AI For Faster Writing w/ Henrik de Gyor My AI Fluency

Afraid AI will replace you? The real risk is being outpaced by people who use it.

Today, I'm joined by Henrik de Gyor, Chief Digital Officer of My AI Fluency and a no-nonsense digital transformation leader. He's an expert in how to streamline content operations, integrate AI responsibly, and scale workflows without chaos. Expect practical lessons on metadata, change management, and building repeatable systems that deliver measurable results.

In this series we break down practical workflows for meetings, writing, health, and career growth. Listen now and tell us: where will you start?

Remote work creates a special kind of writing pressure: you’re answering messages, switching projects, and trying to sound clear while your attention is split. That’s why AI writing tools are showing up everywhere, and why an MIT study finding roughly 40% faster completion on workplace writing tasks gets people’s attention. 

The real promise is not “AI replaces writers”, but “AI supports remote workers” by adding structure, speeding up drafts, and helping you move from messy thoughts to usable words. For anyone writing emails, memos, LinkedIn posts, blog posts, or podcast scripts, the productivity gain comes from reducing blank-page time and getting to a solid first draft sooner.

Looking for Remote Work?

Click here remoteworklife.io to access a private beta list of remote jobs in sales, marketing, and strategy — plus get podcasts, real-world tips and business insights from founders, CEOs, and remote leaders. subscribe to my free newsletter

Connect on LinkedIn 

Transcript

Welcome And MIT Writing Stats

Alex Wilson-Campbell 0:00

Hello, everybody. It's Alex and Henrik once again from the Remote Work Live podcast. Thank you for joining us. And today, I just want to start with a statistic before I before I we we talk about ourselves first. So, an MIT study with 453 professionals doing realistic workplace writing tasks. So, emails, posts, memos, analysis, found that access to AI reduced time to complete tasks by about 40% on average, while quality scores rose by 18%. So I thought that was a pretty sort of stark sort of figure, a stark sort of analysis of how writing with AI can really help. And I really wanted to have Henrik today, Henrik DeGure of My AI Fluency, to be with me on the podcast because AI, Henrik and I, as much as we're friends, we're also we're also podcasters as well. We've met each other in the podcasting world. But we're desk workers, we're sitting down a lot doing a lot of work, but we're always trying to find ways of improving those efficiencies as far as we can. And Henrik of, as I said, my AI fluency, I thought would be really, really good to hear your take as well, Henrik, on how writing and how AI is affecting your productivity with writing. So as I said, the problem I've highlighted already is the well, not it's not just the writing itself. I'm I'm not somebody who enjoys writing really that much. I'm I'm not too bad at writing, I'm pretty good. It feels slow sometimes to me. There's this seems like there's no structure. Sometimes, again, when you're working especially remotely, you know, you're thinking about 10 different things at once. So the ideas sometimes are a bit constrained. So the aim of this episode is to show how we use AI to turn conversations into scripts, posts, you know, emails and other content, and how you can actually hopefully use some of the things that we've that we use and sort of apply it to how you work. And as I said, it's we're not focusing on the tools, we're not preaching to you, we're not experts in terms of certain AI tools, as it were. Although Henrik has expertise in the in the arena of AI, we're not coming up from the standpoint of being a guru in a certain tool or a certain thing, as you probably see quite a lot on on YouTube. We want to we want to share our workflows with you to kind of show you how you can then apply AI. It doesn't really matter which tool you apply, but as long as it does the job, apply AI to improve your workflow. So, yep. So I myself, as I said, I do a lot of writing, Henrik. I I do I write LinkedIn posts, I write emails, I write blogs, scripts as well. I found myself actually, especially with the podcasting, because we as you and I are both podcasters. So I found myself in a way burning out when I was doing manual script writing, or I was trying to come up with ideas from the top of my head or even on the internet. And you know, when I was doing LinkedIn posts, it it same thing, it's just just the ideation, the the process of writing, all these things, Henry, that really sort of stopped me in my tracks, I think, eventually. And I wanted to find another solution. And a lot of the solutions that I use now, you've referred to me. So even things like otter, chat GPTs, the some of the certain things that you use that I now apply in my in my work. So, Henry, I thought we could share a couple of our workflows. I I'll start with one of mine. Or let's see, there's how are you using what sort of things are you writing? And I suppose you can talk about your workflows as we go through.

