AI As Your Writing Partner

[MUSIC PLAYING]

Sara Faradji: When we're teaching English classes, we teach developing a good counterargument and a rebuttal. And I think those are good to keep in mind when working with AI tools. And you may find that some of the things that's pushing back on really as relevant, but it's still good to at least see some of that information and see it kind of leaning into the more negative or questioning tone, rather than just validating everything you say. 

Jill Finlayson: Welcome to The Future of Work podcast with Berkeley Extension and EDGE in Tech at the University of California, focused on expanding diversity and gender equity in tech. EDGE in Tech is part of the Innovation Hub at CITRIS, the Center for IT Research in the Interest of Society and the Banatao Institute. UC Berkeley Extension is the continuing education arm of the University of California at Berkeley. 

AI can brainstorm at lightning speed, clean up clunky sentences, and help you get unstuck when the blank page stares back. But are you losing your own brand voice and your critical thinking skills by offloading tasks to AI? Or are you sharpening your writing and accelerating your productivity with AI assistance? 

In this episode, we explore using AI as your writer and editor companion. How do you decide when AI is used as a creative copilot and when human judgment is needed most? How are communicators in jobs like product development, sales, marketing, HR, using AI to work faster, without-- and this is key-- losing authenticity, accuracy, and trust? 

Let's talk about the art of augmentation, because the Future of writing in the workplace is not human or AI. It's human with AI. To discuss this topic, we're delighted to welcome technical communications strategist Dr. Sara Faraji. Sara believes that powerful stories drive technical innovation, but critical inquiry is what turns innovation into real-world impact. She brings this lens to her technical writing work in cybersecurity by helping research and development leaders communicate technical information in ways that resonate meaningfully with business leaders and customers. 

As a faculty fellow at the University of Maryland and a writing instructor at UC Berkeley Extension, Sara empowers students to build a rhetorical toolkit for career success in a shifting job market. Welcome, Sara. 

Sara Faradji: Hi. Thank you so much for having me, Jill. 

JJill Finlayson: This is going to be fun. And I want to start out with this rhetorical toolkit because it sounds very fancy. What led you to develop it, and why is it needed today? 

Sara Faradji: Yeah, so what I love about being in the English department at the University of Maryland and teaching English is we really ground a lot of our principles on age-old rhetorical concepts. These go all the way back to Aristotle and different philosophers. And I find that these rhetorical skills are tried and true. No matter what technological innovations come, these rhetorical strategies are really great. 

And I think in the age of AI, these kinds of tools are even more valuable. And I kind of have a three-part toolkit that I really recommend for students or for anyone who is going to use AI. And the first one is understanding the question of stasis theory. 
And so what this is is it's looking at writing in a way that is question-and-inquiry-oriented. In thinking about this with AI, it's not just, here's something and give me an output. Here's a paper or a prompt. Now give me an full paper or something like that. 

It's more about, here is a prompt, and now I want you to answer some things for me. What are the facts here? That's called conjecture. And then what kind of thing is this? How do we define it? That's called the stasis of definition. 

And then look at this. Is this good? Is this bad? That's about the stasis of quality. And then there's also a stasis of policy, what should be done. 

So in practice, when you're looking at your prompt, you ask yourself, what do you actually need? If you're not certain about certain key facts, ask, what are the facts of this that are known based on the prompts, and what do we need to find out? That way, it can help prevent you from getting some answers that are hallucinated or not true. 

And when you look at a cover letter or something, you say, is this a good example of the cover letter genre? What are some things I can improve upon? What are the flaws in it? So really looking inward on these questions. 

And the second thing I recommend in the toolkit is looking at rhetorical appeals. So this is looking at-- ethos is establishing a sense of self. So when you first give a prompt to AI, you can say something like, I am writing for this audience. Here's a little bit about me as well. I am a technical writer. You're giving it your persona. You're establishing your sense of self. 

And also, pathos is looking at emotions or what the audience cares about. So establishing who is your audience, what are their values, what do they care about. And things like this are really going to help it to be able to understand how to produce a quality output. 

And then a third part of this toolkit is understanding this issue of exigence, the so-what, the who-cares factor. I think this is something that is missing a lot in first-turn writing when you ask the AI to give you an output. 

So what I like to do is look at, I am writing this for a specific situation. Here is what prompted me to write it, thinking a little bit about what was the situation, what were you being asked to do, why, and really getting into what is the problem that you are seeking to solve, because if we don't give it that information, then it's going to have to assume that. 

