If AI can do the work, who’s actually accountable for what it produces? That’s the question sitting under this final episode of Phuel’s AI miniseries, as Nathy Gaffney and innovation expert Jake James turn from experimentation towards responsibility.
Jake opens by drawing a direct line between AI and the earliest arguments about plagiarism: copying an essay off the internet and passing AI generated work off as entirely your own sit on the same ethical spectrum. His starting point for anyone using AI professionally is simple. Follow your organisation’s governance first, and if none exists, raise that gap with whoever’s accountable rather than assuming it’s fine to proceed. Beyond that, disclosure becomes the real test. If a client or colleague would feel differently about your work knowing AI was involved, that’s the signal you need to have the conversation.
The two push into whether regulation can keep pace with the technology itself. Jake references Gartner survey data showing a strong appetite in Australia for government to step in, but he’s sceptical that legislation alone solves the problem given how far technology development speed outruns policy. His view is that accountability sits with the individual user first: you cannot claim ignorance about not feeding confidential client data into a public AI tool. Legislation might eventually support organisations in making good decisions, but it will not replace the need for people to educate themselves.
Trust surfaces as the mechanism that determines whether any of this actually works in practice. Jake points to Amy Edmondson’s research from Harvard on psychological safety, arguing that teams only surface AI mistakes quickly if they feel safe admitting them in the first place. Leaders who create a climate of judgement around AI use risk teams hiding how the tool is actually being used, which is precisely when things go wrong. Nathy extends this with an image from field marking a football pitch with no lines drawn: without leaders clearly articulating what’s acceptable, people are left guessing where the boundaries sit.
The conversation closes by tying together the series as a whole: embracing discomfort, applying the ten hour rule popularised by Ethan Mollick, using the CAGE framework for prompting, and now layering ethics and disclosure on top. Jake’s closing point is refreshingly grounded. This isn’t really about rewriting common sense so much as building the habit of asking who might disagree with a decision, and who you could check it with, before publishing AI assisted work as your own.
This episode is worth your time if you’re a leader trying to set real boundaries around AI use in your team, not if you’re after another abstract debate about whether AI itself is good or bad.