Henrik de Gyor 4:00

Sure. Yeah. So so I've written nine books, and AI is is not writing books yet, realistically speaking, and not quality ones, the least, but they they are definitely helping in ideation and in cleanup. Whether I'm writing a book, an email, a LinkedIn post, a script, it's definitely helping for ideation, meaning the creation of the ideas. Instead of racking your brain about stuff, you can just plug in a few ideas into AI and it'll give you thousands of ideas, and then you can throw away 999 of them, and then that one kernel that it gives you, you can continue on that. So it doesn't replace you, it augments you and it augments our work, which is super helpful. So it accelerates it instead of like pondering way too long on things and drafting and redrafting and cleaning up. And uh eventually, as writers, where if you write enough, you're blind to your own edits, right? So, so you know, word or or whatever tool you're happening to use it is helpful to a degree for spelling and grammar, all the fun things, but not in the net in the sense of telling a story or or anything like that. It it it if you're if you're down a rabbit hole and you don't realize it, there's very little you can do about it aside from take time away from it or have editors. And one of the editors can be AI, right? To do that, so I I have them looking at everything I write on a regular basis. And I I I use one tool called Grammarly, and it it's it's correct, it gives me stats every week of of how I did. Is so far in the past 10 years, it evaluated 12 million words. Yeah, 12 million words. And and it does it it also tells me like the confidence rating, the the the mood, the the all the different attributes about the writing that I have been doing that week, of whether it's increasing, decreasing by X percent, interestingly enough, what the common errors that I'm show seeing, or what the AI is seeing. And so so if I had to focus on punctuation or whatever, or or you know, whatever the issue is, it it'll say, oh, you you've had a lot of errors around this. This is something I need to focus on. Now I'm more conscious about it. And and I can focus on on bettering that if necessary. Super helpful in ideation, in correction, in redrafting or or drafting the first draft, not necessarily the final draft, because you know that there's there's a variety of things that you want to correct in AI and check, right? So it's still hallucinating on occasion for a variety of different reasons where you you you you don't take something straight out of AI and plug it in. That's that's the lazy way to do it, and and it it's it's really helpful when you use it right.

Five-Step Transcript To Script Workflow

Alex Wilson-Campbell 6:54

On your point about Grammarly, and I think there are there's a certain group of people whose perception of AI is that it makes you lazy or makes you kind of rely upon it heavily to the point that you almost become numb to but that point about Grammarly that you just made, I think the Grammarly has actually helped me to improve my punctuation, it's helped me to improve you know how I construct sentences because it gives you those pointers that you mentioned and it gives you the sort of steer that I guess uh if you had an English teacher next to you would would would do sort of thing. Yeah, I like that example. I use Grammarly actually quite quite a lot. And the the first actual workflow I want to share with you is is related to writing podcast episodes. So from transcripts, and you and I, Henry, we we transcribe quite a few of our conversations. We start off with a I'll just talk you through the I guess the the steps, the workflow that I use in order to produce even something like this episode, for example. So we we've started this with a conversation. So we we planned it, we record you know uh calls, we have discussions ahead of time about what the episode will be about. We we do our research in the background as well. We always make sure there is a transcript, we always make sure. I think one of the things that we do, because obviously we are consenting to having this, you know, the ideation or at least the the capture of the information, we're consenting to that. So whoever, if you are doing it yourself, you make sure you you know you get the obviously the consent of the people who are part of the process, and we aim to stop relying on on memory and just scattered notes because again, when when you're capturing a real conversation with something like otter, for example, I've noticed the quality of otter has improved in terms of how it recognizes words, and the more and more you use it, I think it it actually begins to understand even the dialects and the different accents, and it makes it a lot clearer what what it is, you know, you're what's what's been spoken about. So you and I, Henry, we've aimed to stop relying on memory and notes here and there. I spoke to you about the the the I was looking in my cupboard just the other day, and uh I had like uh reams and reams of paper that I'd used to take notes previously on some of the meetings that you and I had. So Otter was a sort of a real game changer for me. So that's step one for me. Step two is really dropping the transcript. You know, you normally after after we finish a meeting, you'd send me that transcript, we'd share the transcript that we'd you know, we'd collaborate on that transcript, paste that transcript into chat GPT. And what I'm tending to do now is I'm I'm gathering all of the transcripts into one into one document as opposed to having different documents, or some people could could even put the transcript into the instructions. Uh there's an instructions area of chat GPT, because that then becomes like a knowledge bank of all the conversations that you've had that relate, for example, to you and I doing the podcasting. So in this instance, I asked chat, we had a fairly structured conversation on the themes that we wanted to talk about with these particular episodes. And that meant I was able to ask AI to pull out the main themes and the main topics and categorize each topic. And the aim is to see the shape of the episode without really reading everything. And then step three for me was asking for a simple episode outline. So for this one on writing and previous ones that we've had about meetings, for example, the the the uh the AI, I prompted the AI to put together a 15 to 20 minute episode outline based on the transcript. And you can tweak that. It can be an hour if you want, it can be a 20-minute episode. So you can use the AI to prompt in that way and ask it to make clear sections of the conversation. So an introduction, the pain points, stories behind it, your story, Henrik, my story, the workflows, the takeaways, and the structures as well. And step four for me was turn the outline into bulleted points. So again, the introduction, all these different sections of the podcast, with the aim of having something to talk about, which gave us a bit of structure for this conversation and other conversations. And then step five, edit it. Because as you were saying, Henrik, AI is by no means perfect. It can hallucinate, it can make mistakes, it needs training. It's not, it's not, it's not just you could that you can just shove it it shove the uh the transcript into AI and just say, go off and write a transcript, go off and write a sort of a script outline. You have to train it as you're going along. So yeah, editing is important, asking it to simplify the language in a way that's in line with how how we talk, Henrik, and all that sort of thing. And the aim is to make it do all the heavy lifting, really. And that's that's really my process, the the five-step process that I go through when it comes to just writing a script as we are today.