So being able to give it these kinds of constraints-- so this is the toolkit that I recommend. And I think if you learn about and understand situations through this point of view, and use this toolkit, that I think you can be much better at interacting with AI. 

Jill Finlayson: What I love about this is you're starting out with the English department. To me, this sounds like journalism 101. And when you think about jobs and technology, you often don't think about how important the humanities are to really communicating the technological innovation. So I love that lens that you're bringing to it. 

Sara Faradji: Yes, absolutely. I think that with journalism, where you are doing creative writing, all of these different ways that you are engaging with language, I like to compare it a lot to working with code. And in some ways, people who are working with language, human languages, and those who are working with code, we're all working with code in some ways. 

And I think that's what makes these AI tools so interesting is that we're able to get code outputs and generate them much faster, but we still have to analyze the quality, ask it questions, verify, and do a lot of critical thinking still. All of these skills are still so valuable. And we're redefining what that looks like and how we continue to shape and sharpen our skills with AI. 

Jill Finlayson: So let's make you go back to the early days. Do you recall the first time you experimented with AI as a writing partner, and how did it go? What mistakes did you make? 

Sara Faradji: Yes, I can recall the very first prompts I remember using when I was working with AI tools was, first, I started with what I knew, in the sense that I've used different tools that are for spell check, and other than-- I use search. In some ways, these are helping us to refine what we've written. 

But if I'm just taking a paper and then uploading it and saying, fix the typos, or, edit this, and not giving a lot of structural guidelines or information, it might be able to identify issues with grammar, but that's not really the hard part in some ways. There's a lot that's still going to get missed if you do things that way. 

But is it aligned to your style guide? Does it your specific nuances and things like that? And if it doesn't know that and you're assuming that it knows that, then you might have to get into quite a long conversation and get frustrated with the output. 

Another issue I had when I was first prompting was I would just maybe upload a lot of different docs that provided some factual information. Maybe I'll use the source code. Maybe I will provide just a general one-pager of this, all these different types of things, and then say, here, make a knowledge article out of this material. 

Well, all it has is these things. How is it going to be able to validate what is the truth, what has changed, past a specific point? What is outdated? What is new? What are these? How is it going to make sense of all of this? 

And so I realized that I needed to change my thinking about the prompting and not just give it something and, here, give me an output. I need to do a little bit of what I was talking about earlier with the rhetorical toolkit about thinking about prompting as a turn-based conversation, and a line of questioning. I need to be thinking more about, how can I set the stage? 

I'm writing this for this purpose. Here is what good looks like. Here are some examples. And I also ask it to verify and validate information. If you give me the first output, that's not where the conversation ends. I need to follow up and ask it some question. What are three possible flaws with this? 

I think that AI tools are good at editing themselves in some ways, but they have to be prompted. So really thinking of this more as a conversation, that's how it's transformed my thinking a little bit, from, here, I'm uploading it, and I'm not getting it what I want, but it doesn't know what I want. Now I have to train it to know what I want. 

Jill Finlayson: I think this is a critical piece. There's a mindset shift going on because people have grown up with Google Search. And this is not Google Search. Can you say a little bit about more what has to happen to make sure that it's not making things up? What are some of the other things that it's prone to doing that you may not want in your writing? 

Sara Faradji: Yes, so I think when we think of search, and we are just asking a question and we might get a list of links, or we might get some information that is surfaced that's coming from the surface level or interpreted from the links, then we're only seeing a very cursory story. But it's not always from the lens of an actual person using this, or it's maybe not looking at it in the same way that a scholar or a developer or someone might be. 

And it's, again, just skimming the surface. Back when-- if you're writing a dissertation or doing deep research, you might spend a lot of time in the library going well beyond the scope of search to really dig into that research. And I think that with using AI tools, it can help bring some things to the surface. But you have to be questioning and going deeper and deeper. 

You need to be asking, what were the sources, and then actually opening and looking into those sources, if you can, and reading and understanding the context, because, sometimes, with that cursory level, maybe it's just reading the abstract of something, but it's not getting much deeper into the real situational information, whereas, of course, what we're doing research, we need a lot of different facts and statistics put together and analyzed and all these various situations to be able to draw an effective conclusion. Not just a single source proves that. 

