What am I still bringing to the work?
AI has made preparation faster. A reflection on the thinking, challenge and personal judgement that still need to happen before the work is finished.
Building & Experimenting · AI · Critical thinking · Judgement
Using AI at work: is a polished deck proof of thinking?
YouTubeThis article is inspired by real workplace and leadership experiences. I do not use real names, company names or identifying details. Roles, teams, organisational structures, projects and business context may be removed, generalised or combined to protect the people and organisations involved while preserving the underlying leadership reflection. Where conversations are referenced, wording may be edited for clarity and anonymity.
AI has made me more efficient. I can work through material in an afternoon, turn it into something usable and spend another hour editing what doesn’t feel right. I appreciate that. I have no particular desire to make a task take longer simply so I can feel better about having done it.
But that efficiency has also made me question the value I’m bringing. If so much of the visible work can happen this quickly, what was I contributing before? And what am I contributing now?
There is a slightly uncomfortable moment between being delighted with your productivity and thinking, “Fuck. Have I become shallow, or have I just become very efficient at being lazy?” You look at the amount you have produced and feel pleased, then wonder whether you have become better at the work or simply better at generating something that looks finished.
When I look more carefully at how I’m using AI, the useful part is rarely a single request followed by a result. There is a discussion. I challenge an answer, consider the questions I expect to face and work through questions the tool raises in return. Eventually, some of that becomes a slide. Then I go back through it and edit the parts I can’t stand behind or wouldn’t say.
That is still work, even though it looks different from doing everything myself. What I’m trying to understand is which parts of that process are helping me think, and which parts I might be allowing myself to skip.
The presentation can be finished before the thinking is
I’ve seen the enthusiasm that comes with producing a substantial presentation using AI. Forty slides appear, and the sheer amount of material feels like progress. You read the first slide and agree. You read the second and agree again. By slide ten, you realise you’ve met this point several times already, although it has changed its shirt and is now calling itself a strategic priority.
What unsettles me is how reasonable each part can sound while the whole thing remains difficult to use. There is plenty of language, but it can still be unclear what decision is being proposed, what matters most or what the person presenting it actually thinks. Somewhere around the middle, I find myself wondering what the fuck we are actually being asked to do.
I don’t want to pretend this started with AI. We were perfectly capable of building long presentations around an unresolved thought before any of these tools arrived. We just had to manufacture all forty slides ourselves, which at least introduced a natural limit through boredom.
I do think I now need to be more careful about allowing the existence of a polished document to reassure me that the difficult work has happened. It is also easy to direct that criticism at somebody else’s slides. I am considerably more likely to recognise repetition when I didn’t spend the afternoon producing it.
In my own work, I can imagine mistaking familiarity for coherence. I understand what I meant, so I assume the material says it. By the time I’ve read the same argument enough times, I may no longer notice which parts I am supplying in my head.
That is why editing the wording can’t be the whole review. I want to be able to explain why a recommendation belongs there, what information supports it and what would make me reconsider it. If I can only defend the way it has been phrased, I probably haven’t finished.
The same applies when AI gives me something I immediately agree with. Agreement is pleasant, particularly when the answer makes a half-formed idea sound intelligent. It is also the point at which I have the least emotional incentive to keep checking.
I’m spending more time challenging it
Something has changed in the way I use these tools. I find myself spending more time debating the response than I did originally. I’m less interested in specifying exactly what the finished answer should look like and more interested in examining the thinking that might get me there.
Some of that comes from experience. I have had to struggle through problems without a tool producing an immediate answer, and I think that gives me useful reference points. I can recognise when something has moved too quickly past a difficult assumption or when a confident explanation hasn’t really addressed the situation.
But I need to be careful about how much credit I give myself for that. Experience can help me detect a weak argument. It can also make me attached to my own preferred explanation. Spending twenty minutes disagreeing with AI doesn’t automatically mean I’ve thought critically; I might simply be refusing to leave until it agrees with me.
That would be a fairly embarrassing use of the technology. I’ve acquired access to a tool that can help me examine different perspectives, and I’m using it to chair a meeting in which I am also the only acceptable conclusion.
The more demanding version of this process would leave open the possibility that I change my mind. I would want to examine a reasonable alternative, notice what information is missing and distinguish an actual constraint from something I have assumed. Those are useful things to ask of the conversation, even when they make it slower.
That is where I find the idea of AI as an accelerator helpful. There are parts of preparation I’m happy to move through more quickly. I can explore different structures, test an explanation or get a rough version into a form I can examine. The time saved gives me an opportunity to spend longer on the parts that require judgement.
Whether I take that opportunity is another matter. I could just as easily use the saved time to produce another document. I don’t think having access to a faster process tells me very much about the quality of the choices I make with it.
Sounding like me is only part of it
I’ve spent time giving the tools I use guidance about my voice and how I approach things. I want the result to sound like something I might actually say. When I read material that bears no resemblance to the person presenting it, I find myself wondering how much of their own thinking has survived into the document.
That matters to me as a leader. The words I put in front of people create expectations about what I believe, what I’m asking of them and what I’m prepared to take responsibility for. I don’t want to make a commitment in language I would never have chosen, then discover that I haven’t properly considered what it implies.
Still, familiar language can give me false reassurance. A response that sounds like me may be easier for me to accept precisely because it sounds like me. Adding the occasional “fuck” doesn’t establish that the argument is good. It may just mean I’ve successfully personalised the nonsense.
There is a tension here that I haven’t fully resolved. I want help expressing my thinking clearly, but I also want the thinking challenged. If I make the tool increasingly good at reproducing my preferences, I need to watch that I’m not making it increasingly comfortable to agree with.
I would rather encounter an objection that makes me reconsider something than receive another elegant version of what I already believe. That is easy to claim in principle. The real test will be how I respond when the objection concerns a piece of work I’m already pleased with.
This is also why I’m uneasy about treating a final review as a quick check before sending. If my name is attached to the work, I need enough understanding to discuss it without hiding behind the document. I can use help preparing the argument, but the people affected by it will still expect me to explain the decision.
What does this mean for people still learning?
I worry about people entering leadership, or starting their careers, with tools that can produce plausible answers before they have had much opportunity to develop their own judgement. They may appear highly efficient while missing some of the practice that helps a person recognise a weak assumption or an answer that doesn’t fit.
That is a concern I have, rather than a conclusion I can draw about an entire generation of employees. I also need to resist turning into the person who believes everyone should suffer through the same processes because that is how I learned. Some of the difficulty I experienced was useful. Some of it was simply inefficient, and I’m not sure it deserves a commemorative plaque.
AI could be useful in that learning process too. Someone might use it to explore alternatives, ask questions they would otherwise hesitate to ask, or examine an explanation until they genuinely understand it. The important distinction, to my mind, is whether the person is participating in that process or mainly accepting the result.
As a manager, I would therefore want to look beyond how polished the finished work is. I’d be interested in the choices behind it: which alternatives were considered, where the person disagreed with the tool and what remains uncertain. Those conversations could tell me more about development than either praising the speed or disapproving of the assistance.
I’m glad AI has made parts of my work faster. I don’t want to give that back, and I don’t think slower work deserves automatic respect. What I do want is a clearer view of what I’m contributing once the production becomes easier.
For now, the most useful place to look seems to be the parts of the work I still disagree with, the assumptions I have changed and the decisions I can explain without reading from the slides. If all I can show for the afternoon is a larger quantity of material that sounds like me, I would want to spend a little longer with it before calling it finished.
Continue exploring
- Article
The Polished Leader and the Unplugged Leader - AI Skill
AI Workflow Designer - Free resource
Reflection without a record is just mood
