AI-generated work - The real question is not “was AI used?” but “how was it used?”

We are at an awkward stage with AI-generated work.

Nobody wants to receive a piece of work/proposal/deliverable where a human has simply typed a single prompt into an AI tool, copied the answer verbatim and passed it off as their own thinking. That feels thin (because it usually is! - the AI output "looks good" but it is not rooted in the broader context, the experience and human insight). Usually in this type of workflow the problem is not that AI was involved but it's that the human contribution is reduced to almost nothing, no judgement, no context, no challenge/refine loop, no responsibility.

However, I think it is equally wrong to jump from that as a concern to a conclusion that all valuable work must still be written entirely by a human, from the first blank page to the final polished piece.

Good work is shaped. It is researched, tested, argued with, rewritten, challenged, refined and edited. We can now do all of that with AI assistance.

The final deliverable is usually not the place where most of the work is visible. The hard work may have happened upstream, for example framing the problem, understanding the audience, gathering evidence, identifying what's important, drawing on prior experience, making trade-offs, deciding what matters and knowing what to leave out.

AI changes the tools available in that process but it does not remove the need for the process itself.

There is a big difference between AI replacing the work and AI helping humans produce the work.

If someone gives an AI model a vague problem and accepts the first response as the final answer, that is not a good use of AI. It creates shallow thinking, generic output, factual errors and laziness that will be rightly criticised.

However, if someone has done the thinking, gathered the inputs shaped by their experience and insight, reviewed the output, corrected it, challenged it, added nuance and iterated with the tool, then the final AI-assisted write-up may simply be the tidy step at the end of a very human process.

This is also where the debate becomes controversial. Some people will argue that any AI-generated text is less authentic. Others will say that what matters is the quality, accuracy, originality and accountability of the final work.

There are also real concerns about trust, transparency and creativity.

At the same time, AI use is already becoming normal in organisations and is becoming/already a key part of how work gets done. So the useful question is not only “Did AI write this?” but:

• Did a human understand the problem?
• Did a human direct the work?
• Did a human check the facts?
• Did a human make the judgement calls?
• Did a human take responsibility for the output?

AI can generate paragraphs, structure, options, summaries and alternatives but it cannot be responsible in the way a human can. It does not know the client, the context, the politics, the trade-offs, the risks or the consequences. The human still has to own the work. This is not simply a question of “human-written” verses “AI-written”, its work that has been thoughtfully produced verses work that has been lazily delegated.

AI assisted work can be high quality and useful when it is human directed, human reviewed and human owned. It can help turn messy thinking into clear communication. It can help polish a rough argument. It can help explore alternatives. It can help make a final deliverable cleaner and more accessible, but it should not become a substitute for doing the thinking.

AI can help with the final expression of the work. It cannot be allowed to replace the work itself.