“People who treat responsibility, accountability, auditability and transparency as ongoing professional obligations rather than a compliance exercise to be completed and filed are the ones most likely to be considered trusted partners.”
100% this is a fantastic point. for me i would be so much more comfortable with this attitude from my partner.
Thank you! I find it’s easy to start feeling like verifying what AI says is a bureaucratic exercise. In my experience it comes from the fact that the average user doesn’t know that gen AI is probabilistic and very different from a google search - in that AI makes a claim through its response, whereas searches just say ‘here are a few links for you to figure things out’.
100%. and probability distributions work well in some places and not great in others. we are essentially forcing everything into this paradigm, then finding ad hoc fixes
This is a timely and necessary piece – not because “AI wrote it” is new, but because “you signed it” is the part people still treat as a formality. You frame the real shift correctly: authorship is turning into accountability, and responsibility doesn’t vanish just because production got easier.
I also appreciate how you avoid the lazy binary (“ban it” vs “embrace it”) and push the practical question: what does a signature actually certify in an AI-assisted workflow – intent, accuracy, originality, or just publication? Once AI becomes an invisible layer, the easiest failure mode is moral outsourcing: “the model did it,” therefore nobody did. That’s why governance has to evolve around process (review, traceability, disclosure norms), not tool policing.
The auditability section is where this lands hardest for me as an engineering manager.
The Microsoft research you cite describes exactly what I'm watching happen inside engineering teams right now. When junior engineers use AI to generate code they don't fully understand, the review step becomes ceremonial. A senior engineer reads it, it looks plausible, it passes. Nobody asked how it was produced or whether the person who submitted it could explain the reasoning.
The deeper problem is what happens over time. Engineers who never had to solve certain problems from scratch don't develop the intuition to catch errors in AI-generated solutions to those same problems. The audit trail exists the PR, the review, the approval but the human capacity to make that trail meaningful has quietly degraded.
Your point about documentation being the mechanism by which the lineage of a decision is made real is exactly right. But documentation only works if the person reading it still has the expertise to interpret it. When that expertise atrophies, the audit trail becomes a paper exercise.
We talk a lot about AI governance at the tool level. The harder conversation is about preserving the human judgment that makes governance meaningful in the first place.
I agree, it’s probably the hardest part in consulting too. I’ve been working on a Claude Code implementation of thematic coding that gives me far more reassurance compared to the chat version - but that still has gaps naturally!
“People who treat responsibility, accountability, auditability and transparency as ongoing professional obligations rather than a compliance exercise to be completed and filed are the ones most likely to be considered trusted partners.”
100% this is a fantastic point. for me i would be so much more comfortable with this attitude from my partner.
Thank you! I find it’s easy to start feeling like verifying what AI says is a bureaucratic exercise. In my experience it comes from the fact that the average user doesn’t know that gen AI is probabilistic and very different from a google search - in that AI makes a claim through its response, whereas searches just say ‘here are a few links for you to figure things out’.
100%. and probability distributions work well in some places and not great in others. we are essentially forcing everything into this paradigm, then finding ad hoc fixes
This is a timely and necessary piece – not because “AI wrote it” is new, but because “you signed it” is the part people still treat as a formality. You frame the real shift correctly: authorship is turning into accountability, and responsibility doesn’t vanish just because production got easier.
I also appreciate how you avoid the lazy binary (“ban it” vs “embrace it”) and push the practical question: what does a signature actually certify in an AI-assisted workflow – intent, accuracy, originality, or just publication? Once AI becomes an invisible layer, the easiest failure mode is moral outsourcing: “the model did it,” therefore nobody did. That’s why governance has to evolve around process (review, traceability, disclosure norms), not tool policing.
Thanks, Mila! It’s really important to think about all of this when the any tool shifts from amateur/personal use to professional use 🙏🏼
The ChatGPT as buffer story is the sharpest part of this.
In enterprise delivery I have seen the same dynamic - not with client deliverables but with system decisions. The AI suggested this architecture.
The model recommended this approach and the tool becomes a way of distancing the team from their own judgement.
When something breaks at production nobody owns the reasoning because the reasoning was outsourced.
The auditability section is where this lands hardest for me as an engineering manager.
The Microsoft research you cite describes exactly what I'm watching happen inside engineering teams right now. When junior engineers use AI to generate code they don't fully understand, the review step becomes ceremonial. A senior engineer reads it, it looks plausible, it passes. Nobody asked how it was produced or whether the person who submitted it could explain the reasoning.
The deeper problem is what happens over time. Engineers who never had to solve certain problems from scratch don't develop the intuition to catch errors in AI-generated solutions to those same problems. The audit trail exists the PR, the review, the approval but the human capacity to make that trail meaningful has quietly degraded.
Your point about documentation being the mechanism by which the lineage of a decision is made real is exactly right. But documentation only works if the person reading it still has the expertise to interpret it. When that expertise atrophies, the audit trail becomes a paper exercise.
We talk a lot about AI governance at the tool level. The harder conversation is about preserving the human judgment that makes governance meaningful in the first place.
I agree, it’s probably the hardest part in consulting too. I’ve been working on a Claude Code implementation of thematic coding that gives me far more reassurance compared to the chat version - but that still has gaps naturally!
I have also trialing out several chat bots for companies and some people and Claude does feel and perform better than the rest.