-1.png)
What happens when AI can reason through a problem, but doesn't have access to the information it needs to solve it?
For Joel Chan, that question came from using AI in his own work.
As Operations Manager at Western Australian lithium-ion battery manufacturer RENOZ Energy, Joel spends his days working through the practical side of battery projects. That means sizing systems, modelling network tariffs, checking grid constraints and working out whether a project is actually feasible.
AI quickly became useful for some of that work, but Joel started running into a problem.
The AI could also be confidently wrong.
It might invent an inverter specification, misunderstand a system requirement or pull battery pricing that was already 18 months out of date. For something as practical as engineering and commercial feasibility, a convincing answer isn't much use if the information behind it isn't right.
“The models are great at drafting and reasoning, but they kept hitting a hard wall.”
The more Joel worked with AI, the clearer that wall became. The problem wasn't necessarily that the models couldn't think through a problem. They simply didn't have access to the live, verified information needed to make a good decision.
And that information often lives somewhere outside the AI. That’s where aecon started.
From answering questions to doing the work
Imagine an AI agent working on a real project. It might need a current electricity market price or specialised grid data before it can finish the task. A person can find that information and pay for it.
For software, it's not quite so simple.
Today, an engineer will generally have to find the service, create an account, add payment details and connect an API key before an AI agent can use it. Joel started thinking about what it would take to make that transaction happen automatically.
That's the idea behind aecon (Agentic Economy).
aecon is building infrastructure that allows agents to find external services, see what they cost and pay for what they need.
There are still rules around the transaction. A business can set spending limits and decide what an agent is allowed to do. The provider gets paid, while the business receives a normal Australian tax invoice with an ABN and GST.
For Joel the goal is fairly simple: give AI agents a safe way to get the information and services they need to actually finish a job.
That's where the idea gets much bigger than a payment system. If AI is going to move from answering questions to doing work, it needs a way to interact with the world outside its own software.
Building something that can actually be trusted
Joel joined the AI Founder Sprint with a working technical model of aecon, but there were still plenty of questions around how the idea would work in the real world.
The Sprint gave him a chance to step away from the technical mechanics and pressure-test some of those questions with other founders, mentors and people working across different parts of the WA ecosystem.
“Western Australia is practical. People here use AI heavily for compliance, legal drafting, and mining reports. We wanted to test our assumptions against those exact workflows instead of guessing from behind a desk.”
Sessions with mentors like Aidan Morgan, Chief Engineer at Bankwest, also pushed Joel to think beyond the technical side of aecon and focus on what businesses would actually need to make this work.
How much should an agent be allowed to spend? Who is responsible when it makes a transaction? What happens when something goes wrong?
For aecon, those questions are just as important as the technology itself.
The bigger shift
One thing Joel noticed during the Sprint was how often the other founders ran into a similar problem, even though they were building very different things.
AI could increasingly plan what needed to happen, but getting it to interact with everything outside the system was the tricky part.
“The software could plan the work, but it could not reach outside its own code to finish the transaction.”
Seeing that same problem across different startups reinforced Joel’s view that the infrastructure around AI needs to catch up with what people are already asking these systems to do.
The idea of an AI agent isn't particularly useful if it can work out exactly what needs to happen, but then has to stop and wait for a person to step in and do it.
That's the gap aecon is trying to close.
What's next for aecon?
Joel is now preparing to launch the initial alpha of the Agentic Economy platform and start testing it in real-world environments.
The focus is on seeing how businesses actually use it, where the friction appears and what they need before they're comfortable allowing AI to make transactions on their behalf.
The shift Joel is interested in isn't really about AI becoming better at answering questions. It's about moving from AI that can reason through a task to AI that can actually complete it.
“Software is starting to buy services for itself. Our job is to make that completely normal, safe, and boring for Australian users and businesses.”
For aecon, that means building the infrastructure to make AI transactions feel like a normal part of doing business. To learn more about aecon you can visit their website here or connect with Joel via LinkedIn here.
See Joel pitch aecon at our upcoming AI Fellowship and Founder Sprint Demo Night on September 24th. Register here!
The AI Founder Sprint is a Spacecubed program delivered in partnership with Superteam Australia and Solana.