Henrik de Gyor 12:08

And yeah, that's awesome. Yeah, yeah.

Prompt Specificity And Tool Choice

Alex Wilson-Campbell 12:11

And you know, it's something that we're still refining it, aren't we, Henry? We're still refining this. Well, I I am anyway, in terms of refining the process of writing even this podcast podcasting script, because I myself, I'm used to solo podcasting, so it's so podcasting with somebody else is something quite new. So even it when you were asking the when I was asking the AI, I was saying, you know, pull out pieces of the conversation that relate to, for example, writing that I spoke about, or put out elements of the conversation that on writing that Henrik spoke about. So yeah, Henrik, it was all about me trying to get the get a balance. I think we're getting there, aren't we? I think.

Creative Uses And Book Ideas

Henrik de Gyor 12:52

Yeah, totally. Yeah. And I I I would only add to the training when you're trading an AI around what you're wanting to output on it on a consistent basis, like from transcripts of discussions to your point, all the way to talking points for for conversation like this, for podcast notes, for example, of what we're going to talk about. Make sure you add the specificity of what you're looking for and the context, right? Because it unless you create a project or you create a GPT in in Chat GPT or pick your favorite tool in question, you it needs the context repeatedly, especially if you change agents from from one tool to another, because it you can use there's thousands of AI tools, right? Whether it's for writing or or or or ideation or image creation or video creation or of audio editing, video editing, you name it, whatever it you want it to do, each one needs context to one degree or another, and and slightly vary in context of what's applicable to what you're trying to output, right? So it's not creating it in a vacuum because it'll create you something in a vacuum and it won't be what you want. So if you if you give the specificity, if you give the the specifics of what you're asking for, and you give it the context, and then you you explain what you want in that content, then it helps you a lot. I've noticed that ChatGPT has strengths and weaknesses to your point. And then there's there's a number of other tools where there's perplexity, there's Claude, there's uh there's Gemini, there, there's Copilot, of course. It's so it's really pick the tool that's you're comfortable with, but make sure you add that specificity in the context for that content that you're desiring. And the more specific you are, and sometimes it it it it will vary based on the output that you want, it'll help, particularly for writing, since that's what we're covering today. For for writing, because it's it it has all the rules of pick your language, English, right? In this case, instead of referencing a dictionary or thesaurus, which you know, the English majors of of the world, they they do that and they they're like religious, like, oh, I have my you know, fill in the blank branded dictionary and thesaurus right here, and I reference it all the time. It's like mm-hmm, that's great. And and we know how slow that is. Yep, of course. To your point. So back to the MIT study is like if you want to be more efficient and faster in what you're doing, it's it's is it gonna give you more options? Yeah. Can you filter through those options faster? So it's a choice to use AI, and it's a choice to use it efficiently.

Alex Wilson-Campbell 15:33

And it's I mean, I was skeptical at the beginning, but I think yeah, it's like you said, it's a choice, but I think it's a it's something that it's becoming a prerequisite in many areas, I think. And you really have to think about how you can take advantage of tools that are available to make you more efficient, make you know, make you stand out in some way. So it's becoming very quickly a prerequisite to use AI in some form or another. And I I think the way you use AI, Henrik, really I mean, writing a book for me is is is like a an undertaking that I've always thought of thought of. I've started to make notes, I've started to put outlines for books together, and this was pre pre-AI, and I kind of just got so inundated and so so consumed that I kind of the ideas ran out, and then the the it just didn't work for me. But you're using AI for writing books as well.

