3 Chiefs, a Founder, a Futurist and Me Walk Into a Bar

A bar being IMAA's IndiePendence Day panel in Sydney last week. The session was called "The AI Shift – What Indie Agencies Must Do Next." I was there as the tech expert slash agency owner, so my lens on all of this is shaped by spending the last two years inside Australian businesses as BRAIVE actually building automations and AI systems, not advising from a distance. Prior to that 15 years in media & marketing.

What follows is purely based on my notes and what I said. I don't want to misquote the other panelists - they were brilliant and I've put their details at the end of this article if you want to reach out to them directly.

The conversations on stage and in the room afterwards hit a nerve. I'd condensed my prep down to 8 trigger words that I may or may not have written on my hand. So here are those notes expanded and what came up, what I said, and what I'm already second-guessing.

‍

What's the biggest misconception leaders still have about AI?

The quiet hope that an off-the-shelf tool is going to meaningfully solve their specific problems. If they just find the right platform, subscribe to the right AI product, everything clicks into place.

It doesn't work like that. Off-the-shelf tools solve off-the-shelf problems. Everyone has access to the same ones whether that's language models, image generators, predictive analytics platforms or automation builders. They're commodities now. Using them isn't a competitive advantage any more than using Google Sheets is. The edge comes from wrapping AI around your hyper-specific business problems, not subscribing to the same platform as everyone else.

The other misconception I called out on stage is that AI is fast and free. In theory, yes, AI can create fast, cheap, great things for you. But "in theory" is doing a lot of heavy lifting in that sentence. You still have to invest upfront. Time to map the problem properly. Money to build the right solution. Energy to change how your team actually works. The cost isn't in the AI. It's in the setup, the integration, the change management. People skip that part and then wonder why their AI investment didn't land.

And here's the paradox nobody in our industry is being honest about: AI genuinely does create founders' time. It compresses work, automates the repetitive stuff, frees up hours in the week. But show me a founder who actually has the time to invest in the setup. The people who need AI the most are the ones least equipped to stop long enough to implement it properly. It's like telling someone drowning that there's a boat... they just need to swim to it first.

Why do you need to have a short memory right now?

A lot of decisions leaders made 6, 12, 18 months ago are just now coming to fruition. New platform rolled out. CRM migration completed. Workflow redesigned. Team restructured around it. And the uncomfortable question nobody wants to ask is: is the thing that just landed actually good, or is it already the new best worst way to do it based on what dropped last month?

That's the reality right now. You can make a perfectly sound decision in June, execute it well, perfect it by February, and by the time it's live the landscape has shifted enough that you're already behind. Not because the decision was wrong at the time, but because the shelf life of a "right" decision has collapsed.

This requires something that doesn't come naturally to most leaders: zero emotional attachment to your own work. You need a short memory. The process you built three months ago might already be the worst way to do it, and if you're defending it because of the time and money you sunk into building it, you're optimising for sunk cost, not outcome.

I'm not saying throw everything out every quarter. I'm saying you need to hold your workflows loosely enough that when something better becomes possible  (and it will, faster than you expect) you can actually move on it without a six-month grieving period for the old way.

What's happening faster than expected and what's taking longer?

⚡️ Faster: anything that involves pressing buttons in the right order. Manual execution is getting squeezed down to zero. Meta is moving toward goal-only campaign setups, Google is pushing keywordless search (and they always have been and agencies have always navigated value, but it's only headed one way). If your job is clicking things in the right sequence, that job has a shelf life measured in months, not years.

The gap between a good media buyer and a great one is about to get way wider because of this. I've got a lot more to say on what skills actually matter on the other side - I'll be writing about that separately. Short version: the middle of the stack is hollowing out and you need to pick a lane.

Also faster: the zero-click shift. North of 43% of Google searches with an AI Overview now end without the user ever clicking a link. That number is climbing. And it's not just a search problem (tbh I don't even think it's a problem) we need to start thinking about what happens when AI is the buyer, not the human. B2B and B2C campaigns are one thing, but what about B2A? A2A? Agent-to-agent marketing isn't science fiction. It's already being discussed seriously.

🦥 Slower: trust. There's a meaningful gap between what people are using AI for and what it's truly capable of. That's a mix of trust and education, and it's going to take a while to close. The content farms churning out AI garbage for AdSense revenue aren't helping. Every bad AI-generated article, ad or post makes the next legitimate one harder to trust.

