How we deployed Claude to an international team of 40 at MacroActive
A coaching-software company with a team across New Zealand, the UK and Latin America wanted Claude on every seat, the usage data in their own hands, and the governance to match. We installed it in a day, on their own database, and built the library on top.
40
Claude seats, one international team
3-5 / day
Sales calls booked on autopilot
Tens of $k
Saved a year on rebuilt subscriptions
TL;DR
MacroActive builds coaching software for creators, with a team of around forty across New Zealand, the UK and Latin America. The founder wanted Claude deployed properly, the usage data kept in the company's own hands, and the governance to show for it. We installed the harness on site in one day, on MacroActive's own database with their own branded dashboard, and the first usage data was flowing within the hour. On top of it: a skills library for six departments, every session attributed to a named person, and a register of what each skill may reach. Underneath it: a voice SDR booking three to five sales calls a day on autopilot, and software subscriptions rebuilt in house.
What they were paying for
MacroActive runs on Slack, Jira, Duda and Google Workspace, with a large share of the team on ChatGPT and the rest on Claude. Like most teams that size, a few people had got genuinely good at it, and the founder had no way of knowing who, on what, or whether it was doing anything for the business.
He also had a condition before anything went in: he wanted to know exactly where the usage data would live. That is the right question, and most vendors cannot answer it. The answer here was their own database, in their region, under their own key.
What we installed
Two things land in their Claude organisation, both set up once by their admin, the same two we install for every business on the AI Operating System. A plugin that puts the team's workflows on every seat as skills, and a Brain connector that sits in the connector list next to Slack and Jira. Those connect Claude to their systems. The Brain connects it to the company.
The install ran on site at their Auckland office. The collector, the telemetry tables and the dashboard were stood up on MacroActive's own infrastructure, wearing MacroActive's brand, and the first data arrived within the hour. It is the reference deployment for every client who needs their data to stay theirs.
| Piece | What it is | At MacroActive |
|---|---|---|
| Skills library | The team's real workflows, written as skill files and installed on every seat as one plugin. A release reaches everyone at once. | liveEight skills in the library we maintain and three the team wrote themselves, on one plugin for every Claude seat. |
| Brain | A wiki of how the business runs, served to the assistant as a connector so it looks things up before it guesses. | in buildThe company wiki, in their own repository, served to Claude as a connector next to Slack and Jira. |
| Roles and access | Which skills and context each role gets, on top of the client's own directory. | liveEvery session attributed to a named person, so the usage report says who, not just how much. |
| Approval gate | Anything that changes the business waits for a human. Read is open, write is held. | liveSkills read Slack, Jira and the calendar freely. A ticket, a page or a message that leaves the building waits for a person. |
The skills
11
Skills deployed and counting
40
Claude seats, one international team
8
People named in a week
Same day
Install to first usage data
Eight of those are in the library we maintain, and three the team wrote themselves. The library is written from watching the team, not from a workshop: the usage data shows which jobs recur, who does them, and where the generic approach wastes calls, and each skill pins the answer.
slack-to-jira
Product · every requestA product request arrives in Slack and leaves as properly linked engineering tickets. The skill carries the project keys and link types, so it does in two calls what a generic tool spent six rediscovering.
SlackJirarev-evasion-snapshot
Finance · every MondayThe weekly read of revenue at risk, run on a schedule and posted to the channel the leadership team already reads. A scheduled task, not a person remembering.
Slack
Google Sheetscreator-weekly-reports
Creator success · every weekMacroActive's customers are creators with their own apps. The weekly report for each one used to be a person's Friday. Now it is a skill, and the person reviews it.
Slack
Google DocsThe rest cover the working week: the daily calendar briefing for the leadership team, the product request scan, the Expensify receipt run, visual QA on every page change, and a skill for building the next skill. Each one reads from the systems it needs, and nothing else.
SlackJira
DudaGoogle MeetGoogle CalendarGoogle DocsGoogle SheetsMetaExpensifyGitHubSupabaseSix departments, one library
Each department got the skills for its own jobs and the connections those jobs need, and nothing else. Product reaches Jira; creator success does not.
| Product and engineering | Slack to Jira, the product request scan, build-a-team-skill | SlackJiraGitHub |
| Creator success | Creator weekly reports, app page copy, the growth blueprint | SlackGoogle Docs |
| Marketing and web | Page visual QA, Duda widgets, sales page copy | DudaMeta |
| Finance and operations | The revenue snapshot, the Expensify receipt run | Expensify Google Sheets |
| Sales | The voice SDR, calling every lead and booking the call | MetaGoogle Meet |
| Leadership | The daily calendar briefing, the usage report | Google Calendar Slack |
The governance side
Deploying AI to forty people in three countries is an AI governance job as much as a build. Three things make it one you can point at. Every session is attributed to a named person, so the weekly report says who is driving the tools and on what, not just how much was spent. Every skill states which systems it may reach, and anything that leaves the building, a ticket, a page, a message, waits for a person. And the usage data sits in MacroActive's own database under a signed agreement that says what we may see and what we may publish.
