1033. Last year, Dragonfly Editorial had a "no AI" policy, but since then, they've been experimenting with the technology, and the policy has become more nuanced. President Samantha Enslen joins us to talk about what changed, what's working and what isn't, concerns, and how her employees feel about it.
1033. Last year, Dragonfly Editorial had a "no AI" policy, but since then, they've been experimenting with the technology, and the policy has become more nuanced. President Samantha Enslen joins us to talk about what changed, what's working and what isn't, concerns, and how her employees feel about it.
đź”— Share your familect recording in a WhatsApp chat.
đź”—Â Watch my LinkedIn Learning writing courses.
đź”—Â Subscribe to the newsletter.
đź”— Take our advertising survey.Â
đź”— Get the edited transcript.
đź”—Â Get Grammar Girl books.Â
đź”— Join Grammarpalooza. Get ad-free and bonus episodes at Apple Podcasts or Subtext. Learn more about the difference.Â
| HOST: Mignon Fogarty
| VOICEMAIL: 833-214-GIRL (833-214-4475).
| Grammar Girl is part of the Quick and Dirty Tips podcast network.
| Theme music by Catherine Rannus.
| Grammar Girl Social Media Links: YouTube. TikTok. Facebook.Threads. Instagram. LinkedIn. Mastodon.
LIGHTLY EDITED TRANSCRIPT
MIGNON: Hey, Grammar Girl here. I'm Mignon Fogarty, and today I'm here with Samantha Enslen from Dragonfly Editorial, and you may remember her either from a few months ago when she was on talking about how she and her husband recently took over their hometown newspaper.
But even more likely, because she's been a contributor to the Grammar Girl podcast for so long, I can't even remember how long, definitely more than a decade. Recently I attended one of the best AI webinars I've ever attended that Samantha put on for her clients at Dragonfly Editorial.
It was like the best client transparency event I've ever been to, and I just wanted all of you today to hear about it. Samantha Enslen, welcome back to the Grammar Girl Podcast.
SAM: Thank you so much, Mignon. And that is a wonderful introduction. It's so funny to me that you say that was one of the best AI webinars you ever went to because I specifically, when I when, we were thinking of it, I specifically was like, “We don't know what best practices in using AI are, and we're just we're trying stuff. So let's just get together with our customers and tell them what we're trying, and how it's working, and what we're playing with, and what we see, might be coming in the future. And also be honest about concerns that we still have.” So I'm glad that it resonated, and you weren't like, “Why are these people telling us about stuff they don't even know?”
MIGNON: No, it was just incredibly clear. Here's what we're using it for. Here's what we aren't. Here's some problems. So a couple of things for background. So I know a few years ago to about two years ago, you're well, actually, let's back up first. What is Dragonfly Editorial?
SAM: Oh, sure. That's a great question. So Dragonfly is a creative agency. I started Dragonfly Editorial back in 2005, I think. And I was coming out of a corporate job and moving back to doing freelance editing. And at that time, I really had no grand visions other than I had a baby, I had another one on the way.
I was like, “I need to go back to doing freelance editing.” And, so I have a little bit more time, but the opposite happened. And once I started freelancing, I started getting more projects than I could do myself and bigger projects than I can do myself. So I started reaching out to my freelance friends to help, and, here we are 20 years later, and we now have a staff of 25.
We do writing, editing and design. Our focus is mainly on B2B customers. Although we also work with large nonprofits and some government entities and associations, like ACES, the Association for Copy Editing, where I got my start. So that's what we do. And, yeah, it's really a combination of all, like I said, all of the disciplines, writing, editing, and design and creating what we hope is cool stuff for our customers.
And often because it's in those B2B topics or B2B areas, it's explaining things that are not easy to explain, whether it's like in finance or engineering or IT, or science and medicine. That's our specialty, working with individuals and companies to help them get across things that are not easy to explain, and are often very niche topics, but helping them do that clearly and powerfully.
MIGNON: Yeah. And so I think that the invoices go through Dragonfly. So I'm technically a client of Dragonfly, but basically you write for the show.
SAM: Yes.
