If prior cycles and investor loss aversion are anything to go by, we think the supply of available credit is about to contract. For example, investors who have lost money on Bathla are unlikely to feel like allocating again. Investors who have exposure to Bathla and don’t know if they have lost money yet but have had their funds gated are in a similar position. Those investors with no exposure to Bathla and in funds not gated will be reading the headlines and wondering if they should reconsider their allocations – and probably not to the upside. At the same time as these marginal providers of credit withdraw, the cost of debt is rising across the world.

It is one thing to stay quiet in a crowded cinema when you see smoke. It is another thing to assume it is a smoke machine at the disco. After our review of the sector in June (fishing with dynamite) we exited our only exposure to the sector – Centuria Capital (CNI) – in July. It was only a small position (0.5% of the portfolio). In discussion with industry participants, we were surprised by how sanguine many in the property and private credit industry were about the risks in general and Bathla in particular, but having seen this movie before we have come to expect a high degree of confidence from Australian property market operators. We decided our research pointed to fundamental problems no confidence could overcome and exited our position.

We do not think the risks end with Bathla – its size and structuring complexity made it a particularly vulnerable example, but the pressures aren’t confined to just one developer. We think there are likely more entities on the brink in the property development market; exposed to the same risks of rising costs, falling prices, and extension of cash realisations as sales velocity slows. If investors retrench from private credit funding of property developers, this will only exacerbate the problem.

Bathla is a problem the market was slow to price in. Corporate Travel Management (CTD) is a company everyone wanted to be a problem, but reality seems unwilling to comply. After many commentators wrote its obituary on the basis of scarce information and opinion rather than data, here they are in early September about to relist with real cash flow, moderate debt, and new contracts from customers who, a few months ago, were assumed to be hastening the end of CTD as we knew it (UK Ministry of Defence (MoD) £28m contract). CTD will apparently be the longest ASX suspension that neither delisted, went insolvent, or required an emergency capital raising to survive. Under the prior founder management and board, the company was un-investible for many. It will be interesting to see if this changes, or if the stigma is hard to shake.

Confirmation bias is a hell of a drug.

There are lots of ways to interpret reporting season, and this year the lens we choose to view it through is productivity. As we wrote (escape from albotraz) productivity growth is a major problem for the domestic economy, and it is fascinating how differently large and small cap firms approach solving it. For large caps, the actions of Qantas, Telstra and Coles are instructive. Their management teams appear to believe the letters AI stand for “Accenture Invoices” as their best efforts to harness this once in a generation technology amount to making thousands of domestic employees redundant and entering offshore outsourcing agreements with the consulting firm.

Our own experience with AI – in this case Artificial Intelligence – is you need to spend more time understanding the technology, the ecosystem, and having a clear plan of where, how, and over what time period you adapt which parts of your business (or investment process) to the technology.

Familiarity with the detail means as models or harnesses change, the business can understand what these changes mean to the plan. AI in practice looks very different to the breathless LinkedIn posts about agentic employees.

This approach is also what we are seeing in many small caps, whose management teams live in the business rather than in strategy sessions with consultants and whiteboards. Many companies are two or in some cases three years into implementing AI in their business. These are not proof of concepts, demos or meaningless adoption metrics (no one cares how many Microsoft copilot seats you have paid for). These are production changes to operating processes which are evidenced by a real change in productivity – specifically revenue growth at a greater rate than cost growth, including the cost of AI itself.

Below are a series of examples we saw in August from companies across the small cap market in many sectors, and how AI is helping them and their shareholder escape from Albotraz and take control of productivity growth inside their organisations.

Breville Group Limited (BRG.ASX)
6.3x Recipe Creation Velocity: Deployed its proprietary AI Recipe tool in production to generate 660 variant recipes in 67 hours (compared to ~425 hours for 120 Hero recipes under its pre-AI process). 5x Faster Prototyping: UX teams developed consumer-tested prototypes 5x faster, handing off to firmware teams 2–3 months earlier. Rollout Compression: Project Go launched across 40 countries in 8 months using AI versus a comparable pre-AI program taking 14 months to reach just 8 markets.

McMillan Shakespeare Ltd (MMS.ASX)
13% Reduction in Average Handling Time (AHT): Deployed AI and data capabilities alongside real-time agent monitoring across customer operations, driving a 13% reduction in call handling time via dynamic, real-time agent guidance. 18% Drop in After-Call Work: Automated administrative and post-inquiry logging, cutting after-call wrap-up work by 18% and accelerating total inquiry resolution times.

