The commerce infrastructure cycle that Banksia set out to back in 2021 is still in the middle of its deployment phase — which is a different statement from saying it's mature or saturated. The AI-native stacks for personalisation, dynamic pricing, delivery orchestration, and B2B marketplace automation are being built and deployed, and the companies doing that work are now mostly past the "will this work in production?" question into the "how do we scale this to more customers faster?" phase. That's a meaningful milestone. It means the category has cleared the first filter. The second filter — does this build a durable, defensible business, and on what timeline? — is the question the next two to three years will answer for the current generation of companies. But alongside the maturing mid-cycle, a new cycle is beginning to be visible, and it's the one we're spending the most time on in 2025 and into 2026.
The agentic commerce thesis starts from a specific observation: most of the AI capability deployed in commerce today is assisting human decisions — a pricing recommendation that a merchandiser approves, a personalisation ranking that surfaces a product a human buyer then selects, a delivery route suggestion that a logistics planner confirms. The next generation of systems replaces the approval step. A B2B procurement agent that monitors inventory levels, compares supplier pricing, evaluates delivery lead times against production schedules, and places orders autonomously within a defined parameter set is not a recommendation engine — it's an autonomous operator. The infrastructure requirements for that system are different from the infrastructure requirements for a recommendation layer: you need verified supplier catalogues that are machine-readable, payment authorisation flows that support programmatic execution, compliance checks that run at the API layer rather than in a human review queue, and audit trails that satisfy the financial controls requirements of the buying organisation. Most of that infrastructure doesn't exist in a clean, composable form yet. Building it is the opportunity.
The ambient purchasing dimension is a related thread that's playing out in consumer commerce. Smart home devices, wearables, and subscription management systems are increasingly managing low-consideration repeat purchases on behalf of consumers — detergent, pet food, personal care items — using consumption signals and purchase history rather than explicit purchase intent. The consumer experience isn't "search for a product and check out" — it's "this arrived at the right time." The commerce infrastructure that enables ambient purchasing has to handle demand signal ingestion from non-traditional sources, inventory and fulfilment orchestration tuned for replenishment rather than discovery-driven purchases, and loyalty and subscription economics that look different from one-off transaction models. Several of the most interesting companies we're seeing in early conversations are working in this space from different angles.
The convergence of AI recommendation with payment rails is the third thread, and it's the one with the longest time horizon to widespread deployment but potentially the most structural implications. The current commerce model requires a user to discover a product, evaluate it, decide to purchase, and execute the transaction as four distinct cognitive steps. Systems where the recommendation layer is deeply integrated with the payment authorisation layer — where a trusted recommendation from a context-aware system can be converted to a purchase with minimal friction — collapse some of those steps in ways that fundamentally change the conversion model for commerce. We're not describing impulse purchasing dressed up in AI language. We're describing a class of transaction where the trust relationship between the consumer and the recommendation system is strong enough, and the payment infrastructure is streamlined enough, that the act of discovery and the act of purchasing become very close together in time.
APAC's position in the next commerce cycle is, in our view, stronger than its position in the current one. The current cycle was about building the infrastructure layer to support commerce at scale in the region — payments, logistics, data platforms, marketplace tooling. APAC was behind the US and Europe at the start of that cycle and has been closing the gap. The next cycle — agentic commerce, ambient purchasing, AI-native discovery — starts from a more level baseline, and in several dimensions APAC has structural advantages. The density of mobile-first consumer behaviour, the existing depth of B2B digital commerce adoption in manufacturing supply chains across Northeast Asia, and the regulatory experimentation in markets like Singapore that is allowing new financial services infrastructure to be built faster than in legacy-regulated markets all point toward APAC being a generator of the next cycle's leading companies, not just an adopter of what's built elsewhere. That's the thesis we built Banksia around, and it's getting more right with every passing year.