MIGNON: So that's why I was on the email where I got invited to this event that was really for your B2B clients. But I decided to come anyway, just because it sounded so interesting. So one of the things you mentioned is that a couple of years ago, your policy was "No AI. We don't use it." And now you're experimenting with it. So what changed your mind?Â
SAM: Yeah. So what changed my mind? There was not one single thing that changed my mind, but rather observing the evolution of the technologies. Like I said, even over the course of just that year, for example, in early 2023, we felt it was important early on to establish what our ethical policy for the ethical use of AI was for Dragonfly.
At that time, we said things like, “We will never use AI to produce copy. We will never use an AI to edit our content. We will never use an AI to create images." Flat out. And talking about why we think that the importance of human creativity and human thought and the human, and even in B2B communications, bringing our human experience to the table.
Real people talking to other real people. To gather their ideas, and again, help share them in a clear way. So over the course of that year, I think we also started to, as the tools evolved, we started to see them as assistive tools to our writing and our editing and our design rather than things that are replacing us outright.
Or just saying, “Oh, yeah, human work sucks. Now we're just going to use AI instead.” So our policy now is more nuanced. So we say, for example, “We will never use AI to produce copy completely.” So what that means is if you call me yourself for Grammar Girl to write something for you on a certain topic, I'm not just going to put it in ChatGPT and get an answer and email it back to you. That's still cheating or plagiarism.
Same with our customers, we're not going to turn anything into them whole hog. They're not gonna give us an assignment, and we just type it in a search engine or an LLM and turn it back to them. Now in most cases that's not even possible because again they deal with such specific and niche topics, and nearly everything that we write at Dragonfly involves subject matter experts.
So we're talking to people to often learn about their original research that they've done, work that they have done with their customers in the field, their field experience, if you will, and their own thought leadership. What are their personal thoughts on changes in our world, trends, technologies, and that can't be gotten from an AI?
That's all personal information or what you might call company proprietary information. So our work as writers is to work with those folks and help take, distill, take their ideas, their research and distill those into a piece of writing. That said, what we have found is that we can use AI to help us with little pieces of that.
And I'll give an example from the work that I do for you writing blogs. There was one that I was writing a few months ago, and I don't even remember what the topic was, but in pot, but just in passing, I was going to talk about Homer's "Odyssey," reference that in some way, but I didn't need to go into it a lot.
So I just typed into Perplexity. That's the AI I use. Give me a 10-word description of "The Odyssey" and it gave me like a 10-word description of it. I'm like, “Perfect.” That's what people who might not know what this is need to know. So I could just add that little tidbit into my work. We find that we use it a lot for background research of our own, because part of the writing process is often you're interviewing a subject matter expert to understand their ideas and thoughts.
And then you're often also doing your own independent research to make sure that you fully understand the topic that you're filling in the gaps, or expanding on ideas, perhaps that the person you interview didn't have time to get to. And so we're able to also use the AI tools for that. So it's like a combination of search and like I said, maybe writing little snippets of things.
A lot of time, what I will do is, I'll ask one of the engines a question and have it, maybe it'll provide 2 or 3 sentences of something back, and I'll be like, “Okay, yeah, those are some good ideas that I want to incorporate into what I'm writing.”
So that'll give me a concept that I can work from and write it in my own words. And then I'll get to another point where I'm like, I need a little bit more information. And so I'll put a query back into the AI entity, and it'll give me a couple more sentences and I'm like, “Ooh, that is a good point of something I should include.”
And then I can go back and work in again. So it's almost like the rhythm that I am using the AIs for when I'm writing, it's almost like a call and response. It's like I query the AI, and it gives me something back. That queries my brain and then I can type something till I get stuck, and then I query the AI.
It gives me something and I'm like, “Ooh, a query for Sam.” And then I can write, it is becoming more of a collaborative process and we're really seeing it as a tool in our writing work right now, not a replacement for what we're doing, but just another tool.
MIGNON: I'm curious if your clients were asking about AI. Were they? And it could go either way. Were they wondering if you were using it to be more efficient, or were they concerned you were using it, or were they not asking about it at all?Â
SAM: Really people hadn't asked about it much at all. It was just something that I think that we felt like you said, in the name of transparency, that we wanted to initiate the conversation and say, “Hey, just so you guys know, here's what we're doing. Here's what we're trying out and if that's something, for some reason, you don't want us to do on your work, tell us. Roger that, we will not use any AI tools.”