And so I think these are things that we still need to keep in mind when we're researching. We have to be asking those types of questions. We can surface and find things at the initial level, but we still have to do that work of reading through, understanding that context, probing a bit more, to be able to make sure that we appropriately fact-check and put information out there that can be trusted by other people. 

Jill Finlayson: So we're architecting the intent. We're providing more context. We're talking about the audience but also what we want to attain as an output, these guardrails. It sounds like it's taking an incredible amount of time. So why does this onboarding of AI feel slower than just writing it yourself? 

Sara Faradji: I think that's a really good question, because we keep hearing about how AI is going to make our jobs so much faster. We're going to be able to generate articles and knowledge so quickly. We're going to be able to produce twice, three times the output that we did before. 

But I think there is a truth to that, but at the same time, there is this long Durée or this longer relationship that we have to have with the AI. When we first start using it, like I was mentioning, it's like training a baby in some ways. It only knows what you give it. 

And what I found really valuable is we have to let the AI deal with our frustration a little bit, be able to voice information. And the more information it has, the better it's going to get. So that future state of your being able to really turbocharge your workflows and everything, I wouldn't expect the very first thing you produce, the first question, the first output, to be able to make it like that. 

You really have to train it and be able to give it your constraints. It does take some time, but I also think that you can really, as you get to understand it more, and as you use certain tools, there are some options where you can create documents or a folder and have all the source information there, or you can add it to a project. And then you don't have to keep repeating all of this information every single time you start the conversation. 

Gradually, it is going to learn. But it has to baseline and understand you a bit more. So it really does take that time. 

Jill Finlayson: So upfront costs for longer-term efficiency, but also probably more completeness, because it's accessing a lot more information than we as an individual might have. 

Sara Faradji: Right, exactly. It can pull from a lot of different sources. I think back to when I was doing research on my dissertation. I would have all of these documents. I would go and read many, many books and have all of these lists of quotes, all these highlighting across all of the different sources. And there's still so much value in that and being able to do that deep research. 

But we can have a coworker to help us to find and organize some of that information and to make sure that we can find what we need. We're still doing the work, but it might be surfaced or organized a bit better so that we can maintain focus on being able to draw important and sophisticated conclusions. 

Jill Finlayson: So you work in a very sensitive field, cybersecurity, where data and accuracy is essential. Tell us about that work and how that context made you use AI perhaps a little more cautiously. 

Sara Faradji: Yes, I work in the ever-changing field of cybersecurity, where technical documentation is very important in this field because if you misconfigure something, if you don't follow a correct step, then that could lead to big issues in your environment. And so it's very important that users trust this documentation. 

I didn't want to just offload all of that thinking and the complexity of technical writing to AI, or to ask it to do the technical piece. Something I found really interesting when getting into these types of roles is I think there's an assumption that if you are an engineer, then you have very strong technical skills, but maybe you don't necessarily have the best communications or writing skills in some ways, and you can use AI to offset that type of work. 

And then there's an assumption that if you are a writer or a communicator, then maybe your strength is in being able to translate complex information into easily digestible information, but you might not necessarily have the deep technical background. But you could offset that to AI. 

And in some ways, I think that there's some flaw in that kind of thinking, because we can use AI to help make up for those gaps. But at the same time, it can't replace in the sense that, oh, I'm just going to trust what the AI gives me as the technical output, and I don't need to verify that information. 

I'm still working-- I use the AI tools to help me get first drafts based on my style guide, my templates. But then I ask it, what are questions that I need to ask the product managers, or what are questions I need to ask our CISO? What are the privacy considerations? If I'm a user going through this for the first time, then what problems might I encounter? What are some troubleshooting? 

And then it can really start to get into some interesting information beyond just, here's the factual information I'm giving you.

Make a guide. It's more, how can we fill in these gaps and be able to answer key questions? And when I'm reviewing the output, if there's something that I know that I have not verified or wasn't from the text, I'm going to flag that and ask more questions about it. 

So certainly in cybersecurity and other fields, you have to be careful about what you put out there and not just using the AI tool as the Subject Matter Expert. But there are certain ways that it can help with changing your workflow a bit. I'm now in a place where I don't just write the information all on my own and then have a long list of questions for a Subject Matter Expert. 

I have very focused questions that I might ask certain stakeholders who know more. And we can also use different kinds of technologies so that once we have the verified information, we let the AI tool know that that's the correct answer. And it'll learn from it later on. 

