How upskilling a 5-person agency turned four-hour tasks into one

We did not replace anyone. We taught the team to drive specialised Claude and ChatGPT agents built around their exact jobs, and connected them through one operating layer.

The Ecommerce Accelerator (Magnitude Group) coaches ecommerce brands across Australia and New Zealand. Five sharp specialists: a media buyer, a creative, an account manager and an admin. On paper, a strong, lean team.

In practice, each of them was running their own workflow by hand. And while they had Claude and ChatGPT, they used them like a search box, a long way from the leverage the tools could actually give them.

Why was a sharp team still capped?

Because every specialist was a bottleneck on their own corner of the business. The creative built every carousel by hand. The account manager pulled P&L together manually. The media buyer ran campaigns the slow way. Smart people, repeating the same work every week.

The AI that should have unlocked them sat mostly unused, because nobody had set it up around how they actually work. Generic prompts, no real workflow leverage.

How did we get the team actually using AI?

Not with a course. We ran hands-on sessions, sitting with each person and having them walk through exactly what they do day to day. Then I went away and built an agent for that specific job, and got added into their Claude and ChatGPT organisations so the agents lived right where the team already works.

The result was AI shaped around their real workflows, not bolted on beside them. Each specialist got a tool that does their actual job, and the training to drive it.

What did each person get?

We built three specialised agents, one per core workflow:

Creative: a carousel content agent

A fine-tuned ChatGPT agent that produces the creative carousel content the team used to build by hand, in their style.

Account manager: a P&L agent

A fine-tuned Claude agent wired into Xero and Google Sheets, so it builds P&L views on demand instead of someone pulling the numbers together manually.

Media buyer: a campaign agent

A custom-coded media-buying agent built into a custom application on the Meta API, so campaign work that used to be slow and manual runs through a purpose-built tool.

What ties it all together?

All of it runs through one central application that holds each agent's instructions and connects the specialised agents together. Instead of a scatter of disconnected prompts living in different heads, the team has one operating layer they work from, so the agents stay consistent and easy to maintain as the business grows.

What was the return?

The work landed where it counts. Core tasks that used to take four hours now take about one, roughly 4x faster, and across the team that added up to 40+ hours a week returned to the work that actually grows the business. All delivered in six weeks.

Could this be your team?

If your team is sharp but capped, doing skilled work by hand while the AI you pay for sits mostly unused, this is exactly the kind of problem we solve. We start with a Discovery Week: we find your single biggest win, scope it, and ship a first working piece fast, so you feel the value before any bigger decision.

Frequently asked questions

More case studies