Validation Day, 16 October 2026
On GreenSpark Validation Day we spoke with mentors and industry representatives. Below is what we heard, grouped by who said it. Names are omitted unless the person has agreed to be named.
A Horizon Power representative
- Planned outages need long approval and notice (up to two weeks). If a major weather event such as a cyclone is coming, planned outages are cancelled and teams move to emergency response. They rely on the Bureau of Meteorology.
- Earlier weather warnings would support proactive customer outreach, for example reminding prepay customers to top up their credit before a heatwave, because prepay power disconnects when credit runs out.
- Disconnecting a Life Support customer is treated as a very serious breach. Life Support customers must acknowledge a planned outage notice before it goes ahead (message, then call, then in-person visit); otherwise the outage is cancelled or delayed.
- For a real product, data security and compliance matter most. A proof of concept on synthetic data, with clearly stated assumptions, is the right approach.
- Find the problem first: talk to people who know it and to the managers who would decide to deploy a tool.
A Rennie Advisory mentor
- Vulnerability means being at risk of harm, not just being unable to pay. Harm can be made worse by extreme heat, family and domestic violence, housing instability, job loss or changes in family circumstances.
- Victoria's Essential Services Commission is a good model for defining and acting on vulnerability.
- A tool like this would most likely need to sit inside the utility's own systems.
- Of the three sub-problems, Detect would most interest a company.
- The value for a utility is evidence that it is proactively identifying risk (compliance), plus lower costs over time. Financial data would likely be limited to bill payment history.
Other GreenSpark mentors
- Combining data is not the final product; the intervention is. A human must make the judgement, for example when contact is unsafe because of family violence.
- Look for patterns and indicators rather than labelling a customer as vulnerable.
- Compare with how banks assess loan risk, but used to help people rather than to refuse them.
What we are doing about it
- Focusing on 3.1 Detect and 3.2 Explain.
- Keeping a person in charge of every decision, with a separate pathway for safety concerns.
- Using synthetic data only, and stating our assumptions clearly.
- Prioritising conversations with the managers who own hardship and customer support.
A project developed through GreenTech Hub's GreenSpark innovation program, Perth, 15 to 30 October 2026.
Team Connect is an independent student and early-career team. Organisations named on these pages are referenced from public sources or our own conversations; naming them does not mean they endorse, partner with or have reviewed this project.
Last updated: 17 October 2026