Henrik de Gyor 16:39

So my last few books I did use the early versions of pre pre-Chat GPT for ideation and for outlines only. They were not powerful enough to write chapters nowadays. If you want to create a sci-fi book, you could probably write it with Chat GPT. The one example uh that I heard recently was grandparents, instead of reading the same bedtime story for the billionth time to your you know under 10-year-old, you pull out Chat GPT on your tablet phone, whatever, and you have their imagination of that young boy or girl create the story, and then you have it type typed out for them, or you speak to it, and then it'll generate that story for you. And then you monitor it as the adults, right, in the in the room, because you know it's it's still an AI, and and you can filter it to whatever you desire, right? And and of course you're writing for that said aged child, right? But then it's it there is no end to that story, right? So it's very imaginative, it's extremely creative on some of the professional versions of of the tools. You have a slider of how much you want it to hallucinate on purpose, right? So so that it's like I think it was Amazon. There, there's a slider where you can go to like 0.01% as hallucination or 100%. So if you want full sci-fi, full bore, you can do that and and it'll go hog wild on you on whatever you can imagine.

Spotting AI Hallucinations

Alex Wilson-Campbell 18:16

On that point of hallucinations, uh, there may be some people out there who don't necessarily know what that is, hallucinations of AI. What is there an example of what to look out for in terms of well detecting that your AI is hallucinating?

Henrik de Gyor 18:32

Well, if you know something that's factual and then it start adding non-facts in there, like one example is is I was I was looking up for various events, I was looking up locations with coffee shops, for example, like a list of coffee shops in my area, and and it created some, a list of them. And I was like, I I the last few they were didn't exist. There were places that didn't exist. And I I went to Google Maps and uh checked it. It's like, no, these places don't exist. It's like there's no address. That's the first clue. But it it'll create things that don't exist, it'll create papers that don't exist, it'll create new papers, so research papers and things like that. And granted, you know, the the reason why AI is is increasing so fast is because there are so many research papers coming out. It if you look at the exponential, it's like literally a curve that's that just skyrockets because there are so many research papers written about AI. It used to be like, you know, dozens that you know, only PhDs and and developers. We're neither neither of us are or that. But anyways, uh, and neither of us are writing papers about that. But but but it's it's like thousands per month are being written. So so that's why we're seeing an exponential growth in the past hand handful of years. Less than literally like it was 2023. Um that when it it started exploding, and it exploded in popularity faster than any other social media tool or any other form of media ever in the number of people using it. So there's over 100 million people at the last time I checked, using it regularly.

Drafting Emails And Slides Safely

Alex Wilson-Campbell 20:16

And I guess going back to the the hallucinations bit would it's quite useful if you're writing a fiction book, I guess. But totally. But for you, what what is it do you have like a set process when it comes to writing your your books or whatever it is that you're writing using AI as a sort of like a process that you have in your mind?

Henrik de Gyor 20:39

Well, so not necessarily for books because books are I I've parked most of the books temporarily. I've already written nine, so so until until something really, really sparks my my interest and and piques my curiosity, I'm I'm gonna park it for now. But as far as writing to your point and and using AI, um a lot of emails, a lot of drafts are started via ChatGPT or or or or a variety of different tools. I'll it depends on what I'm trying to write. Like sometimes it's it's a it's a playbook, sometimes it's a a slide deck. So if I'm trying to do a slide deck, I'll go to Manus. And uh, Manaus is pretty amazing. I can output it to anything I want. There's several other tools that'll do that. Gamma will do that as well. And it it it really depends. So it it'll it'll generate me one that's branded if it needs to be branded for a client. It'll generate me one that's branded for me. It'll generate one that's not branded, just straight out of the box, as long as you supply that branding, right? That starting point. And then it's also like, well, how many, how many slides do you want? Or what what what do you want in each slide? And and then you can curtail the text first, the text content of of this of each slide, slide one through, fill in the blank. And then what graphics do you want in there? Do you want charts? Do you want imagery? Do you want icon iconography, iconography? Whatever that looks like. So I've been doing a lot of that recently. And then there's also email, email drafts, like, oh, I need to email draft something. Like, let me check it. Like here, here's the here's the the original email. I might even remove the person's name and email address. So I'm not sharing the PII. So no personal identifiable information is being put into the AI. So so there's no connection to that, to that individual or that company. And I might be putting the first name in there, right? I'll put that in there if that matters, if that's part of the context. And have you have a draft a response that's you know positive about whatever response in that sense. And then based on that, I'll copy it and put it into my an email and draft it, fix the few nuances there, unless it's it's terrible. And then I just have to add more context or more specificity in into the AI. And then I literally I just copy it and put it into the email or into a PowerPoint or whatever the output is, right? Whether it's writing an email, writing a uh a message of some sort, writing a uh a playbook, all the different things that you can write, right? Sometimes I'll write reports, and yes, uh the reports will be pumped into there and they'll be checked regularly because you know there's gonna be X number of versions of that that reports, right, until you you finalize that report for a variety of reasons. And you can check it and make sure that it's it's doing what it's supposed to be doing. And it's uh obviously you have to check it that there's that's that's you're still the human in the loop. And you should be worried if there's no human in the loop. That's our purpose is to be the human in the loop. It's the re you're the requester as the human in the loop, and you're the receiver of that output. And if you're not the human in the loop, checking to make sure A, it's still working, B, it's it's outputting what you're supposed to be doing, why are you there?

