Key Points
- AI is currently a net positive in my work, but I think that is because I am slowly seeing the nuance of where it is helpful or not.
- In my core areas of expertise, I’m cautious. AI can definitely speed things up, but it might come at the cost of actually getting less skilled in those areas.
- In satellite skill areas, AI has been energising. It has helped me re-engage with things like web design, data analysis and communications, where I had started to feel too far behind to properly participate.
- It isn’t so much about productivity as it is feeling like you can play in areas that you’d otherwise not have done so.
I’ve been using AI in my work for a little over two years now. My main platform is ChatGPT, though I’ve also experimented a bit with Gemini and I keep eyeing off Claude in a suggestive way.
Overall, it has been a net positive for me. But not an uncomplicated one.
One of the first places I started using AI was in writing blog posts for work. I manage the content on the student wellbeing portal BetterU. I’ve been writing on there (and its predecessor) for 9 years. The benefits of adding AI to my workflow were immediate: I could draft faster, clean up structure, smooth out clunky sections and get from idea to publishable piece more quickly. Given that a significant chunk of the content on there is informational (e.g. promoting an upcoming event), AI could help me create promotional content quickly and easily.
But there were definite downsides. I remember about 6 months ago, comparing my older posts to my newer ones and realising that whilst AI was helping me create cleaner content, it didn’t always sound like me. My particular “voice”, for better or worse, is a bit chaotic, absurd and messy. When working well, it attempts at dumb humour and pointless tangents and sometimes succeeds. And I found that voice missing from the pieces that had heavy AI input. Worse still, because I was outsourcing content generation to AI, I was losing the ability to imbue (my favourite word at the moment) content with my voice. It wasn’t just that my voice was disappearing in the content itself. It was disappearing in me as a writer because I wasn’t requiring it of myself on a regular basis. As all dads know, stop making dumb jokes and you stop being able to make dumb jokes.
| This year particularly, with my writing, I’ve been trying to find a decent hybrid approach. I dump all my ideas into AI initially. It produces a starter draft. I then take that out of AI to work on separately, then back to AI for pesky sections or proofreading. There might be a few interactions of this sort. The goal is to use AI for speed, structure and momentum, but then deliberately bring myself back into the piece. What you’re reading now is one such article. |
That is the caution that I mentioned in the title. Outsource to AI too much of the work that defines your work identity, and you may find yourself deskilling. Losing parts of yourself that you spent a long time cultivating. This is certainly a threat I’ve noticed in my work. I am a psychologist working in mental health promotion. That job requires a lot of writing – for articles, for presentations, for campaigns and resources. I think to do it well you need to find your ‘voice’ – the styles/vibes/personality that characterises your writing. I’m not a particularly good writer – functional at best. But when my writing works it is generally because it has some personality. It is the boring looking dog that seems to somehow charm you anyway (please note, I achieve this rarely). And I was seeing that personality disappear from my work AND finding it harder to access that personality in any writing I was doing.
I’ll give you another example – reading and digesting research articles. Definitely a part of the job of a mental health promotion psychologist. Now, you can ask AI to summarise articles for you, in great detail and for specific purposes, and sometimes that is exactly what you need. I’ve even got a detailed set of questions for AI to break an article down into core components (table 1).
Table 1: Feel free to cut and paste my research article digest questions for use in AI
| Research Digest Questions and Prompts |
| 1. Big Picture / Context What’s the broader context of the research? Why is this topic important or relevant right now? What problem or gap in the literature is this research addressing? What issue or challenge is this study trying to better understand or solve? Why might this be important to people working in education, wellbeing, or mental health today? What conceptual models or theories underpin this study? 2. Research Aims & Objectives What specific questions or hypotheses does the study address? What did the authors aim to achieve or clarify with their research? Is the study aiming to compare, evaluate, propose, or consolidate knowledge? 3. Methodology & Design What was the overall research design? (Experimental, correlational, qualitative, etc.) Who were the participants? (Demographics, how recruited, how representative?) What methods or procedures did the researchers use? (Interviews, surveys, experiments, observations?) What measures or instruments did the researchers use to capture key variables? Are there any notable strengths or weaknesses in the study’s design or methods? For secondary research (like reviews): New: What types of studies were included (e.g., RCTs, observational, qualitative)? What inclusion/exclusion criteria were used? How was study quality or bias assessed (e.g., AMSTAR, GRADE)? What methods were used to synthesise findings (e.g., meta-analysis, narrative synthesis)? 4. Key Findings & Results What are the primary outcomes of the research? Were the results statistically and/or practically significant? Are the results consistent or inconsistent with previous findings in this area? What unexpected findings or insights emerged? Which findings were strongest or most consistent across studies? What findings were less conclusive or had mixed results? 5. Interpretation of Results How do the authors explain their results? Are their interpretations reasonable given the data and methods used? Are there alternative explanations for the findings? What assumptions do the authors make in drawing their conclusions? How does the research build on, challenge, or extend existing frameworks? 6. Implications & Applications What are the practical implications of this study’s findings? Who could directly benefit from these findings, and how? What actions, recommendations, or changes might be inspired by this research? If someone were to implement these insights, what first steps could they take? What types of actions (e.g., program design, policy changes, personal behaviour change, teaching strategies) does this research seem to support? 7. Limitations & Critical Considerations What limitations did the authors acknowledge? Are there other limitations or critiques not mentioned in the study? How generalisable are these findings beyond the specific context studied? What caveats or cautions should readers bear in mind? What are the limitations of the evidence base itself (e.g., missing populations, outcomes, contexts)? What cultural, social, or systemic factors might influence these findings? Are there equity considerations in applying this research? 8. Future Directions What unanswered questions remain? How could future research build on or clarify this study’s findings? What could be the next steps in practical or applied terms? What are the research gaps, and what applied or policy directions could address them? 9. Accessibility & Communication How clearly and engagingly are the findings presented? What terminology or concepts need simplifying for broader understanding? What analogies, metaphors, or practical examples could clarify key concepts? What image, metaphor, or campaign message might capture the core insight of this study? If you had to explain this study in under 2 minutes to a colleague, what would you say? |
But each time you get AI to digest an article for you, you’ve sidestepped doing it yourself and practising the important skills of research engagement: slowing down, reading, contemplating, connecting it to existing knowledge, being critical, feeling confused, asking “what the fuck?” to the statistical methods, and occasionally enjoying or battling the authors’ language.
