ChatGPT and Claude at Work: what you're not using yet |
Melbourne

Building context that changes the output

Documents, examples, prior work, connected sources. The single biggest lever on quality and the one most people skip. Includes turning existing reports and internal material into context you reuse.

Projects and persistent working environments

Instructions, files, knowledge and memory set up so the tool already knows the client, the team or the recurring task before you start typing.

Memory management

Three different things get called memory: what carries within a conversation, what a project holds, and what follows you between chats. How to review, correct and clear each, and how to move memory between platforms.

Skills: teaching the tool your method

Both platforms now call them Skills, and both do much the same job: capture a recurring task once — your method, your format, your review steps — and have it followed the same way every time. Built live, on a real task.

Agents: ChatGPT Work and Claude Cowork

What "agent" actually means once you strip out the noise. Both platforms will take a brief, work for an extended period and hand back a finished file rather than a chat reply.

From tender document to draft submission

A full worked example. Interrogating the tender documents, pulling compliance requirements into a structure, drawing on past submissions and capability material, drafting against each criterion and reviewing the result. Days of assembly work compressed into an afternoon.

Large spreadsheets and messy data

Big workbooks don't work if you upload them and hope. The preparation that makes the difference: getting data into a shape the model can reason over, deciding what to send and what to leave out, working across sheets, and structuring the analysis so the answers hold up.

Building tools without code

Calculators, selectors, decision aids and internal tools built in plain language.

Claude Design: a first look

Setting up a design system (in effect a brand kit or style guide) so output looks like your organisation rather than generic, and a look at what it can produce. Enough to know whether it's worth your time.

Desktop and browser

What the desktop apps do that the browser versions can't, including permanent access to your own files and the ability to operate your computer.

Model variant and thinking effort

Which model to use and how hard to make it think. Most people leave this on default, which means either wasting time on trivial work or under-powering the work that matters.

The final pass

Getting a draft is the easy part. The discipline of the last review before something leaves your desk: what to check, what to hand back to a stronger model, and what only you can judge.

Source quality

Knowing what's grounded in your documents, what came off the web and what the model filled in on its own.

Where your judgement still has to sit

The work that stays yours regardless of how good the tools get.

Prerequisites and access

No formal prerequisites. A paid ChatGPT or Claude licence is strongly recommended — either is fine, and most of what's covered assumes paid access. Without one you'll follow the day, but much of it won't be available to you afterwards. Note that some ChatGPT features, including Skills, are only available on business and enterprise plans.

Bringing a laptop or device is optional. This is a demonstration-led session, not a hands-on lab.

The program is updated regularly as the platforms change.

Pair this with a foundational course

This course can be taken as a standalone program, but it also works well alongside Foundations of AI at Work as a two-course sequence. Together, the two courses provide a practical progression from general AI understanding into more advanced professional use.

A 20% discount applies when booking two or more courses, or for two or more people. Many organisations treat the two courses as parts of a single learning program.