Anthropic released a free, self-paced course called AI Fluency, aimed at people who use AI tools daily but have never had formal training on how to get good outputs from them. For a Sydney or Melbourne small business owner who's been prompting Claude or ChatGPT by trial and error, it's a genuinely useful starting point. It's also, on its own, not enough to change how your business actually runs.
This kind of free training is showing up more often as AI tools move from novelty to daily use. The logic is sound: a person who understands how to work with a model well gets far more out of it than someone repeating the same three prompts. The gap most owners hit isn't in the course content, it's in the jump from 'our people are better at prompting' to 'our business runs a step faster because of it'.
What the course actually covers
The course is built around the idea that working with AI is a skill, not a one-off trick. Rather than teaching prompt templates to memorise, it walks through the underlying judgement calls: when to trust a model's output, when to check it, and how to structure a task so the AI has enough context to do it well. That framing matters, because most of the AI mistakes we see in client businesses aren't about picking the wrong tool. They're about handing over a task with too little context and then being surprised by a generic answer.
Broadly, the free material covers:
How to frame a request so the AI understands the goal, the audience, and the constraints, not just the topic
When to delegate a task outright versus keeping a human in the loop for review
How to spot confident-sounding but wrong output, and build a habit of verification
How to iterate on a response instead of accepting the first draft
Where AI tools are a poor fit for the task at hand
None of this is specific to one industry or one piece of software. It's aimed at building judgement that transfers across whatever tool a person ends up using day to day, which is exactly why it's a reasonable starting point rather than a finishing line.
Where the free course stops and implementation starts
This is the part that matters for a business owner rather than an individual employee. A course builds a person's judgement. It doesn't build your business's workflow. Those are different problems, and conflating them is the most common reason we see AI adoption stall after an initial burst of enthusiasm.
Take a bookkeeping-adjacent example. A course teaches an office manager how to prompt well when asking Claude to draft an overdue invoice reminder. It does not connect Claude to your accounting data, decide which of your 40 overdue accounts actually need a firm tone versus a gentle nudge, or make sure the drafts land in a queue someone reviews before sending. We recently scoped exactly this kind of setup for a Sydney services business with roughly $45,000 a month in receivables sitting in aged debt. The gap wasn't AI literacy — their team had done the reading. The gap was the plumbing between the AI and the actual data.
A few places we consistently see this gap show up:
Understanding a good prompt, but having no connected system for the AI to pull real data from
Knowing how to review AI output, but no defined checkpoint in the workflow where that review actually happens
Individual staff getting fast at using AI personally, with no shared process across the team
No decision about which tasks are safe to fully delegate versus which always need a human sign-off
None of this is a criticism of the course. It's simply outside its scope. A general AI literacy course is written for millions of people across every industry and country; it can't tell a Brisbane retailer how their point-of-sale data should feed into a Claude-drafted stock reorder, because it has no idea that system exists.
Turning course knowledge into a working setup
If your team has already worked through Anthropic's free material, or something similar, the next step isn't more training. It's picking two or three real, recurring tasks and building them into an actual workflow with Claude sitting in the middle — one that pulls from your real tools, has a human checkpoint where it matters, and runs the same way every time rather than depending on whoever remembers the right prompt that week.
For an Australian small business, that also means building in the context a generic course can't cover: privacy obligations under the Privacy Act when customer data flows through an AI tool, what your industry regulator expects if decisions get automated, and which tasks genuinely need a person to sign off rather than a model. None of that is covered in a general-purpose course, because it can't be — it depends on your business.
A reasonable order of operations looks like this: finish the free course material with whoever's keenest on the team, pick one workflow that currently eats real hours each week, map exactly where the data lives and who needs to review the output, then build that one thing properly before moving to the next. Three well-built workflows will do more for a 20-person business than a team that's individually fluent but working from a blank page every morning.
If your team has done the AI Fluency course and you're wondering what a proper setup looks like on top of it, that's a conversation worth having. Book a brainstorm and we'll map out where Claude can actually plug into how your business runs.