On the surface, it looks great. You appear to be getting way more done. Your colleagues are temporarily amazed at your increased output. But underneath that you are getting stupider. Use AI too much (or crudely) and you start to outsource your actual core knowledge and skill areas. You aren’t learning as well. Your expression becomes too sanitised. Your actual abilities atrophy. I can see a world in which you get worse at your job, the more you let AI into it.
The title indicated a positive story – where is that?
So yeah, I did lead with the bad news, but I do think there is a positive side here as well. It relates to using AI for what I’ve found myself calling “satellite skills”. These are skills that aren’t your main expertise but are still likely to be useful to your work. For me, at the moment, these are web design, data analysis and communications.
I stopped doing much web design in my twenties, once it became clear the field was moving much faster than I could keep up with. I was studying psychology and that took up most of my bandwidth, and so my old HTML skills (mostly derived from a Dummies Guide to HTML) quickly atrophied. Had I been better at web design than psychology, then my career might have headed in a different direction, but that skillset got frozen in time.
A similar thing has happened with data analysis. I did a PhD in Clinical Psychology and data handling and analysis was definitely a part of that. I then worked in projects and research for the subsequent decade which meant more data handling and analysis. I was never a gifted statistician (my unique gift in this world is being average at everything) but I was handling data regularly enough for it to be a part of my work. But when I shifted from a research role into mental health promotion (which has a more educational and clinical flavour to it) in 2017, there wasn’t really any regular data to handle, so those skills shrunk.
In both cases, the longer I spent outside of those areas, doing those tasks, the further behind I felt. Eventually, opportunities to re-engage didn’t feel exciting. They were just reminders of how far behind I was. Modern websites were built using technologies that were well beyond my old skills. Modern research papers were using statistical methods I hadn’t kept up with. Tools like Power BI sat in the category of “probably useful, but not something I have time to learn properly.”
Using AI has shifted the dynamic on both.
On the website front, my skill level meant being relegated to making small changes within existing templates or putting in requests to very busy teams for more advanced assistance. But with AI, I can now explore and often carry out more significant modifications, mock up design changes, imagine new functionality and even develop working prototypes. I can get guidance on how to implement some of those changes in the different systems in which I work such as AEM and WordPress.
That doesn’t make me a web developer. But it does make me much less passive in the process. I can be less reliant on the web team for every small idea, and when I do need their help, I can be clearer about what I’m asking for. For example, we’ve been needing to update the website for our mental health campaign Good Vibes Experiment for a while now. Rather than give the web-team some text instructions for what I was after, I was able to give them a detailed mockup. [just as a quick aside, I’ve found Netlify incredibly useful for hosting side ideas].
I am now noticing a similar thing with data.
Recently, I had to work with service data from our counselling service. Without AI, I was relying on a fairly limited manual, some past guidance from software representatives and whatever Excel skills I still had lying around. I could do a basic export and tidy things up, but anything more complex would quickly need to be handed over. Now with AI, I’ve been able to go further. I can discuss with it specific things I want to do with the data, specific analyses I want to run and troubleshoot more complex questions about how we use the data for service improvement (I want to be clear that I do not put service data into any of these systems). AI is like having a pretty capable data scientist sitting next to me.
And that, for me, is where AI has been shining most in my workflows. Less about replacing my core skills and more about giving me the scaffolding I need to develop satellite ones. So I find myself, at the time of writing, thinking that, in your core craft, AI should probably be used carefully. It can speed things up, but it can also tempt you to skip the very activities that deepen your expertise. But in satellite skill areas, AI can be energising. It can help you return to things you had quietly abandoned because you felt too far behind. It can let you play, experiment and prototype without needing to become an expert first.
I’ll finish up with a few comments about the third skill area that I haven’t mentioned yet: communications.
I manage a few newsletters, and we are currently upgrading the main one that is attached to the BetterU site. AI has helped me create newsletter mockups, test HTML templates, explore different designs and think through different communication scenarios. So not just the technical side of putting a decent looking newsletter together but also thinking through the bigger challenge of how we communicate in an information-saturated landscape. This is knowledge at the intersection of a field I wasn’t trained in (communications) and a field I was (psychology). Playing there (which I probably wouldn’t otherwise do) means growing my circle of competence, so feels like genuine professional development.
Take home message
On any given day, I think I can construct arguments for and against AI’s spread into our work lives. And I don’t yet have a meta conclusion regarding its value.
But I have been using it long enough in my workflows to arrive at, at least one nuanced perspective.
That is, for the core skills that make up your professional identity, be careful not to outsource them too much to AI because something will be lost in both the output and your ongoing skill level in those areas. Continue to push yourself to grow in those areas. Overuse here risks a deskilling over time.
For satellite skills, I think you can maybe lean in a bit more, especially if doing so, allows you to play in areas that you simply wouldn’t have previously because of time, knowledge, or the steepness of the learning curve. Use here has the potential to help you connect your core skill set to other fields and areas.
What is a skill area, outside of your core competencies, that you think you could develop if you used AI?
Dr Gareth Furber 