Hey there,

Today we’re doing something a little different.

Last Thursday, Glitch hosted our inaugural Annual Review to give our community of LPs and Founders an update on our portfolio and the market. The event was for our limited partners. As such, much of the content was confidential and under the Chatham House Rules.

But we felt some of what was said on the panel was very useful, so we wanted to share as much as possible.

The panel was hosted by Glitch Partner and Firmable co-founder Leigh Jasper and featured portfolio founders Herk Kailis (Cadmus), Adam Laitt (Traild), and Sam Senior (Monarch).

Leigh, Herk, Adam, and Sam are all scaling companies globally, targeting different customers, from high-velocity SMEs to large, slow-moving enterprises.

The panel focused on AI in Australia vs. the US, AI adoption vs. the real world, and how each company is thinking about distribution and defensibility.

Sam (Monarch): The US is roughly 24 months ahead of Australia in adopting AI. Australia has real anticipation about what's possible, but it's "tightly coupled with fear and governance" that will kill innovation quickly. The US had already been through this phase two years ago and moved to "screw it, what can we start [automating] today?"

Herk (Cadmus): You have to be AI-focused even if your customers aren't. Universities "may be the last people to adopt AI in the world after the government." The job for startups is still the same. To find the enterprise-level pain, the budget line (AI transformation, integrity at scale, student pass rates) and the person who signs, and ensure your product helps those different stakeholders in a coherent way.

Adam (TRAILD): America isn't San Francisco. Most of the US market is more like a regional motel overlooking a Walmart, not Waymo. While in Australia we feel that the US is at the cutting edge of AI, most of America is still very early on the adoption curve.

Adam: COVID changed distribution more than AI has. Pre-COVID, TRAILD's customers wanted an in-person demo at their office. COVID forced those sales calls onto video, back to back. The AI revolution hasn't had a similar quantifiable effect on distribution for TRAILD.

Adam: The productivity story isn't only about engineering teams. Within TRAILD, the internal legal team has scaled its output dramatically without adding headcount by using AI.

Herk: Technology adoption hasn't changed. Cadmus serves universities where we’ll still see a thousand academics and a thousand ways of working, so it's the classic adoption curve: innovators, early adopters, then the majority.

Sam: Change management is a major moat. Even Anthropic (via Boris Cherny, creator of Claude Code) says enterprises are begging to be told how to start using AI. Anthropic is responding by deploying more engineers to the field. Bringing people, processes and organisations along is still the job to be done for startups.

Herk: Cadmus went from point solution to owning the whole platform, which makes point-solution competitors easy to displace and the product indispensable as the data compounds.

Sam: Security posture is now part of our infrastructure work for AI companies. Monarch is doing everything like SOC 2, pen testing, spending far more time on terms of service, and running models inside customer clouds via AWS Bedrock or Azure Foundry (often a version behind the latest release). That being said, though, SF is very "risk-on,” with nobody slowing down unless it's customer-facing work and they ask for it.

Adam: The mid-market now cares about cybersecurity. Once, around 20% of deals had a cyber review in the sales process. Now, nearly every material deal has data security, cyber and AI governance reviews, which is something we’ve had to adapt to.

Sam: Cost-out thinking is short-termism, but it’s rife in Australia. Australian enterprises are targeting 20–40% headcount efficiency through the use of AI. The US had that conversation a year ago and moved on to the "jobs shift" and to looking at AI as a value-creation tool.

Herk: Talent is the constraint, not cost. Cadmus is desperately looking for quality AI-native talent, which is scarce in Australia. As a business, Cadmus shifted away from hiring hungry junior people to hiring AI-native senior people with "20 times the capacity” for work and context compared to a new junior.

Herk: We aim to be efficient with software costs but over-invest in customer-facing humans. As contract values grow, you want more people servicing revenue. Over-service to get to scale fast, then let the servicing load come off the back end.

— Will Richards, Head of Platform at Glitch Capital