Concept evaluation · v1 · March 2026

Campaign communications, handled by AI. The work that pays starts again.

Cadence sits between your inbox and your campaign data. It reads every email, drafts the reply in your voice, and routes each thread by financial, legal and relational risk. A human stays in the loop on every decision that matters.

0%
of a coordinator's day spent on email admin today
$0
AUD lost per coordinator, per year, on zero-revenue email
0
emails to deliver a single mid-size influencer campaign
0
email types Cadence classifies, routes and drafts against

Agencies run on relationships. The email that sustains them earns nothing and eats 65% of a coordinator's day.

How it works

Three moves, on every email.

Cadence does one job: it takes the email load off your coordinators without taking away their judgement. Each incoming email goes through the same three moves.

01 / CLASSIFY

It reads the email.

Every incoming email is classified into one of 39 types, checked against 24 risk flags, and matched to its campaign.

Go deeper: classification →
02 / DRAFT

It writes your reply.

The draft sounds like you, not like AI. It is grounded in live campaign data and everything the agency has learned before.

Go deeper: the memory →
03 / ROUTE

It knows what it can't decide.

Routine mail can send itself. Most drafts wait for a one-glance approval. Anything commercial, legal or sensitive goes straight to a human.

Go deeper: the MVP →
Pick an email
What Cadence does with it
The relationship matrix

It knows exactly who it's writing to.

Every contact carries a living relationship score: twelve dimensions, rescored after every reply, re-read every Sunday. Not a CRM field somebody forgot to update. A memory that moves.

And the voice moves with it. The same chase-up relaxes for a partner of three years, tightens for a difficult negotiator, and stays properly formal for someone you met yesterday. Pick a contact and watch the draft change.

Go deeper: the twelve signals →
Same task · drafted for this person
The agency brain

Ask it anything. It answers from memory, with receipts.

Cadence builds an institutional memory that's yours alone: your brands and their rules, your playbooks, every rate ever quoted, every promise ever made. It reads that memory before it drafts a single word.

Ask a question in plain English and get the real answer, cited to the exact email or record it came from. Week one it's helpful. Week twelve, switching it off would mean losing the brain.

Ask Cadence
Ask me anything about your campaigns. I read the Agency Brain and the Fact Ledger before I answer. Try a question below.
39
email types classified & routed
explore →
36
deal archetypes understood
explore →
24
risk flags that can escalate
explore →
learns from every edit you make
see the curve →
The morning briefing

Your day is triaged before you sit down.

An agent runs before you open the app. It builds a stack of cards for your day: drafts that need a glance, go-lives in the next two weeks, budgets near their envelope, anything else that needs you.

You swipe through. The stack gets shorter. By the time it's empty, every decision the day needed is made, and you haven't opened the inbox once.

Swipe to dismiss Runs before your work-start time Instant on-demand Chat-scheduled briefings
› “Brief me now”
‹ here's your stack: 4 cards, 2 need you.
› “Give me a briefing at 3pm”
‹ done. the next stack is ready at 3pm.
Morning briefing 07:42 · Tuesday
→ Drag to dismiss Tap Open queue for drafts
The review loop

From a 90-second rewrite to a 3-second glance in 30 days.

Every edit you approve teaches Cadence your voice. Day 1, a draft takes about a minute and a half to read and rewrite. By Day 30, most drafts are a three-second glance-and-approve. Hover the curve to see any day.

Day 1
~90s
Read the draft, rewrite it in your voice
Day 7
~34s
Lighter touch: openers and sign-offs already match
Day 14
~12s
Skim, tweak a line, approve
Day 30
3s
Glance and approve. It writes the way you would
THE BUSINESS CASE

65% of a coordinator's day earns the agency nothing. Cadence gives it back.

A coordinator costs AUD $112K a year, fully loaded, and most of that buys email a trained system can draft. Recover it and the same team runs 67% more campaigns, with margin that finally expands.

The time imbalance

A mid-size activation takes 300 to 500 emails. Around 24 hours of a human's time, per campaign, on correspondence.

The output of that time is not a worse coordinator. It is a coordinator who never gets to do the work that pays. Toggle the chart to see where the day goes, and where it could go.

Coordinator time allocation
Source: Cadence internal modelling · Australian mid-size agency average · n=12 coordinators
Traditional scaling

Linear headcount, flat margin.

Campaigns↑ 50%
Staff↑ 50%
Cost base↑ 50%
Marginflat
Revenue grows. Profit doesn't.
Cadence scaling

Capacity grows. Cost base sits still.

Campaigns↑ 67%
Staff0 hires
Cost base+ Cadence licence
Marginexpands
3 coordinators → 30+ campaigns / quarter.
The Harvey AI parallel

Harvey didn't replace lawyers. It removed their lowest-value work, then built an $11B vertical AI business in three years.

Cadence applies the same thesis to marketing agencies: keep the coordinator, remove the 65% of their day that produces no client value. Same model. Same defensibility: voice data, relationship memory and workflow context that compound per agency.

$0B
Harvey AI valuation, 2026
$0M
ARR · 3 years from launch
0%
of coordinator time we recover
$0
AUD recoverable per coordinator / yr