I was talking about writing a minute ago and the same thing. Let me jump over to our editing side. Again, a year ago, we had said straight up, “We will not use AI to edit your or anyone's work whatsoever.” So now, our policy has changed a little bit. We won't use it to edit anything without careful review of what the AI outputs, by our own human editors.Â
And from that perspective, using an AI assistive tool for editing is not much different than what we already do with the suggestions we get back from spell check, you use something like Grammarly. I have Grammarly as a plug in for everything I'm typing, whether it's in my email or a Word document, and I'd say maybe a good 75% of the suggestions I take that it gives me, but there's another 25% that either aren't necessary, are goofy, change my meaning, or create an even bigger absurdity beyond goofiness.
So really, again, on the editing side, we're starting to see the AIs as assistive tools, not something that's replacing us. Just, saying, “Oh, we're not going to use an AI for editing” is like saying, “I'm turning off the spell check on my Microsoft Word because it's not human or real.” You wouldn't do that because you recognize that it's a tool that does help you.
Those little squiggly lines under words, you realize, “Oh, I did misspell that.” And that's helpful.
MIGNON: Yeah. At the transparency webinar, you had people, the people from each of your departments speak about what they were using it for in their department. And one of the things I took a lot of notes on was on EditGPT. Cause I've tried it too, as an alternative to Grammarly and just, one more, it's always good to have one more check, you know?
SAM: Yeah.
MIGNON: But you, the person who talked about having some technical problems using it too, so it's not quite there yet.
Can you talk about how you found it useful, but what the barriers have been too?
SAM: Yeah, I can give you a compare and contrast. For example, we talked a couple months ago about the newspaper that my husband bought recently. And that I help him with here in our hometown in Tipp City, Ohio. Well, when the stories come in for the newspaper, I would say they're, maybe anywhere from 250 to 800 words about a page of text.
They're short and sweet. So something like that is very easy to put into a EditGPT. That's it's EditGPT is a standalone app. You can also use it as a plug-in to ChatGPT. That's a whole conversation of how that works. But let's just say it's an editing tool. And for the newspaper, it's fantastic because it's catching lots of technical errors.
It's like a spell checker, but a little bit better. It catches inconsistencies and things that a spell check might not get. So we can pop it in there, get some results, check them, confirm we want to do it, and move on. Now on the Dragonfly side, our documents that we're editing are often 25 pages long, 50 pages long, 500 pages long.
And so those tools aren't quite equipped right now to handle documents that are that large effectively. Moreover, at Dragonfly, for some of the technical documents we work on, let's say a 500 page document might be split between 10 different editors, each editing 150 pages of that. That math probably doesn't add up, but you know what I'm getting at.
And so the logistics of, are we all going to put that into a LLM and each person get their own results. There's just some barriers to using it on large documents right now, both from the file management size and because on — the file management side — and also because my best understanding is that, most of the large language models right now just can't handle being dumped in a document of that size to process and analyze. That may change over time.
MIGNON: Yeah, that's right. And then the track changes are a really nice thing about it, but then it sounded like they weren't working for you when you were transferring things back to Word or something like that.
SAM: This is exactly, this is a huge problem, that we haven't figured out how to get past for the work that we're doing or that, not just our agency, but anybody who's a freelance or professional copy editor, 99% of the time when you're turning work back to your customers, you're going to want to do it with track changes showing.
And so that means in a Word document, new text is highlighted, removed, you can see all the removed text over on a bubble on the side. So it's very easy for a reviewer to see exactly what changes an editor made to your document. Problem is, so right now, just plain old ChatGPT doesn't create track changes.
When it, if you could put a document in there and say, please edit this page for me, it'll edit it, but it won't show you what the changes are. First of all, you have to use a plug in the editGPT to create track changes, but then when you try to copy and paste that back out into word. It's a whole thing. So yeah, you know what? What I imagine, Mignon, will happen in the not too far future is that these tools will exist as plug-ins it right into Microsoft Word or right into Google Docs or whatever you're working in. And then you'll just be able to do your work without having to go into without having to put your content in a separate platform and then bring it back into whatever you're working on.
