Tuesday 19 May 2026

Shadow AI is running your agency: Cat McGinn

As AI reshapes the way agencies work, Advertising Council Australia has partnered with Time Under Tension to launch AI Academy, an eight-week course designed to help agencies build practical AI skills. We caught up with course facilitator Cat McGinn ahead of the first intake on 16 June.


As creator of humAIn, the industry’s specialist conference on AI in media and marketing, what are you seeing that has convinced you agencies need to build their AI skills?

It wasn’t what they were doing with AI; it was the gap between what they thought they were doing and what was actually happening. 

Most agencies I speak to have shadow AI running through them: individuals experimenting with tools in isolation, invisible to leadership, ungoverned and uncoordinated, with no shared approach, no agreed standards, and no real understanding of where AI should and shouldn’t sit in the workflow. 

Someone in strategy using a tool for research, a producer using something else to cut turnaround times, a junior creative using something without considering brand safety to get unstuck – none of it connected, none of it intentional. 

That’s not capability, it’s improvisation, and improvisation at scale creates real risk: inconsistent output quality, over-reliance on tools people don’t fully understand, and a creeping homogeneity in the work. 

It also means they’re missing the category-redefining creative possibilities that happen when a whole team has levelled up together. The agencies that pull ahead won’t be the ones with the most enthusiastic early adopters; they’ll be the ones that have made deliberate decisions about how AI fits into their work, and have built the skills to back them up.

 

When agencies attend HumAIn, what are they coming for? 

humAIn is built for collaboration, because no one – not the AI research labs, not the most sophisticated marketing teams, not the most progressive agencies – has all the answers about AI. 

The technology is moving too fast, and the implications are too broad. So, humAIn is designed to be a space where the creative industries can come together to share what’s working, showcase boundary-pushing work, talk tools and tech, but also interrogate the bigger conversations about impact – cultural, organisational, ethical, environmental – that don’t get enough airtime elsewhere.

The conference has always attracted the people who are paying attention – not the hype-chasers, but the ones who understand something structural is shifting and want to understand it at a deeper level than the usual conference circuit offers.  

 

Why AI Academy?

It’s designed to help agencies get practical, stay practical, and give people time to actually build something over eight weeks of live sessions. 

I’m ex-agency, and spent years as an industry commentator, which means I understand the process, the silos, and the gaps in ways a purely technical trainer often doesn’t: how a brief actually works, how client service teams think, where the disconnect between strategy and creative emerges, and which parts of the workflow are genuinely painful, versus which ones just feel that way.

That context shapes everything about how the course is designed, moving from foundations like prompting and research skills, through to building custom AI assistants for recurring tasks, and ultimately to a real adoption roadmap that participants can take back to their specific teams and implement, not a generic framework they’ll file and forget. 

Simon Hillier (Co-Facilitator) and I are both practitioners, and the course is built for the reality of agency life.

 

Claude, ChatGPT, Gemini and more — which tools are your go-to, and why?

  • Claude is my primary thinking partner for anything requiring sustained reasoning
  • Perplexity is my go-to for research because it cites its sources, which matters when you need to verify rather than just generate
  • ChatGPT (on the paid tier) remains the most versatile day-to-day tool
  • Gemini earns its place inside the Google ecosystem
  • NotebookLM has been a game-changer for the book I’m currently writing, working across large volumes of source material and data in a way no other tool handles as well, and 
  • I’m actively experimenting with Runway for video and Lovable for building lightweight digital products, both of which are moving faster than many people realise.

One thing I do that most people don’t have time for is run constant comparisons across the major LLMs as their capabilities shift; part of what I’ll bring to AI Academy is that I’ve done that work, so participants don’t have to.

But the tool is always secondary to the approach: if you can’t clearly articulate the task, set appropriate parameters, and critically evaluate what comes back, you’ll get mediocre output regardless of platform.

 

How do agency leaders encourage AI adoption without losing the critical thinking and creativity that set great agencies apart?

The risk isn’t that AI makes agencies less creative – it’s more subtle than that, and more insidious.

Researchers have identified a process called “cognitive surrender”: the tendency to accept AI output without using your own judgment. When that becomes habitual across a whole team, you don’t just lose quality control, you start to lose the habit of critical thinking itself. 

The leaders getting this right are being intentional about where AI is and isn’t appropriate, investing in the skills to evaluate output critically rather than just generate it, and actively protecting the conditions in which original thinking happens. 

My view is that AI should absorb the repeatable, time-consuming, and cognitively expensive but low-value tasks, so human thinking can flourish where it actually matters. It can also offer us ways of bringing ideas to life that might never have been possible before – because of cost or technical limitations, which is so exciting. 

That’s not a threat to creativity; it’s explicitly making space for it. But it requires active, deliberate decisions and systems thinking, rather than letting adoption happen by a process of slow drift and hoping for the best.

***

Ready to move beyond experimentation? AI Academy starts on 16 June for ACA members. Secure your place here.