The loop earned its keep in week three. A team member pulled a public third-party skill pack into the organisation and proposed it to everyone. The register caught it the same day: unreviewed skills, from a stranger's repository, at a company with a data agreement. The decision went to the founder with the evidence, rather than becoming the default by nobody noticing.
What MacroActive owns
A repository in their own GitHub organisation. In it: the skills library, the collector, and the dashboard in their brand. Beside it, the telemetry database they hold the key to. Versioned like software, owned by the company. Models will change and staff will move on. The library keeps compounding, and a competitor can buy the same Claude seats but not the architecture built around how this team works.
The result
An international team of forty on one Claude library, installed in a day on infrastructure the company owns, with every session attributed and every skill's reach on the register. A voice SDR booking three to five sales calls a day without a person touching it, and tens of thousands a year back from subscriptions rebuilt in house.
The custom development underneath
Some workflows need more than a skill file, so alongside the library we did the custom development necessary to enable them. The largest is a custom voice AI sales SDR. It calls every lead from MacroActive's Meta campaigns within about five minutes, while they are still warm, qualifies them, and books the sales call straight into Google Meet. It averages three to five booked calls a day on autopilot, for a fraction of the cost of a salesperson, and it was trained on the same call data that cut new-rep ramp from about three months to about one.
The second is quieter and pays every month. A growth company accumulates software subscriptions that each do one narrow thing. Where a skill or a small in-house app does the job as well, we rebuild it and the subscription goes. Across the estate that is saving the team tens of thousands of dollars a year, and every one of those rebuilds is now something MacroActive owns.
Could this be your team?
If you are paying for Claude your team barely uses, or you would deploy it tomorrow if you knew where the data went, this is the problem we solve. We start with a free AI Opportunity Audit: enter your company, connect the tools you already pay for, and see your three biggest opportunities mapped before you spend a dollar.
Frequently asked questions
Where does our usage data live?
In MacroActive's own database, in their region, under a key they issued and can revoke. The founder asked that question before agreeing to anything, and being able to answer it with their own Supabase project is what closed it. A signed data agreement sets out what we can see and what we can publish.
What does governance actually mean here?
Three things you can point at. Every session is attributed to a named person. Every skill states which systems it may reach, and anything that leaves the building waits for a person. And when a team member pulled a public third-party skill pack into the organisation, the register caught it the same day and the decision went to the founder rather than becoming the default.
Does this work for a team spread across countries?
Yes. The plugin reaches every seat on the organisation regardless of where the person sits, and the telemetry reports on one clock. MacroActive's team spans New Zealand, the UK and Latin America and runs on one library.
What is the voice SDR?
A custom voice agent that calls every lead from MacroActive's Meta campaigns within about five minutes, while they are still warm, and books a qualified sales call into Google Meet. It averages three to five booked calls a day on autopilot.
What do you mean by rebuilding subscriptions in house?
A growth company accumulates software subscriptions that each do one narrow thing. Where a skill or a small in-house app does the job as well, we rebuild it and the subscription goes. Across the estate that is saving MacroActive tens of thousands of dollars a year.
Who owns the skills and the wiki?
MacroActive. They live in a repository in their own GitHub organisation, alongside their branded dashboard, versioned like software. If we walked away tomorrow all of it would stay exactly where it is.
More case studies
Nature Baby: 7 Departments Migrated Into Claude
Seven departments, from finance to the shop floor, running their day-to-day through Claude: 50+ skills deployed, 12 data connections, and a weekly eval that ships the fixes.
The Ecommerce Accelerator: 40+ Hours a Week Saved
We trained a five-person agency to roughly double its output, then built one shared platform its members access through their own Claude or ChatGPT.
AI-Powered HR & Health and Safety Training Platform
An AI training platform that generates assessments, scores candidates, and automates compliance workflows.
Caregroup: Sovereign Claude Across 15 Healthcare Practices
Every inference pinned to Australia on AWS, Claude from reception to the board, a GUI automation layer built from clinician recordings, and a voice receptionist for 180,000 patients.