Repurposing Transcripts Into Many Formats

Alex Wilson-Campbell 24:03

Yeah, and I think that's that's a key thing, isn't it, with the operators that there's this term going around AI operator. Uh and I think that's that's how I'm trying to to see myself now as somebody who is operating the AI for the for whatever end it may be, for a positive outcome for myself. Using it, you know, I'm not I'm by no means am I using it perfectly, but I'm I'm using it because I want to try to to improve, to, to, to stay relevant, to to stay up to date. And it things are changing so quickly. It's not it's not possible to to completely be on the bleeding edge, is it? Because things are things are changing so so rapidly, but at least the very least, just trying to immerse myself in as much as possible without going down rabbit holes and without without drowning is the key, just a bit, bit by bit each day, you know.

Henrik de Gyor 24:57

Really?

Always Be Capturing And Farewell

Alex Wilson-Campbell 24:57

But for me, I think what what I've decided is that as a somebody who creates content quite a lot, I think there's something you said on the last show about capturation, it you know, capturing your IP, capturing the you know, the ideas of other people with their consent, always taking sort of electronic notes so that can be used and then sort of repurposed into other content. So repurposing, you can repurpose anything, can't you? You can repurpose even images. You can not that we're gonna talk about that today, but repurposing transcripts is a big thing for me. Even conversations that I've had on the remote work-life podcast from from years ago, transcribing them, you know, because YouTube itself, if you upload to YouTube, it will transcribe. I mean, it's YouTube is not the best when it comes to transcriptions, but it makes it possible to transcribe conversations, it makes it possible to transcribe monologues, and you shouldn't waste that information because you can change it into transform it into multiple pieces of content as I do. So I I would choose after doing I've done about 250 podcasts. Henry, you've probably done a lot more than that. That's like a mammoth amount of data. It is a mammoth amount of information. It shouldn't be wasted in terms of turning that into something else. You could use the AI, ask the AI to pull out the main ideas, two or three clear ideas, let AI label the ideas in a simple plain language. And then another step would be to draft specific content types. I've repurposed podcasts into LinkedIn posts, into articles or newsletter articles or blog posts, show notes, so many different things. But again, as you said, it's important to edit and make sure that it remains as factual as possible and relevant, as as little hallucination as possible. So, yeah, transcripts and repurposing is something that's been really, really, really big. And I think that's all I want to share today in terms of how we're using it to write. But what I'd say is just, you know, as as Henrik and I have said it in pre in a previous episode, is to capture capture your notes, even if you're you know setting yourself a session where you're just doing a brainstorm on a certain activity or a certain idea, whether you're trying to write a blog or a LinkedIn post or an email, you know, you can capture some of your ideas about what you want in that email, in that blog post, by using something like Otta or the thousands of other sort of transcription software that are out there. So pick and pick one, one of those recordings, get the transcript and use AI to ask it for a simple outline of a I don't know, a LinkedIn post. Ask it to do that, but give it as Henrik was saying, give it context and then turn it into a LinkedIn post. If you if you're trying to do LinkedIn, and it can turn it into different versions, it can turn it, turn it into like a list style LinkedIn post or a contrarian take or or you know an informational post. So it there's so many different things that it can do. It can duplicate the the you know your your thoughts, uh, it can really help with with with content production at scale. So that's what I'd suggest doing. But uh Henrik, I just really wanted to say thank you for joining me today for this episode. Of course, always be capturing. That's the that's the phrase.

Henrik de Gyor 28:32

Always be capturing.

Alex Wilson-Campbell 28:33

That's on my that's something I learned from you this this this year, so I'll be doing that as well. But I think we have been doing that, but it's a matter of just sort of going through with it and and using everything at your disposal to to create content for our from our perspective. Thanks again for everybody listening. That was another episode of the Remote World Live podcast, and we'll see you on the on the next one.

Henrik de Gyor 28:52

Thank you.