Steal This Move
Short answer: A rescue team tuned its drone’s AI to report possible people at low confidence because a pilot confirms each one. Teams screening contracts, expenses or customer messages can copy the setting: ask AI to flag broadly with a confidence level, assign one owner to check the flags, and tighten definitions rather than raising the bar.
How did the rescue team find the hikers?
Two hikers missed a pickup on the Dead Horse Gap track in Australia’s Snowy Mountains as darkness fell. Fire and Rescue NSW launched a drone from the search command post. Its AI scans the camera feed and estimates how likely each object is to be a person, and the team uses it to report possibilities, not certainties. Inspector Phillip Eberle described the AI saying an object “has a 35 per cent chance of being human.” At a 100 percent setting, he said, “it wouldn’t pick it up.”
The pilot checks each report. That is how this pair was found: the AI raised a possibility, the thermal camera confirmed it, and the drone’s spotlight guided ground crews in.
Why does a low bar work here?
The setting only makes sense because of what sits behind it. A cautious threshold would keep the pilot’s workload low and miss the people the search exists to find. A generous threshold gives the pilot more to check, and every check has an owner. The two decisions depend on each other.
Many organizations configure AI screening the other way. Contract review, expense checks and customer-message triage are often set to surface “only real problems,” because a long list feels like noise. The result is a short list that looks reassuring and hides the uncertain cases, which are precisely the ones that need a person.
What should a team copy?
Flag broadly, with a confidence level. Ask the tool to report anything that might be an issue and to say how sure it is, sorted from most to least likely.
Give the flags one owner. The same person or role checks them every time, so the review is a job, not a hope.
Tune definitions, not the bar. If the volume is unmanageable, sharpen what counts as an issue. Raising the threshold until the list goes quiet recreates the blind spot.
Habits like this hold when a team practices them together. Start with a guided session →
Keep reading: The reviewer folds when you press hardest. That is the moment you need it most. · Nearly every team uses it. Look at who marked the answers.
This is the thinking behind a Trusted AI Culture.
Sources: Region Media, "How AI saved two hikers lost in the unforgiving alpine wilderness," June 29, 2026. Fire and Rescue NSW, "Fire and Rescue NSW drone team locates missing hikers using AI technology," June 25, 2026.
Written with AI in the loop: my idea, AI drafted and sharpened, my judgment on the way out. Every word is mine to stand behind. — Darren