Same thing with style guides-- if there's something that I'm realizing is a bit off, then I'm going to train the AI. From now on, use this type of response, or, know this fact, and make sure that's correct in our master materials. 

Jill Finlayson: So it can help you identify a little bit what you've missed. It has the complementary skill sets. How do you use it to find blind spots in your own arguments? If you think you've done everything correct, how might you use it to explore if you actually do have it correct? 

Sara Faradji: Yes, that's a good question about how can it find some things that I might have missed. So I might actually ask it some questions. I'll give the information, and then I will give it some roleplaying advice. 

First, imagine that you are a CISO, and you're looking at this, and you need to understand, how do I find the high-level metrics, or, what are the privacy considerations? In the event of a certain issue, then what is my concern going to be? 

And then I might have it review from the perspective of a security analyst and say, what are some different types of issues that you might find if you were this persona? What questions would you ask back at me? And then I might even ask some questions, like, are there any errors in this-- or is there any information in this document that isn't answered by the initial resources that I gave you? Is there any logic that seems unsound? 

I just go and ask it these questions and ask it to verify and give me really clear answers. And in some ways, that's a great process, because I think both the AI tool and myself are both learning at the same time. 

I'm building in my technical knowledge in some areas, and it's learning how to write better, how to better explain how to surface gaps. So in that way, it's a mutual learning when we think about it as your copilot, but also your kind of learning partner. 

Jill Finlayson: I like the focus on the process and being able to use it to really analyze and to give you feedback, ask questions, so that two-way street that you were talking about. I see this having real applications for people who are maybe doing a startup and they have a pitch deck. 

They could say, hey, shred my pitch deck. Tell me what questions I'm going to be asked. And it allows you to not only anticipate the questions that you might be asked, but prepare answers for those questions. 

Sara Faradji: Yes, absolutely. I love seeing how you almost ask it to question everything that you have written and say, what's the flaw here? I actually think AI tools work a lot better that way. Rather than saying, is this good, or, did I do everything right, I think that AI tools are a lot better at finding what's wrong. 

And they're good at self-correcting, too. It's interesting. I've been experimenting with using AI tools, looking at GitHub repos and being able to, on an automated schedule, self-audit a certain thing. And it'll be interesting to see how maybe through the first self-audit, it didn't find some errors. 

Then it goes back and says, I actually found this other source that contradicts what I was saying. And so it's really interesting to be able to continually question, even have it question itself, to be able to find what the flaws are. 

And I would highly recommend that type of mindset of going into the negative or having it find what's wrong rather than just create something or, is this good, type of thing. So really start with those negative-leaning types of questions, and you'll be, I think, more impressed with the output. 

Jill Finlayson: One of the things I think about is, it is hard when you face a blank page. It's much easier to edit someone's work than it is to start with that blank page. But there's also a flip side to this as well, where you have just a jumble of documents and ideas and you need to get it organized. How would you use AI in that situation? 

Sara Faradji: Yes, this happens to me a lot. You will get all kinds of things-- meeting notes or tickets and various sources that might not be organized very properly. And you need to make sense of the mess. And I think it's OK to start with that. It's better to start with that using an AI tool than without one, in some sense, because, otherwise, there's a lot of manual reading. 

And like I said with that rhetorical toolkit, I think that the stasis theory comes into play really well here because you have this note pile. You have a transcript, just a brain dump. And you bring all these materials and then say, can you look at these documents through the lens of stasis theory or start asking some of those questions I was mentioning. 

What are the factual claims that we can pull from all of these documents that we need to be surfaced in the article? And then what are the quality judgments that we should ask? You could even ask it just very briefly, sort these into facts, definitions, judgment calls, or optional types of configurations and things like that. 

And so that way, we're starting from a place of brainstorming and not just, create a document out of this. We're saying, can you pull out the facts first? And it's a gradual type of conversation. And then you can shape it and evolve it. 

And maybe it's not until the fourth or fifth prompt where you really start to get into, OK, let's start to build a draft from this. So I really like starting with that to try to say, before I start manually reading through all of these things, can you help me pick out what the most salient points are? And then we start from there. 

Jill Finlayson: One of the other challenges is AI has been trained as a-- it's a service. They want people to keep using it. So AI tends to be very flattering of your work and telling you've done very good things. Have you had any issues with AI not telling you when you've got something wrong? 

Sara Faradji: Yes, this happens a lot. It is very much wanting to please you and say, yes, this is great. It's very quick to call out what you're doing well, or, this is gold, all of these things. It's very aspirational in that way. 