MIGNON: Hmm.Â
SAM: I feel that's a barrier right now, again, particularly if you're working on really large documents. If you just need to edit a page at the time, no big deal.
MIGNON: Yeah. And I'll say that you still have all your human editors. This is just, plugging it in and finding, okay, there's this tool that will find 80% of the commas that are in the wrong place. And that just saves a lot of tedium of looking for commas. That's been my experience with it.
SAM: I think you're getting at an important distinction, which is that ultimately, I think these tools are often good for helping edit and spot the technical errors that are wrong in a document. Like you said, spelling, grammar, punctuation, typos, a word might be hyphenated in this part of the document, but it's not hyphenated in another.
What we have seen, at least so far, is it's not really great at helping with the meaning of a document. For example, a fair amount of the work that we do in, on the editing side might, is what you might call substantive editing, which means not just, don't just fix the commas and periods and spaces, but also take a look at this, whatever I've written and tell me, “Does this make sense?”
Is this getting the point across that I want to convey? Is this in a logical manner, written in a logical manner? Are there important points in this conversation that I might have left out? When customers come to us and say, “Hey, we've written this thing, but we're not quite sure it's right. We want you to do a substantive edit or a substantive review of this and give us feedback and help us make it better and stronger.”
What we have seen so far is the AIs aren't quite good at doing that because they don't know what the important point is. They might, again, it's through that human conversation with our customers of finding out what's the meat of this piece?
What's the hook? What's the critical conversation we're trying to have here, and then we can help them edit it, rewrite it, revise it accordingly when an AI just isn't going to know that.
MIGNON: And here's another thing. So I don't think about visual AI essentially at all, you make graphics and things like that for your clients. And it was really an also an interesting discussion about how you are and aren't using it for that aspect of your work.
SAM: Yeah, so once again, a year ago, we had just said, “Nope, we're not. Our policy right now, we're not using AI to create any images.” And now our policy is, I think we say, “Okay, we're not going to use it to create entire images or designs.” And I'm sure you have seen on the internet all kinds of examples of really goofy things that an AI has created.
If you say, “Show me images of 10 mammals.” And it's outputting mammals with made up names, and some of them have six legs, and some of them are like combinations of cats and turtles. It doesn't necessarily do a very great job of image creation, especially when AIs are trying to create videos. There's lots of issues with fingers missing and mangled heads being on other bodies, all kinds of unpleasant stuff.Â
So what we have been using it for again, is more of an assistive tool in the work that we're already doing. For example, one of the things that AIs can be good at is fixing pictures, photographs that is.
So you might be able to give an AI, say like a 1970s era photograph that is blurry, it's pixelated, it's possibly faded over time, and you can input that in to one of the visual AIs and, essentially say enhance and clean up this photo and it can do whatever magic it does with all the pixels and all of a sudden you have, a really beautiful picture that's suitable for publication.
We can. We also use an AI for helping us brainstorm sometimes. Okay, it does sound weird to say that we're using like a robotic brain to help us brainstorm, but I guess that's the way to put it. One of our designers, for example, was working on an image for one of our customers that had to do with cargo on a shipping container.
So she asked an AI to generate a picture of that, and it gave some rough ideas of what the design could look like. So she's okay, much as I said with my writing, she's okay, that gives me something to work from. I can visualize now what this graphic might look like, and she could then render it through her own artistry in a more professional and appropriate looking way for this particular customer, maybe using their brand colors.
Designing it in a way that made sense. Oh, another example that Casey said that they do sometimes, the icons that there are on websites or that people use to represent little things like a, like a light bulb might be used to represent an idea or something. So sometimes we get asked to create icons for something.
Let's say it was for an idea, a light bulb has already been used as that imagery 10,000 times and 10,000. So you're like “ChatGPT, give me some ideas for an icon that represents a new idea.” So it might make some suggestions, and then maybe there's something that resonates for the designer, “Ooh, that would be a good idea. Maybe I could use that for my icon” and then again that informs their design. So again, we're starting to see these as back and forth tools as little pieces that can help us in our creative process, rather than eliminating our creative process and just turning it over to a robot, if you will.
