You don't have to know what to automate.That's the part we do first.

Most automation tools hand you a blank canvas and assume you already know what you want. We start by finding out.

You talk.
We ask how your agency actually runs, then hand back a ranked list of where the hours are going.
You read.
What we think should happen comes back as a few lines of plain English. Change anything that's wrong.
You watch.
We run it over your last thirty days with everything switched off, and show you what it would have done.
You decide.
Nothing goes live until you press the switch, and one press turns it off again.

Part one · Before you build anything

Nothing is connected and nothing is built. The only question on the table is which hour of your week is worth getting back.

Tell us what your week looks like

We ask about how your agency actually runs. What happens when a new client comes in. What you do every Monday. What fell through the cracks last month.

You don't connect anything, and you don't need to know what any of it is called. You answer questions about your own business.

At the end you get a report: where your hours are going, ranked, with a number against each one.

The audit is free, permanently. No card, and nothing connected to it.

Where your week goes

6.5Chasing proposals
4.0Weekly client reporting
2.5Onboarding a new client
1.5Chasing invoices

Hours a week, across everyone who answered.

You know where your time goes. You don't know where theirs does.

Most founders can name their own worst hour of the week. Almost nobody can name their account manager's.

Send your team a link. It takes two minutes to answer, the answers are anonymous, and we report back what people agree on rather than what any one person said.

9 of 14named weekly client reporting, without being offered it as an option.

Answers are never shown against a name, or against a role small enough to work out whose they were. Nobody sees anybody else's, including you.

Part two · Before it runs

You read what it will do, in your own language, and watch it happen to your real work with everything switched off. Then you decide.

Read it before it exists

Pick something off the audit and we write back what we think should happen: what the process is now, what should happen instead, and what we still need to know.

It reads like a sentence. If we've got something wrong, you say so and it changes. If something's ambiguous, we ask rather than guess.

You agree on the problem before anything gets built, then agree on exactly what it does before it runs. Two separate decisions, on purpose.

Quote follow-up

When: a deal moves to Proposal Sent
Wait: three business days
Only if: it's still sitting there
Send: a follow-up in your voice
Stop: after two

Nothing connects until you've seen the plan

Most tools want access to your CRM before you know whether you even want the thing. We ask afterwards.

One click per app, on the same sign-in screen you already know. Nothing to copy between browser tabs and no settings pages to find.

And we tell you what we're asking for in the same words as the plan, including what we are not asking for.

The permissions we ask for and the sentence describing them come from the same place, so they cannot disagree with each other.

To run Quote follow-up

Read your deals
Notice when your deals change
Change your deals
Send email from your Gmail account

Not asking for

Delete anything
Read your other mailboxes

This is the part nobody else does.

Before it does anything, it does nothing

Before an automation goes live, we run it across your last thirty days with everything switched off. No emails sent, nothing changed. Then we show you what it would have done.

Real clients, real dates, real reasons. And if the second line looks wrong to you, if a won deal should still get a thank-you, then you have just caught something before it reached a client.

Last thirty days, with everything switched off

Would have actedAcme Pty Ltd, email to jane@acme.com, “Following up on the Acme retainer”
Wouldn't have actedBright Digital, because the deal had already moved to Closed Won
Couldn't tellKirribilli Group, because the record doesn't go back far enough
Still in flightSouthbank Media, because the three-day wait lands after this window ends

If we can't work out enough of them to be confident, we say so and tell you how far back to look instead. We would rather say “not yet” than let you switch on something we can't vouch for.

Part three · After it's live

Automations are easy to build and hard to keep. This is the half nobody demos, and it's the half you live with.

One switch, and it's yours

Arming an automation is its own deliberate act. Nothing runs before you press it, and one press turns it off again.

You also get a plain-English note describing what it does, written for whoever is still there in eleven months when you aren't.

Armed since Monday, 9 runs

The note that comes with it

What it does, in the same words you approved
Which accounts it touches
What not to rename without telling us
How to pause it

Not “Dear valued client”

Where an automation sends something, the wording is written fresh each time, in your voice, learned from real emails you have actually sent.

It knows the client's name, what the quote was for, how much it was and when it went out. It doesn't invent a date it wasn't given, or sign off as somebody who doesn't exist.

Elsewhere it can read an incoming enquiry and work out whether it's urgent, or pull the sense out of a long note, so the automation can behave differently depending on what arrived.

AI is included in every plan. There are no credits to buy, nothing to top up, and nothing extra to work out.

Written fresh, not filled in

Subject

Following up on the Acme retainer

Hi Jane, just circling back on the retainer quote we sent through on the 4th, the $4,800 a month one. Happy to jump on a call if that's easier than reading it.

Not: “Dear valued client, we are following up regarding your recent enquiry.”

Things change. You hear it from us, not from a client.

The tools your agency runs on change constantly. A field gets renamed, a permission lapses, an app updates something.

Most of the time we handle it and you never know. When it needs a decision, you get told what stopped, what we think will fix it, and a button to approve.

We never repair something quietly. An automation that fixed itself wrong doesn't fail, it succeeds at doing the wrong thing, over and over, to real clients. So a fix always waits for you.

While something is waiting on you, the automation is paused rather than left half-working.

Paused, waiting on you

The Gmail connection needs signing in to again.

While that's true, anything this would send through Gmail isn't happening. So it's paused rather than left switched on and quietly doing less than you think.

Every version, kept

Every change is kept. You can see what an automation looked like when you approved it, and go back to any earlier version.

The report you read before switching it on is stored exactly as you read it. Same lists, same numbers, same names.

Version 3Running nowYou approved this on 14 March
Version 22 March
Version 119 February

“You approved this” only means something if the thing you approved is still there to look at.

You can work out your bill in your head

You pay for the number of automations you are running. Not credits, not tasks, not per person.

One automation is one thing that starts it

“Follow up on unanswered proposals” is one. “Welcome a new client” is another. Describe three separate things and that's three, and we build them separately on purpose so one breaking never takes the others down with it.

See pricing

Ready to save your team time?

Tell Nrth Star what your week looks like. It works out where the hours go, builds the automations that get them back, and keeps them running.