I’m a total convert to ChatGPT. It’s got to the point where I struggle to imagine completing many work tasks and some personal tasks without it. While not quite daily in reality, I do use ChatGPT every week, and often very extensively. The productivity it makes possible is insane. On the best days, I feel like there are 3 or 4 of me. Perhaps more.
Tens of millions of people fit that description, but that still leaves many more millions who have either never used ChatGPT, or only dabbled with it without fully realising its potential.
For those readers, in this article I’m going to detail (in no particular order) where I’ve found ChatGPT to be massively game changing. Note that other capable AI assistants exist – and this list is almost certainly likely to be equally applicable to the likes of Claude, but I haven’t used them to the extent that I have ChatGPT.
1. Data preparation
Suppose you have data in rows that need stacking into columns or vice versa, break down messy data formats or do a text to column operation where the separator is longer than 1 character or varies throughout the string. Normally, I’d be writing Excel formulas, SQL or Python for the task – and that’s still very much worth knowing and investing in long term. But for those quick jobs or when you’re on a tight timeline and don’t need to build something repeatable, ChatGPT is incredibly helpful thanks to its built in ability to read and execute code on uploaded data. This isn’t a case of having it write formulas or teach you how to do something, but just doing it and giving you back a reformatted and downloadable output.
I’ve not only been impressed by its ability to do the data prep, but also the brevity of explanation required. For the aforementioned exercise of turning data in rows into columns, I used to write out in quite a lot of detail where I expected each bit of data to end up, but have now come to realise that a very short and objectively quite poor instruction suffices.
In practice, this means I can solve data formatting issues on the fly, with minimal disruption to my workflow.
While I don’t recommend skipping the fundamentals of tools like Excel and SQL (you’ll absolutely need them), ChatGPT’s efficiency is hard to ignore when it comes to one-off data wrangling.
2. Summarising text
Copying and pasting vast text tracts, or uploading long documents (multiple of them if needed), and simply asking for a summary (providing a specific structure for that summary if desired) has saved me hours and hours both in my professional and job seeking life.
At 128,000 tokens, or around 85,000 words, the “context window” (the amount of content the AI can consider when drafting its response) is huge. To put that in perspective, that’s longer than a typical novel. When you stop to think about it, the ability for anyone to have a text that long read and summarised within seconds is absolutely insane.
As a recent job seeker, I’ll give you a concrete practical example of this in action. As a job seeker, especially in today’s job market, you’re probably looking at a lot of job specs every day. It’s important to quickly understand what each job actually is, what you’ll be doing on a day to day basis, and what type of skills and experience successful applicants will have.
If you work in any sector where job titles are too ambiguous or weird to do this for you (recent highlights for me have been “ghost buster”, “chief of staff” and “founders associate”), you have to rely on the spec. Unfortunately lots of recruiters and companies write incredibly long and not very well structured job specs, meaning a lot of work to read and parse these specs (perhaps the worst examples of this I have seen have been put out by the civil service, which usually feel like a novel and where even a careful read has sometimes left me largely clueless).
What to watch out for: Ask ChatGPT to include references to specific documents, pages or paragraphs to make it easy to fact check its summaries.
3. Answering questions by consulting a knowledge base
Depending on how complex your needs and how big your knowledge base is, you can use any of the following approaches to provide the AI the relevant info and then extract answers about that info.
- Custom instructions – standing information that ChatGPT applies to all your chats with it. There is a low limit on the length of these instructions, but this approach can be useful if you only use ChatGPT for one specific purpose and the information is not very extensive.
- One off chats – if your knowledge base need is a one off, then you’re probably going to be using this approach and just uploading or pasting the content in a normal chat window before firing off your questions. Often, this will continue naturally from my second use of ChatGPT above (text synthesis).
- Saved conversation – a particular saved chat that you regularly return to with the knowledge base uploaded or pasted near the start. This chat will function similarly to a custom GPT but if you use it very extensively, its possible that the importance of the originally uploaded content will start to fade in the AI’s “memory”, or even vanish from it if you manage to go beyond the context window.
