In this article, I want to explore something I think most people haven’t quite caught up with yet: the real use of AI in their work. Being more immersed in the AI field than most people around me, I sometimes feel permanently behind; everything moves fast enough that catching up feels like a full-time job. However, when I talk to people who are not as deeply involved, I notice something else: how much most people genuinely don’t know about what AI can already do for them.
To some extent, and maybe I’m wrong, I think most people still picture AI as a chatbot. Or at the very least, they know it is capable and still have not changed how they actually work. Useful for a quick answer, a bit of email help, rephrasing their writing, and helping them think. Essentially, a smarter search bar sitting on the side of their screen. This picture is already outdated. AI is not staying inside the chat window. It is becoming a layer that wraps around how work itself gets done: building ideas, organising, automating, and in some cases even covering roles you would have hired for. The shift is less about a clever tool and more about what one person can now reach for on their own.
Three areas have caught my attention recently, places where I keep noticing the change in what individual capability looks like. They are exciting but also scary in a way, as AI becomes increasingly dominant in how we work. My main aim here is to show how fast this is moving, and if any of it prompts you to start using AI differently, I am happy to have given you some insights.
1. Building an App, Software or Dashboard
The distance between having an idea and having a first version of that idea has collapsed. Honestly, it is absurd (in a positive way) that if you want to, you can build most types of apps or software you want to. Something that would have meant hiring a developer, or learning to code over a weekend, or giving up, can now be a working prototype in a day.
What this changes, beyond speed, is the relationship most of us have with software. For years, if you needed a tool, you bought one. A subscription. An app. A piece of software designed by someone who imagined a user roughly like you and built for the average of all those imagined users. However, now with AI using tools like Claude Code, Codex, and Lovable, you can build the tool yourself and adapt it to your specific needs.
Of course, ideally, software developers will still be needed if what you are building will go to market. Moreover, specialised software still has its place, don’t get me wrong. This is just how the law of efficiencies in our economy works: letting someone else be good at it saves you time and money. For instance, a platform built for the solar industry, a dedicated literature review tool, or anything that depends on deep domain logic will still have its place. However, the threshold for “I’ll just build this myself” has changed massively. Last month, I built a small expense dashboard for the way I think about money, instead of downloading yet another app and reshaping my habits around its categories. It is faster, more specific, has no compromises, and it aligns with what I want. To give you another example, I am also building my own health dashboard, connected to all my different devices like Whoop and blood tests. Again, I would have normally downloaded from the app store, but why not just build it myself for my own needs? I get it: convenience matters, and even building something with Claude requires a level of learning which some people might not be bothered with. However, the main point is that you can now do it, and the business applications are huge.
The cleanest analogy for me is Excel or Word. AI is becoming a universal layer in the same way those once were: something you open and bend to whatever the project needs, except that the output is no longer just a spreadsheet or a document. It is whatever you actually need it to be.
2. AI as your Personal Assistant
If building feels like the more obvious story, this one is quieter and probably more relevant to most people’s actual day: AI as your personal assistant.
Much of modern work is not deep or creative. It is coordination. Sorting. Formatting. Searching and drafting the same kind of email for the fourth time this week. AI can be used as a personal assistant, not for the occasional task, but as a second brain running alongside your day.
Is AI already at a point where it can actually be like a proper personal assistant, as we see in futuristic movies? No, it is not there yet. However, it can already handle many specific functions if you set it up with clear instructions. It needs a bit of setting up using systems like Claude, writing SKILLS, and other steps we can cover in another article. However, it can already be done. Many of my friends and colleagues ask me: how, and in what situations? Some examples I can give you are: optimising a schedule, drafting and triaging emails, tidying folders, reformatting a document that someone sent in a strange shape, pulling information from a long PDF, or turning a page of messy notes into a clean list of next actions. For instance, I have developed specific skills and custom Claude setups so that it can do a series of tasks for me every day or when I need it to: going over my emails to flag what is important and what is just spam, summarising notes and actions from a meeting, helping me with literature reviews, and organising folders. Honestly, all the stuff that I would usually either need to do on a Sunday before my work week began or find time for during the week. In a way, all this is priceless: when I was a consultant, we had a timesheet, and the less time spent on non-billable organisational work, the better. This is gold for time-based work.
The small, invisible frictions that drain attention every day quietly add up in a working week. When you find yourself wishing you had an assistant for certain tasks, it is worth thinking about whether AI could handle them instead. In fact, there is an even harder question to ask: what if AI didn’t just help with the task but owned the whole process?
3. Not just a Personal Assistant but Whole Workflows
There is a step beyond the assistant idea, and it is where things start to feel different. It is not just about helping as a personal assistant would, but almost like hiring someone to handle a specific workflow entirely. The way I view it is that there’s a difference between AI helping you with admin tasks and AI owning a process entirely. Think of it like a personal assistant doing a specific task you asked for, except this goes a step further, because the work requires deeper expertise for the task at hand, not just admin.
In my opinion, most people underestimate how much of their work is already in this category. Not just obvious admin, but whole recurring workflows: the kind of thing that happens every week and quietly consumes hours. If you have not thought about it yet, stop and think: what are the things I keep doing every time? Even if every time it is slightly different, that is fine, because that is the power of AI: flexibility.
A friend of mine works at a real estate firm where every apartment listing comes with a custom brochure and a social media post. Brochure layout. Copywriting. Image selection. Scheduling. The whole sequence. Most of that can now be handled by a system that runs almost end-to-end, with her stepping in at certain points to review, make decisions, and tweak. Now she can focus on other more important things, and what was a big chunk of her week is now optimised. The job has not disappeared. The hours have.
The point is not one dramatic automation. It is that dozens of small recurring tasks can stop needing manual attention, and that adds up fast. What gets freed up is the part of the work that actually requires judgment, taste, strategy, real human input: the part most people would say is the reason they are doing the job in the first place. Maybe soon we will be at risk of AI taking over many of our jobs. I agree it is a concern, but we are not there yet. The point is to identify where certain workflows can be handled by AI, fully or partly. The right question to ask yourself is: if I had to hire someone to cover part of my job, what would I want them to do? This is the best way to identify where workflows can be handed to AI to make your work more efficient.
The Next Step: AI Covering Roles
There is another development worth paying attention to, one that some have started to implement but that is not quite there yet: the idea of using a set of AI agents to cover a role that in the past you would have hired someone to do. The distinction from workflows is this: a workflow automates a specific task. What I am describing here is covering an entire role. Not one process handled by AI, but a whole function: everything a writer does, or an analyst, or a junior developer, or a researcher. Not perfectly, but enough to change what is possible for one person sitting alone with a laptop. Of course, this is also happening with all the new AI start-ups out there using AI for a specific workflow. Products positioning themselves as an AI writer, an AI analyst, and an AI researcher. The category exists; the question is how quickly it becomes reliable enough to trust with real work. In my view, we are not quite there yet, but this area is developing quickly for certain types of roles.
Closing
The exciting part is that more people can now create, experiment, and build without waiting for permission: from a manager, from a budget, from a piece of software somebody else designed.
The unsettling part is that many people have not yet noticed how much of their own work is already in motion around them. A workflow they own is being quietly reshaped. A task they took pride in becoming a one-click operation. A skill they spent years on, starting to look like one capability among many that anyone can summon.
I am not sure how I feel overall. I am definitely excited to see how we can make our work more efficient and to have capabilities beyond anything previously imaginable. Nevertheless, I also feel a little scared at the rate AI is moving, to the point where it could replace many jobs. Maybe that is the real question. Not what AI can do for us, but who we want to become, now that so much of what we used to do for ourselves can be done for us.