Also slower: the collaboration between AI capability and life experience. A 19-year-old likely knows how to harness AI. But a 30-year veteran in our industry knows what to harness. That combination of domain expertise plus AI tools is where the unfair advantage sits. And it takes time to develop. For what it's worth, as someone who started in this industry at 16, I hate when people assign value of opinion or impact to age it's just an example of the dynamic, not a rule.

One thing to watch: middle managers getting in the way of progress. Not maliciously but they're often the ones most invested in the current process because they built it, they own it, and their role is defined by it. When someone comes along and can automates half of it in a day, that's not a celebration for the person whose job description just got thinner. Be aware of that. The biggest blocker to AI adoption in most organisations isn't the tech, it's the org chart.

‍

What's one decision leaders should make now?

My answer on stage was boring and operational. Sorry.

Audit and write down every workflow or process you have. Every one. Assign each a pain score and a rough estimate of how long it takes. That's your initial automation hitlist  and more often than not, the solutions will utilise AI to get it done.

You will not be able to create space for true agentic capability within your business until you automate everything that honestly, should be automated. Along the way you'll find that AI and Automation work really well together too.

Make a decision to focus on operational hygiene before the shiny stuff and commercialisation. I know that's not the sexy answer. But the businesses I see winning with AI aren't the ones chasing the latest tool launch. They're the ones who looked inward first, found the friction, and built around it.

Start with the friction, the tool reveals itself. And hint hint it's likely not a $29.95/month SaaS product. It's 16 hours of "coding" it and iterating it yourself and then $X per day in token and hosting costs.

The thing I'm already wrestling with

On the panel I said that vibe coding and hard coding are no longer dirty words. I meant it. The ability to build a quick, scrappy tool specific to your business even if nobody else ever uses it is genuinely powerful. It's lowered the barrier to solving problems that were previously too small or too niche for anyone to bother with. The compounded effect of solving enough of those? Wunderbar.

Before you roast me yes I do believe you should iterate and have your code reviewed by a human for any proper applications, especially customer facing though.

As a semi-counter to that though. People acting like having quick access to code was the only thing that had been holding them back from the next level of their business. As if the bottleneck was always technical, and now that it's gone, everything else just falls into place.

It doesn't. Here's what I think is actually happening: ideas being expensive to implement was acting as a quality filter. When building was hard, only the ideas worth the investment survived long enough to get built. Now that filter is gone. And we're finding out that a lot of ideas probably should have stayed on the whiteboard.

Being able to build something doesn't mean you know what to build, or whether it's worth building, or how to make it stick inside a team's workflow. The hard part was never the code. It was understanding the problem deeply enough to know what the right solution looks like (and cash... likely). That hasn't changed. If anything, the flood of people shipping half-baked tools is making it harder to distinguish signal from noise.

I don't have a clean answer for this one yet. I still believe vibe coding unlocks real value. I've seen it shave hours and hours off a teams menial task time each week. But I'm watching closely for where the line is between genuine problem-solving and just building stuff because you can.

What does competitive advantage actually mean now?

Every workflow you've perfected is potentially a product. That approval process you spent two years refining? That reporting cadence your clients love? Those aren't just internal efficiencies, they're assets. Agencies aren't just services businesses anymore. They're product incubators, whether they realise it or not.

Agencies should think less service, more product. "We compete on service though, ours is better than anyone else!" - sure buddy.

On the flip side, I think this industry needs to let go of the idea that headcount equals capability. The 10:1 agent-to-human ratio isn't theoretical, it's happening right now in companies that have leaned in. Revenue per employee shouldn't be a dirty word. It should be a metric every agency leader is watching closely.

And yeah, letting go. That's the thread running through all of this. Let go of the process you're emotionally attached to. Let go of the decision you made 12 months ago that felt right at the time. Let go of the idea that building your own tools is someone else's job. The leaders who'll thrive in this environment aren't the ones with the best strategy. They're the ones with the shortest memory and the least ego about what they built yesterday.

Where does this all start?

If there was one thing I kept coming back to across the whole panel, it was this: do not start with tools. It's also the #1 question i get asked daily - "what's your favourite AI tool" and my answer is quite simply "the one you build for yourself".

Spend more time mapping out your friction points and the tools reveal themselves even if those tools end up being scrappy vibe-coded apps specific to you and your business. Nobody else needs to use them. They just need to work.

Stop looking for one tool that does everything. Start with one problem that costs you everything.

The other panelists (go say hello):

Bel Harper — Chief Product & Marketing Officer, oOh!

Katie Rigg-Smith — Futurist & CEO, Say Robots.

Kathryn Illy — CMO.

Margie Reid — CEO, Thinkerbell.

Louise Cummins (moderator) — Co-Founder, Australian Centre for AI in Marketing.