- Custom GPT – a version of ChatGPT saved for your use (or the use of your colleagues) that always has access to and instructions regarding an uploaded corpus of text and data files. This is what you need if you expect to query the knowledge base repeatedly or many people need access to it (and frankly iss not much more complicated to setup than starting a normal chat).
There are obviously countless potential uses for a knowledge base but by way of example I’m going to extend the job seeking use case above. Having summarised a job spec via ChatGPT, what could be a more natural next step than working out if you’re a good fit for the job, where you have gaps in necessary skills or experience, and which achievements in your career you could highlight for each of the requirements of the spec?
Here, the knowledge base is your CV, personal statement, linkedin profile or any other record of your skills, experience, career history and achievements. Now not only have you shaved off a chunk of time in working out what the job actually entails and who they’re looking for, but you’re already positioned to make a decision on whether to apply and what to include in a custom CV and cover letter.
Note for this application, you want a fresh one off chat with the basic ChatGPT or with a custom GPT – custom instructions won’t cut it for a knowledge base this long and you don’t want to use a single saved conversation as it will risk polluting the context window of the current job of interest with other job specs you have discussed before.
4. ChatGPT as an Excel formula assistant
This works when you already have a good grounding in Excel or can at least describe the input and expected outputs. In fact, my most common use of ChatGPT for Excel formulas is not to do things that I can’t do, but to write lengthy formulas that I can do but can’t be bothered to write out – those long nested formulas where writing them is a pain and you’re probably going to misplace a comma or a parenthesis somewhere. For example the other day I had cause to iteratively xlookup values from about 15 columns to try and find a match, and writing or clicking that out manually is much slower than just saying “nest 15 xlookups, in turn looking up cells B2-P2 in column A and returning column D”.
In a way, ChatGPT is like having a patient tutor sitting beside me, guiding me through the logic behind each function and formula step-by-step. And because it provides explanations, I’m not just copying and pasting; I’m actually learning why certain formulas work in certain ways. So it’s not only a time-saver but a learning tool.
I recommend standing instructions to have ChatGPT explain the formulas it’s providing, which helps to catch the rare errors it makes and also makes the tool act as a tutor not just an assistant. An important side effect of this for me has been that I’ve learned a lot of formulas and functions that I wasn’t aware of – or at least learned how to use them better. For example, in the last couple of weeks I’ve picked up or honed my use of the lookup function and formulas that output and handle arrays like unique and transpose.
What to watch out for: Sometimes the AI will take things too literally. For example, I once asked if there was a way to improve a formula so that I wasn’t writing out the same xlookup twice, with the intention of finding something faster or more efficient. It found a solution of course, but the replacement was more complex to write and execute.
5. Coding
Actual developers and software engineers will want to explore dedicated AI coding tools (many of which are interfaces into GPT built specifically for the job), but for casual or occasional coding I’ve found ChatGPT to be incredibly useful.
I’m nobodies idea of a programmer, but I know enough to be dangerous. I can describe what I want, and read code well enough to usually understand what it’s doing. For SQL, Python, PHP and other languages, I’ll often sketch out the logic I want in simple sentences, and ChatGPT generates code that does just that. While it’s not a substitute for knowing how to code, it’s an exceptional companion for building functions, fixing bugs, custom tools and more. In fact, it’s largely thanks to this application of ChatGPT that the site you’re reading exists.
And it’s not just writing code from scratch; if I have a piece of code that’s behaving oddly, I can paste it into ChatGPT, explain the issue, and it’ll provide a modified and easily copied block of code to make it work or run more efficiently.
In my view, as implied above, this only works well if you have a basic understanding of what the code should look like. Knowing a bit about the structure, syntax, and logic of the language you’re working with is key – enough to follow what the code’s doing and spot if something seems off. With that in place, ChatGPT becomes an invaluable coding partner, speeding up development and making troubleshooting smoother. But like it’s use as an Excel formula assistant, with explanations required as a standing instruction its also going to act as a tutor at the same time.