Superloop Limited (SLC.ASX)
~400,000 Calls Avoided: Live AI agents (Teddy & Mo) fully deflected and resolved ~400,000 incoming customer calls. >500,000 Automated Fault Fixes: AI diagnostic tools (Refreshify & Exray) processed >500k fault resolution events, the majority completed without human intervention. 75% Same-Day Activation: Operational AI (Processify) achieved same-day service activation for 75% of orders, with <10% requiring any manual touchpoints. Headcount Containment: Avoided volume-driven call centre labour additions, keeping employee cost growth modest while revenue grew 21.6%.

NIB Holdings Limited (NHF.ASX)
11.5% Higher Customer/FTE Ratio: AI automation and workflow simplification lifted the Customers-per-FTE operational ratio by 11.5% year-on-year. >26,000 Manual Hours Saved: Production AI frontline tools (nibGPT, 700+ users across ~340k queries) eliminated >26,000 hours of manual contact centre work.

Netwealth Group Limited (NWL.ASX)
50% Autonomous Query Resolution: The in-production AI assistant NOVA (Netwealth Online Virtual Assistant) now resolves approximately 50% of all external “how-to” queries submitted by financial advisers and their clients. Accelerated Feature Delivery: Deployed AI-assisted development and code generation across engineering squads to materially reduce cycle times and speed up the rollout of platform functionality.

Australian Clinical Labs Limited (ACL.ASX)
20% Efficiency & 30% Faster Turnaround: The production Lab of the Future AI/ML platform operates 24/7, processing +50% of routine laboratory volume with a 20% efficiency gain and 30% faster critical result delivery. OCR Automation: Live AI/OCR systems deployed in back-office billing and operations to drive cash savings into FY27.

Temple & Webster Group Ltd (TPW.ASX)
$3M Fixed Cost Reduction: Realized $3 million in fixed cost savings in 2H FY26 vs 1H FY26, driven primarily by live automation and AI insights agents. Operating Leverage: Fixed cost ratio declined from 10.6% to 10.1%, with fixed overheads growing only 5.5% against 11% revenue growth.

Megaport Limited (MP1.ASX)
2x to 4x Developer Output: Tooling engineering staff with generative AI systems drove a 2x to 4x year-on-year increase in output per contributor.

Smartgroup Corporation Ltd (SIQ.ASX)
+32.3% Lift in Operational Efficiency: AI-enabled process automation and customer interaction tools within the broader transformation program increased Customers per operational FTE from 1,389 (in 2023) to 1,837 today. Scalable Operating Leverage: Scaled digital omnichannel service delivery while holding headcount efficiency, contributing to 13% top-line revenue growth outpacing operating expense growth.

Electro Optic Systems Holdings Ltd (EOS.ASX)
Less than 10-Second Threat-to-Neutralisation Loop: The NiDAR AI platform automates the end-to-end loop from threat detection to kinetic solution in under 10 seconds. 1-to-Hundreds Operator Leverage: AI target-assignment algorithms enable a single human operator to track and engage hundreds of incoming targets, cutting operator training time to a few days.

SiteMinder Ltd (SDR.ASX)
Over 2x Code Deployment Throughput: Core engineering teams doubled their production code-change throughput. 10%+ Support Cost-per-Case Reduction: Delivered a >10% reduction in support unit costs alongside higher customer satisfaction. Gross Margin Expansion: Directly linked AI tooling efficiencies to a 62 bps expansion in adjusted subscription gross margin (to 87.0%).

AMP Limited (AMP.ASX)
Up to 2 Hours Saved Per Review: In-production North AI Workbench and AI FileNote cut administrative time by up to 2 hours per client review, directly unlocking adviser capacity.

Zip Co Limited (ZIP.ASX)
86% AI-Assisted Code Production: AI tools assist in 86% of code written in US teams and 57% in Australia, accelerating time-to-market. Core Process Automation: Live deployment of the ZedAI agent layer automated operational tasks across customer service, merchant onboarding, and collections.

Steadfast Group Limited (SDF.ASX)
5x to 10x Faster Analyst Insights: Deployed Cortex (conversational AI with a semantic layer) and two targeted AI agents for broker and insurer analysis, enabling analysts to extract data insights 5 to 10x faster. 40% Platform Performance Boost & Cost Reduction: Applied AI optimization across its underlying platform architecture, delivering a 40% performance improvement, lowering computing/processing power usage, and reducing ongoing platform operating costs.