And so that's why I think kind of leaning into the negative is important. I think that when you give it a draft, its inclination is to give you a slightly more polished output, but it's not necessarily going to give you a genuine assessment of what's wrong every time. 

And so I sometimes will ask it questions like, what is the weakest part of this argument. Or what would a skeptical reader push back on? Try to look at this from the perspective of someone who's convinced I'm wrong. 

And these go back to rhetorical strategies. When we're teaching English classes, we teach the conversation about developing a good counterargument and a rebuttal. And I think those are good to keep in mind when working with AI tools is don't just come to it looking for the praise and saying, here's some information. Really ask those deep questions. And you may find that some of the things that it's pushing back on, maybe they're not really as relevant. But it's still good to at least see some of that information and see it kind of leaning into the more negative or questioning kind of tone, rather than just validating everything you say. 

So I think you have to give it some of these types of prompts to look at it as a place of that critical inquiry, the questioning side of things. And you're going to get some better results. 

Jill Finlayson: This is where your own domain expertise really can help you. There have been some articles talking about how people who are older are better at using AI because they can interrogate it more effectively, and young people maybe don't have as much of a lived experience to see. So why is our own domain expertise important? 

Sara Faradji: I think the domain expertise of humans is more relevant than ever, because if I'm just working with an AI tool, and I get an output, and I believe it must be valid because I am not able to question it otherwise, then that's where the danger can lie. 

And in some ways, I think if you're just delegating or offloading a certain type of expertise to AI, then you're really doing yourself a disservice. And again, we have grown up in a culture where we have very specialized types of knowledge. Someone is trained to be a writer. Someone is trained to be an engineer. 

But I think AI is kind of breaking down these silos in some ways. And I think I've heard the phrase "collapsing the talent stack," or someone can be a jack of all trades and do many of these things with AI kind of helping to fill in the gaps. 

And I think that is exciting in some ways, because you can make a tool, and you can vibe-code or something and help to fill in the gaps that you might not have. But if you don't have that domain expertise to know this is going to work, or this is going to scale well, or there's a problem here, then it's not going to be as successful. 

So I think the domain expertise is important, but I do think that we can use AI tools to help us get that expertise and not just take the output at face value. Something that I've been doing is when I'm looking at steps that are very technical in nature, I look at the output that it gives me. 

And then I say, is this supposed to be all one command? What would happen if they didn't do it this way, if the user did it this way? What would happen if the user approached this in a different type of angle? What would happen if the user clicked this instead of that? 

And so I keep kind of going through these different scenarios at every step, and then let the AI teach me and fill in the gaps. And so I'm building my knowledge and expertise at the same time, as I'm also using my expertise in the writing to be able to frame what it should look like. 

Jill Finlayson: The democratization of the technology and the growth mindset really does allow people with problem space expertise to now actually go forward and solve some of these problems. I like this positive spin because I think a lot about atrophy, that we're going to lose some of the skills because we're offloading it to AI, or, for young people, foreclosure, where they're never going to learn some of these skills because AI is going to do the work for them. 

I say that, but when you are describing these rhetorical toolkit, that's really teaching critical thinking skills. So maybe we don't have to be as concerned. 

Sara Faradji: Exactly. I think that someone who starts working with AI tools, it's going to be great to see how they approach it with a skeptical mindset and not just an evangelist mindset. I think there's a lot to be excited about with AI. But if we go into things with the perspective of a skeptic and being able to ask more questions, then we're only going to be able to use the tools better and we're only going to learn more from it. 

And so I do think that these skills that you're learning in English classes and whatnot, they still come into play. And they're even more valuable, in some ways, because we're not just taking the output at face value and saying, this must be the fact. We always go into it with that skeptical mindset and being able to help us discern what's fact from fiction, what's credible, what are the edge cases. 

I think that, really, the edge cases are most important when it comes to documentation or any type of writing. We can have it give you an output or say what it is, what it's not, or what's on the edges, or what are the outliers. That's what's going to make these documents and this writing much more valuable. So I'm excited to see how we continue to push on the tools to be able to think more critically alongside us. 

Jill Finlayson: Let's talk more about these tools, because there are things you can do with AI that will lead to those efficiencies you mentioned earlier. Another definition for skills is what are you teaching the AI, specific skills, or sometimes they're called Gems in Gemini. Can you talk a little bit about what you can do to train the AI? 

