds the nightly voice diff agent, every reply feeds the relationship matrix. The trade-off isn't permanent. It's a head start vs a flat start. Both end up running.
What gets deferred to "in-product nudges":
The Fast Path covers what you need to start. The product fills in the rest as it encounters real situations — one question at a time, in context.
- Deep client-side mapping — top 5 client contacts seeded; the rest populate from inbound mail. Day 7 nudge picks up the new ones.
- Edge-case walkthroughs — only the most common 6 configured upfront. Rare cases surface as in-product modals on first encounter.
- Full SOP customisation — default SOPs used initially; Day 14 review prompts a walkthrough.
- Brand-by-brand mandatories — top 3 brands captured; additional brands prompted on their first email.
- Multi-coordinator workflow — single-coordinator setup by default. Teammates trigger a 10-minute mini-onboarding.
Three things people ask:
Isn't 90 minutes too short to learn me properly?
It is. That's why we tell you the accuracy is lower than the 10-day path. The product keeps learning every day after onboarding — the Fast Path just starts you on the curve sooner.
What if I get something wrong during the session?
Everything you tell us is editable from Settings. Nothing said during the Fast Path is locked in.
Can my team and I both do the Fast Path?
Yes — each coordinator runs their own session. Voice profiles, relationship matrices and routing configs are per-coordinator. Shared campaigns and contacts get merged at the workspace level.