Straight from the unicorn's mouth: Lessons on AI-first product leadership
from the 2025 Rampersand Product Counsel LIVE events.
We kicked off with the first-ever Product Counsel LIVE events last month, bringing together founders and heads of product from Australia’s most successful tech companies, including Canva, Culture Amp, Xero, and Seek, to discuss product leadership in the age of AI.
Across two packed rooms in Melbourne and Sydney, early-stage founders sat shoulder to shoulder with seasoned operators and product leaders. The conversations were sharp, honest, and generous, with Chatham House rules creating a safe space to share gritty realities rather than polished stories.
This is the purpose of Product Counsel: bringing world-class founders and operators together to help grow faster and better. What emerged was a rare peek behind the curtain into how Australia’s top product builders are navigating the AI wave, and what early-stage founders can put into practice today.
It’s impossible to capture the raw value of the information shared on the nights. Here are four clear lessons from our panellists and members that were really valuable:
1. Pressure to “Have an AI strategy”
A consistent tension emerged: the pressure to “have an AI strategy” is enormous, even when the product isn’t ready, or the customer doesn’t actually need it.
Boards want to see AI on the agenda, customers expect AI baked in, and investors increasingly treat an AI narrative as table stakes. Several founders described fundraising environments where simply having (or not having) an AI story dramatically shifted investor interest:
“It went from nobody caring to everybody caring almost overnight.”
This pressure pushed many teams to “shipping something safe fast,” leading to a wave of support bots and low-risk features released not because they created the most value, but because they were the least likely to break and the easiest to sell internally.
But the product leaders with the most scar tissue weren’t chasing ✨shiny✨ features. They were clear: AI should sit behind the customer problem, not in front of it. When anchored correctly, AI accelerates the product story; it doesn’t replace it.
“The fundraising market only seems to care about AI technology right now... you need a compelling AI story alongside your strategy.”
KEY TAKEAWAY - AI shouldn’t be the story; it should accelerate the story.
2. The new AI tempo: Speed, iteration & building for failure
AI has rewritten the speed of product development. Prototypes that once took weeks can now be done in minutes, and teams are learning faster than ever.
But this new tempo introduces fragility: models behave inconsistently, degrade without warning, and react unpredictably to messy real-world inputs. Shipping faster means encountering the unknown faster.
“I had something that would’ve taken 4–5 weeks… and I got it in 30 minutes.”
But speed comes with instability. One Product Counsel leader described models mixing units in financial tables, misinterpreting formatting, and even shifting behaviour from concise and direct to super chatty overnight, with no explanation:
“They started presenting thousands on one page and millions on another… the LLM went, ‘What do you want me to do?’”
One story described a high-performing algorithm failing when repurposed for a different context, a reminder that AI cannot be assumed to generalise:
“Data solves the problem it solves - not the problem you wish it solved.”
This new environment has reshaped quality practices.
But the “testing” that matters most is watching real user behaviour, because generative systems are inherently non-deterministic. Understanding how customers use the system has become the core of quality.
“If you’re writing generative AI code, it should be 20% code and 80% code that tests the code (evals).”
KEY TAKEAWAY - In a probabilistic world, real usage (not theoretical QA) defines quality.
The teams that win are those who ship small experiments quickly and invest deeply in evaluation and user behaviour.
3. New PM responsibilities = Team culture and structure shifts
AI has expanded the product role. PMs now navigate model selection, evaluation frameworks, cost implications, legal/privacy alignment, and the operational realities of model drift. But the shift is cultural as much as technical.
Product Counsel leaders highlighted how privacy, legal, and security now sit at the heart of the product development process from day one. PMs are coordinating decisions that span technical feasibility, ethical risk, regulatory interpretation, and cost.
PMs are now expected to understand evaluation frameworks, something few teams had even considered a year ago:
“Our PMs now need to know evals… nobody was thinking about that a year ago.”
Teams are levelling up together. Many companies are embedding structured AI development goals across engineering, improving velocity without sacrificing quality.
AI-generated early drafts are also pulling engineers into problem definition earlier, strengthening collaboration and engagement.
KEY TAKEAWAY - The PM role is more complex - and more essential - than ever
AI won’t replace product thinkers, but it will separate the strong from the weak. It amplifies real product craft and exposes gaps fast. The winners will be PMs who layer AI onto solid product fundamentals.
4. Differentiation and trust in an AI world: Moats beyond the model
As access to the same foundation models becomes ubiquitous, differentiation shifts to what only you can own.
AI products strengthen over time as users contribute more data - not just content, but preferences, corrections, annotations, and contextual signals. These behavioural loops become powerful moats.
Trust also emerged as a defining differentiator. Some enterprise customers demanded AI opt-out mechanisms, but barely used them once implemented; they just wanted transparency:
“Nobody actually wants to opt out… but they want to know exactly what you’re doing with their data.”
Distribution, credibility and trust matter too. One member captured it succinctly:
“Owning the data, the behaviour… these are the moats. A kid in a basement isn’t selling to a board.”
UX also plays a key role. A team attempted to merge search and generation into a single elegant interface, only to learn it confused users.
And on the strategic front, several founders emphasised that thinking “AI-first”, rather than adding AI features, can reveal entirely new product categories:
“How would I solve this in an AI-first world? You may build something fundamentally different.”
KEY TAKEAWAY - Moats no longer come from access to models.
They come from proprietary data, trust, workflow embedding, and differentiated user experience: The parts competitors can’t copy.
BONUS: AI makes human judgment more valuable
AI accelerates execution, but the human craft of storytelling, synthesis, and problem framing matters more than ever. Tools can produce content instantly, but they can’t create clarity, strategy, or resonance.
One Product Counsel leader warned against outsourcing creativity entirely to AI, referencing cognitive science:
“The part of your brain used for brainstorming is the one that decays first… don’t outsource it entirely.”
KEY TAKEAWAY - AI can generate words, but humans create meaning.
The Rampersand Product Counsel (RPC) is unlike any product community in Australia. It is a high-calibre network of founders, product leaders, and operators who’ve built and scaled product-led companies - many to unicorn status.
The mission of the Product Counsel is simple and bears repeating: To bring world-class founders and operators together to help grow faster and better.
If you’re building a product-led company, consider this your invitation to join the community and learn from the people shaping the frontier.
More Product Counsel events are on the way! Keep an eye out and subscribe to stay in the loop:
 |
|