Sara Faradji: Yes. So I highly recommend using Gems and also projects to be able to build very focused initiatives where the AI is going to learn over time. You're not just retraining it. I think with a lot of tools, we have a chat feature, where that's just a back-and-forth conversation. Those are good for just initial kinds of research, or you just want to have almost a search function to be able to ask a question and just get a general answer. 

But if there's a project you're working on-- you're working on revising your resume or your cover letter, or you're working on a very specific type of article project-- then I would make some kind of project or a Gem where you're going to have a style guide. You're going to have work samples. You're going to have objectives and almost instructions that you want this tool to know and repeat and understand. 

So every conversation you have with it is not going to be starting from scratch. It's going to have all of these resources. And the answer it's going to give is going to be closer to the output that you want. 

And over time, you can even talk with it and say, for example, I have my style guide. But I'm realizing that my style guide, I want it to evolve over time. I don't want it to be static. There are these just different edge cases that I want to keep in mind. 

Tell it that in the project. And it'll say, OK, I'll update the master file. It'll do that for you, and then it'll help to format the output from now on using what you've given it. 

Jill Finlayson: You've mentioned style guide a couple of times. How do we teach AI to speak in either our own voice or our corporate brand voice? 

Sara Faradji: That's a good question. I think that what I would suggest is first starting with the personal voice and what you like. And you can even use an AI tool to give it an initial document. Maybe there is a style guide that you have. Or even if you don't have one, just say, I want to make a style guide. 

And it might give you a quiz question. And it might give you two side-by-side documents and say, which one looks more like you? And you go through this process, or you can ask it to. And then, eventually, it's going to be able to make your own style guide based on your own voice. And it'll say, does this sound like you, and give you options like, this sounds like me. 

And be very straightforward. It'll get to learn more about what are your convictions. What would you never say? What would you say what sounds more like you? And then, from there, you can really be able to have the AI start to sound more like you. 

And then when you have the corporate style guide, that might emphasize more mission, vision, values, and things like that, and seeing if you can merge them. Say, here's my company's mission, vision, values. Here's some examples of documents that they have and their writing style. Are there ways that we could merge them? 

I almost make a kind of hybrid style guide. And maybe you're also building your style guide off of MLA, AP style, like a master style guide that a lot of organizations use. You can almost use AI tools to find that middle ground among these style guides and adapt it over time. 

Jill Finlayson: This might be technical, but if you've given it a bunch of examples of your voice, the way you write, and you've asked it to create this, how do you save that style guide? And then how do you use it when you want to draft an article or write an email? 

Sara Faradji: Yes. I think the style guide really has to have multiple parts. One is, what are the rules, and what are the formatting choices that you want it to follow? And then also, I would have as part of it, what would I never say? 

For example, I think there are a lot of even writing choices that AI tends to gravitate toward. The em dash is a classic one. And lately, I'm noticing the "it's not this, it's that," these types of styles that are very almost cliche. 

There are things that probably come up in a lot of writing. If we're looking at all the writing in the world that AI is picking up on, it probably sees these repeated. It's modeling after human language. But now I think we're getting at a point where that's becoming too repetitive and it doesn't make us unique. 

So in your style guide you have, maybe you want to set some constraints. I don't want to be using too many of these. I don't want the whole document to be kind full of this repetitive type of word choice or sentence structure. 

And then also, I think having the examples-- let the AI know what good looks like to you and show it, these are examples of things that I have written without AI tools. And this looks good to me. This is natural. This is something that is ideal. 

It needs to be able to understand that as opposed to all of the writing in the world. So again, having, with your style guide, the examples are good, and you can, with certain tools, projects, maybe you just have this file that's like a markdown file or something that is in a folder that your AI tool can read from, or you can upload it to the specific project. So it's always going to reference that. 

Jill Finlayson: So you can tell it to avoid these sort of robotic tells that it's been written by AI. You can tell it to, hey, I don't want to be trite. I don't want to use, common phrases. And then you give it the positive guide. 

But at the end of the day, how do you maintain authenticity? If you're an HR person and you're trying to rewrite a job description, and you want people to want to work at your company, how do you avoid not sounding generic or rote? 

Sara Faradji: I think that sometimes with these documents, it can be very easy to let the AI slip and use language like, our ideal candidate wants to be X, Y, and Z. Or it might say the cliche kind of language, like, I need the candidate to wear many hats or to be able to work in a fast-paced environment and things like that. 

