Joe Stevens.
Joe Stevens, Group Digital Director, Match & Wood
The problem with starting
There's a lot of noise around AI at the moment. And no, this isn't an application for the role of Captain Obvious. It's an acknowledgement that, with job displacement headlines coming out of places like Amazon, Atlassian and WiseTech, sci-fi associations, and the relentless speed of change and sheer volume of tools out there, the intimidation factor can be real when starting your AI journey.
This is, perhaps, why in February of this year only 14% of Australians said they use AI daily, and 36% say they never use it at all.
Add to this, coworkers showing off polished time-saving workflows, YouTube videos of business owners automating entire functions, and daily news articles about ever-expanding capabilities, and it can feel like that first step keeps getting higher.
The barrier to entry is lower than almost any other technology introduced in the last 20 years. You don't need tech skills or fully automated workflows. You just need to start. And to do that, all you need is to be a good communicator.
You'll be surprised at how quickly the ball gets rolling.
Using AI & I
AI may feel like magic, but it isn't. It's not omniscient either. Despite its sophistication, it relies on YOU for context and YOU for expertise. It amplifies and supplements your thinking, and I don't see that changing, even as the tools become more powerful.
My advice: pick one tool (don't overthink it, Copilot is the most integrated across most organisations) and just open it. Look through your workflow and find something you need help with. I like to start with an unenjoyable task, any time saved there is time I'm happier at work.
Start by breaking the task into subtasks and, for each one, identify where it falls in the "AI & I" framework:
AI = AI can do this
AI & I = I need to work with AI to do this
I = only I can do this
Then work with AI on the "AI" and "AI & I" subtasks. Prompt it the way you'd communicate with a colleague. Be clear, concise, and give the right context. This lets you chip away at small wins rather than trying to automate an entire day's work upfront (there's plenty of time for that).
For example, take a brainstorming session for a new client brief:
- Digesting the brief (I - though this could become AI&I with the right privacy and security protocols in place)
- Picking a suitable channel based on client context and business goals (AI&I)
- Pulling case studies or industry examples (AI)
- Strategic understanding (I)
- Refining response structure and writing the proposal (AI&I)
- Presenting to the client (I)
Just like that, you have a set of starting subtasks mapped to your capabilities.
From there, craft your prompts using the C.R.E.A.T.E. framework:
Context (what role should the model adopt?)
Result (what output do you want?)
Explanation (what information does it need?)
Audience (who is this for?)
Tone (how should it sound?)
Exceptions (what should it avoid?).
Add tasks, refine them, rinse and repeat. The progress is in the experimentation, and those who win won't be the most technical, but the most curious.
Agents
While there's plenty of debate around agents, APIs and CLIs, you don't need to get caught up in it. Start by logging what tasks you're using Copilot for. After a couple of weeks, you'll likely spot patterns, recurring tasks where you're entering the same context, the same explanations, the same requests.
That's your cue to build an agent.
As I mentioned earlier, AI relies on your context. Creating an agent gives you the opportunity to expand on that context and build in relevant information that gets used automatically throughout your workflow. It increases speed, improves quality, and reduces the chance of hallucinations.
In the brainstorming example above, you could build a 'brainstorming agent' that already knows your preferred structure, your go-to sources, or even just the channel outputs you typically work toward.
And just like that, you're building your own agents, and with it, embedding AI into your daily work.
Beyond
Then the fun starts: vibecoding, building internal tools, lightweight low-code app creation, synthesising data at speed, synthetic research, sense-testing strategies.
But don't worry about those yet.
Go. Start. Experiment. Take the first step, because the only way to beat the intimidation is head on. You'll be amazed where it leads.