Skip to content
Growth Machines

AI agents built inside your own Claude or OpenAI workspace

Growth Machines builds AI agents and workflows inside your own Claude or OpenAI workspace, or in tools you already use, like Google Sheets. They handle research, outbound drafting, CRM entry, reporting and content, and a person on your team approves what goes out. We are based in Montreal and build in French and English.

Run it in the agent

Which of your team's tasks should AI take first?

Give the agent your website, then list the tasks your team repeats and the hours they take. It maps each one by effort, time back and risk, and writes the first three set-ups. Free, with a free account.

What we do

Built inside your own Claude or OpenAI workspace, or in tools you already use, like Google Sheets.

Research and enrichment

Agents that gather and organize the research your team now does by hand.

Outbound drafting, with agents built on Grok Bot

Outbound messages drafted by agents built on Grok Bot. A person on your team approves each draft before it is sent.

CRM entry and reporting

The records and reports your team updates every week, filled in by a workflow instead of by hand.

Content drafting in your brand's voice

Drafts written in your brand's voice for your team to edit and approve. Inscora's marketing team of one publishes three posts a week with a drafting workflow it runs itself.

Bilingual content workflows

French and English content produced in one workflow. Altitude Sports' bilingual product content now goes out the same day.

Results from this work

As published on davidzagury.ca/consulting.

Altitude Sports

E-commerce · outdoor retail
Same dayturnaround, down from 2 to 4 days

Bilingual product content goes out the same day instead of two to four days later, and the external partner's fee is gone.

The problem. Bilingual product content went through an external partner: two to four days per batch, and no way to check it.

What we did. A Google Sheets-native automation. One script, one spreadsheet, nothing new for the team to learn.

50%+
faster end to end
$10Ks
saved per year

“Tu as parfaitement compris nos problématiques internes complexes. La solution concrète que tu nous as proposée nous apporte déjà une aide précieuse au quotidien.”

“You understood our complex internal challenges perfectly. The solution already helps us daily.”

Hélène Soares, Altitude Sports

Inscora

Cyber-insurance startup
3 / weekon 15 minutes of their time

A marketing team of one now publishes three posts a week on fifteen minutes of their own time.

The problem. A category to own and one person to do it. Content happened when there was time, which was never.

What we did. Built the whole content system: a voice guide, four content pillars, a 72-post calendar, and the drafting workflow the team runs itself.

72
posts planned across six months
< 5 min
from rough bullets to a finished post

The details

Where it usually goes wrong
  1. Agents nobody reviews

    Output goes straight to clients and the first mistake costs more trust than the agent saved time.

  2. Built outside your accounts

    Your data leaves for a vendor's platform and the work disappears when the contract ends.

  3. Pilots that never reach daily use

    A demo works once, then nobody owns it and the old way returns.

  4. No proof it saves time

    Without tracking, an agent's value is an opinion.

The first 90 days
  1. Build

    From the shortlist

    • The first workflows from the workflow map, built inside your licenses
    • A person approves every draft that goes out
  2. Calibrate

    To your bar

    • Quality tuned with your team on real output
    • Fixes made on the cases your team flags
  3. Hand over

    Yours to run

    • Training for the people who run each workflow
    • Workflows in use, drafts sent and hours saved tracked
What you get
  • Agents and workflows built inside your own Claude or OpenAI workspace, or tools you already use
  • Outbound drafting agents on Grok Bot, where outreach is in scope
  • A person approving every draft that goes out
  • Calibration on real output, with fixes on the cases your team flags
  • Training for the people who run each workflow
  • The agent instructions and settings, documented and yours
  • Workflows in use, drafts sent and hours saved reported to you monthly
Calibration

We tune quality with your team on real output until it meets your bar.

A person approves every draft that goes out
Nothing reaches a customer or your CRM without review.
Fixes made on the cases your team flags
We tune quality with your team on real output until it meets your bar.
Adoption and handover

Your team runs what we built, and you can see that it does.

Training for the people who run each workflow
So the workflows keep running after we step back.
Workflows in use, drafts sent and hours saved reported to you monthly
Usage and tool spend before and after, where you can check them any time.
What we measure
Workflows in use
Built is a start; in daily use is the result.
Drafts sent
Output that went out after a person approved it.
Hours saved
Against the baseline counted in the workflow map.
Cases flagged and fixed
How calibration closes the gap to your bar.
Where the work happens
  • Claude
  • ChatGPT and the OpenAI platform
  • Grok Bot
  • Google Sheets
  • Your CRM
Why Growth Machines
  • We build inside your own licenses, so your data stays in your accounts and your team keeps what we build.
  • A person on your team approves what goes out, before anything is sent.
  • The collective behind Growth Machines includes a Head of Software Development.
  • David Zagury, the founder, built MediaPlan.ca, an AI-powered media planning system.
Who it is for · What we need from you
  • Teams with repetitive work in outreach, research, CRM entry or reporting.
  • Companies already paying for Claude, ChatGPT or OpenAI seats.
  • Leaders who want it done well and safely, with a person approving what goes out.

Step by step write-ups of the tasks, your tool list and costs, seat details, and read-only access where needed. One short intake covers it all.

How fees work

A fixed fee, agreed before work starts. No hourly billing.

Read-only access, built inside your own licenses, scoped after the discovery call.

The guarantee

If your team is not using a system within 60 days, we rebuild it for free until they are.

Written into the agreement, not only this page.

Questions

Where does our data live?

In your accounts. We build inside your own licenses and connect read-only where we need data.

Which tools do you build in?

Your own Claude or OpenAI workspace, or tools you already use, like Google Sheets.

Who approves what the agents send?

A person on your team, before anything goes out.

Do we need new software?

Usually not. We start with the tools you already pay for. The Altitude Sports automation runs on one script and one spreadsheet.

Can the agents work in French?

Yes. We build bilingual content workflows. Altitude Sports' bilingual product content now goes out the same day instead of two to four days later.

What happens after the build?

A calibration period with your team, then adoption reported to you monthly.

See where you stand in minutes.

Open the growth agent and give it your website and your goals. It reads your site, looks at your first screen like a visitor, asks ChatGPT, Gemini and Google AI Mode what they say about you, and ranks what to do. Then do the work with your team, or hand it to the studio.