MIGNON: Yeah. It makes me curious how do your employees feel about it? Are they all really enthusiastic or has there been some concern?
SAM: I think it's some of everything. And that, what it is that Dragonfly is probably no different than any other representative sample of the population. You have some people who are like, this is the coolest thing ever. And they have seven different AIs that they're using, some for personal life, some for work that they're trying out and are really excited about it.
Other people are like, “I'm very happy doing things the way I do things. I'm very skeptical about this”, and then everything in between. But as a company, we're doing the best we can to educate our employees. We've done trainings with Erin Servais, who has a class called, AI for Editors.
The designers, there's a lot of training available. One thing that people might not know, our designers use the Adobe Creative Suite, mainly for their work, sometimes Canva also. And Adobe actually has a lot of built in training functionality already, so they're not having to go out and use a separate platform necessarily.
A lot of it Adobe helps them with. So their point being, we're giving them training for Adobe, the writers, there's all kinds of classes on writing for AI that are out there, we're training the writers in that. And what we have told everybody this past year is that, We don't know what the future perfect tool is going to be or the right tool is going to be.
And moreover, all these technologies are changing so rapidly. So this is just a year to experiment and try things. And we're going to be seeing what works. And as we see that over time, we may be able to refine more specifically "Here's the tool that we're using at Dragonfly for this function in this way." And then it'll become like standard operating procedure right now. It's still, I would say, in more of an experimentation and trying and seeing what works phase.
MIGNON: Yeah, talking about images and Adobe and everything reminded me of a story I did hear about images, and it'll be a great transition to talk about bias, because I saw this scenario where a woman had sent in a square headshot for a conference bio, and the organization needed a little bit more of a vertical photo.
And so they ran it through one of these AIs that has this great "expand the image" thing, where it just fills in what's missing, like around the edges that is supposed to be very neutral. And she noticed that it unbuttoned an additional button of her shirt, and her bra was showing a little bit.
And she was like, “Was my bra showing in my headshot? Like I never noticed that before.” And she went back and looked and it wasn't, it was the AI had done that. It had essentially sexualized her just a tiny bit. But that is something that you do hear about, from AI is that it has these biases, these built in things, because it was trained on all of the internet and all of the internet has a lot of terrible ideas.
How are you working with that, just ensuring that those really subtle biases don't slip in when you're using it for your business.
SAM: I think this is a great question. And so far I've been talking in a very positive way about how we've been using AI and how we're seeing it as a tool to help us with our work. But there is a negative side. I'm going to say lots of negative sides. That's only one. Yeah, the fact that you can get on YouTube and see all kinds of quote, “sexy AI ladies dancing”.
I'm like, why is that the first, is that the, did that really have to be the first thing that we go to as soon as we have a new technology is. Making videos of fake sexy ladies?
MIGNON: Seems predictable.
SAM: Indeed. All too predictable. But there are many other issues going on with AI right now that are cause for concern. One is that it's nice for me to say, “Oh, we're not using AI tools to replace our creatives. We think that the human input is still the most important ingredient in the equation”, but lots of other people aren't.
They're like, “Oh, if I can just type such and such into ChatGPT and have it spit out a blog for me every day on different topics, I'm going to do that from now on.” I guarantee that there's, same thing going on with images and photography. How is this going to affect, how's this going to affect the careers and livelihood of writers, of editors, of people?
Photographers, videographers, designers, everybody in the creative professions. It's so interesting to me because my kids, the girls right now are, let's see, 15 and 20. And so you would think that they would be more naive about this kind of stuff, and they're less naive and like super anti-AI.
Anything that they see that they think is AI, they simply say that “Oh, that's AI generated.” They like, have a real distaste for it. And I think they crave authentic content, real books that they can hold on to real things created by real people. We talk sometimes about, again, that pinpoint in time, let's say, March, 2023, when the GPTs really started hitting the scene. That's going to be a marker, I think, for the future when before that's going to be the only time we know that things were actually really created by a human.
After that everything created from henceforth, there's always going to be a question, did somebody really write that novel, or did they work with an AI that kind of wrote most of it based on their ideas, and then they just touched up a few things here and there. And I don't know if I can even pinpoint why, but I think there's something very sad about that and very soulless about that.
And all issues of art and soul aside, it also comes down to just a plain old question of theft. What you could call theft in many cases, there's all you need to do is Google. There's multiple lawsuits that are open right now against all the major AI engines, whether it's OpenAI or Anthropic or what have you, lawsuits from artists, YouTube creators, computer programmers, record labels, novelists, news organizations who are essentially saying, “You have taken our content without our consent, without compensation, and you're using it to train your product that then you're selling on the market for profit. And, we're not, our creative works are being used without our permission for your benefit.” Essentially, plagiarism writ large on a grand scale. So again, we don't know how all that's going to play out over time.
But I think it's a very valid concern. It's even going on, I think the most recent actor, Robert Downey Jr. just came out and said, “All my contracts henceforth say, my image, my voice cannot be replicated by an AI and used in future movie productions without my consent. Just because I make a movie for you, Marvel Studios or whomever, doesn't mean that I am giving away my virtual soul my identity for you to now just do whatever you want.”Â
MIGNON: Yeah. Thank you. You've articulated that so well. I am. I feel the same way. Like I feel like we, I have great concerns about the future of workers, especially creative workers, writers, editors. That's why I've been spending so much time learning about it and reading about it and trying it because I feel like we have to know what this thing is that's coming for us.
And also understand if it is a useful assistive tool, but I also have all the same concerns about the plagiarism and the bias and all the problems with it. But I do think it's essentially unavoidable at this point, for people, working professionally in these fields.
SAM: I do think it's unavoidable at this point. And I explained it to my staff. Okay, this is dating myself somewhat, but when I started editing, we still did it on pencil and paper and marked up, comments literally with a red pencil, and then we handed them to somebody who was a quote desktop publisher who input our changes into the system, and then they came back to us and proofed well, then at some point they had editors started who editors … ooh, got a computer themselves.
And they invented track changes. Microsoft had track changes, and then all of a sudden we could start doing it. I remember when that was a huge technological shift, and we had editors at that time who were like, no, that's not how I do my work. I can't work on a computer. And there was a period of time when we would train people to work on Microsoft Word and train them how to use track changes. And then there came a period of time when it's like, “Sorry, if you don't know how to do that, I can't hire you at this point.” So I do think that, even though we don't know right now where all these AI tools are going to take us and what form they will take.
I do think that there's pretty good evidence that they are going to become part of our day to day work. In the same way that we couldn't imagine having day to day life without our phones now, or without having Microsoft Word or Google Docs or email, those were all weird changes, when they occurred at the time. Now we've just got used to them and can't imagine day-to-day work or life without them. So I do imagine that's where we're going to get before long. And yeah, nobody in any workplace is going to be able to say, "Not for me."
MIGNON: For a while I was trying to avoid AI search, but then, I'd be like, “Oh, I'll go to Google instead of Perplexity.” But then now we have the Google AI summaries, it's unavoidable in that too. There are smaller search engines that still don't have it,simple and I want to clarify one thing.
I don't think it's unavoidable for everyone. I think, novelists and poets and painters, it's avoidable for people like that. And it, we want the humanity in those works, but I think in the business context, it's, you're right. It's going to, and it's going to be built into the major tools that we all use, Sam, thanks so much for joining us today, sharing your knowledge and putting on that wonderful transparency event for your clients.
That was just, it was wonderful. And I hope more companies and people do that continue to do that. Because I think it's a necessary educational event that really helps everyone.
SAM: You're welcome. I'm so glad you found it. so glad you found it valuable.
MIGNON: And where can people find you?
SAM: Oh, you can go to our website, which is just dragonflyeditorial.com. or you can follow us on X @dragonflyedit.
MIGNON: Great. And now for our Grammarpaloozians, we're going to have a bonus segment where Sam will talk about her book recommendations. And I think I want to talk a little bit about what you think the future holds for AI and what you think you'll be doing in hiring in the next couple of years. But for the rest of you, thanks for joining us.