And those all might be true. But if they're all put together in the job description, then it feels vague, or it might even have a negative connotation that, oh, this job might not be too specified. And so something I recommend if you are writing a job description or working with any kind of corporate language is to ask the AI tool to push back and say questions. 

Just start asking the AI questions and think about, is this job description written for somebody who really loves this type of environment? What would a candidate find confusing about this? What would a candidate find vague? 

And be able to ask it those. And how can I write this job description in a way that sounds engaging? And you still might have to go back and question on that, because then it might give you a lot of kinds of language with exclamation marks or with just general, vague, happy language too. When you describe the tone, just be specific and clear about what you want things to look like in some ways. 

Jill Finlayson: The sea of sameness happens on the flip side as well. When people are applying for jobs, they're using AI to help write their cover letter. They want it to be specific to the job, and they want to make sure they're mentioning the keywords so they can get through the filters. What is your advice to people using AI to write those kind of career applications? 

Sara Faradji: Yes, I think the best way to start with writing a cover letter or resume is to first look inward and not outward. And so by that, I mean you might give the AI tool your current resume or your cover letter, or even if you don't have these documents yet, just start explaining a little bit about your background and some work experience you've had. And ask the tool, what types of jobs would I qualify for in this job market? Or what would you say are my strongest skills that would look attractive to employers? 

And sometimes the AI tools can flag things that maybe you didn't even realize were the most valuable skills and things like that. And so first, to start, even before you look at a specific job ad, just start by looking within, and think about how you can really build a good kind of master resume or cover letter template, almost like that style guide. 

See, how can I answer certain questions? Maybe set up an interview with your AI tool to be able to do some of that introspective work and learn more about yourself. And then, when it comes to starting to apply, maybe you ask the AI tool to look at different job ads and think-- I would maybe ask the AI tool to consider, what types of job roles would I most qualify for, and what skills and experience are required or recommended for these jobs? 

And have those conversations to better understand, how would I tailor my resume for these fields? And then, after you have that back-and-forth, then you start to look at actual job ads, and then bring those in. And even before you start matching the resume and the cover letter with the job ad, say, here's the job ad. What are the top skills that this employer is looking at? Out of all of these different requirements, what is the employer really asking for? 

These are kinds of good questions to maybe get some good results to surface. Then, after you just start learning a little bit more about the job ad and trying to surface from this long list what's hidden or what's interesting, then start to put your existing resume or cover letter against the tool and think about, what should I change? What should I surface? 

Maybe you need to tell more of the story. I use the STAR method-- the Situation, Task, Action, Response. I like to work with AI tools to say, can we look at my bullet points and help me flesh out the STAR method for these? 

And then maybe I work backwards in that way to think, OK, now that we have more of that context and that rich information, now can we go back and reformat that bullet point with the context in mind to really show what the values and what the key metrics, what the ROI, all of these different things were that maybe weren't very evident in the first draft? And of course, give it some constraints. Don't make up experiences or information to fit what they're asking for and whatnot. 

Jill Finlayson: I like the focus on specificity. If you can write something in your cover letter that anyone can write in their cover letter, that's not doing you any favors, because I will generate that sentence for everybody. So tying back to that STAR method, and what have I done specifically that demonstrates this skill, rather than saying, I'm good at managing details. 

Sara Faradji: Yes, exactly. I think the specificity is important. An example I use sometimes where AI tools can help is, maybe you need to start thinking about your resume and your cover letter less in a task-oriented format and more in, what was the value that I brought, like a value-add resume. 

But again, if you don't have enough information to share, it's going to sound like every other resume, and then it's not going to help you succeed with the ATS and all these different factors, because if it's seeing the same thing across 100 resumes, you're not really going to stand out. 

You can use the AI tool to help you understand what the value really was that's beyond the task. But you have to do that work of being able to balance what is clear and correct with what is unique, what stands out, what would be impressive to this audience. I think also asking the tool those questions of, would this appeal to this hiring manager and things like that would be important. 

Jill Finlayson: AI can really help you prepare for interviews. It can help you to network. Is there any fear of being too prepared, a little creepy, knowing a little bit too much about the person you're talking to? 

Sara Faradji: It's funny because you can certainly use AI tools to do the open-source intelligence, or the OSINT work on your hiring manager, the recruiter, the company, things like that. And what I would recommend is you can do that research. That's great. But try not to come into the interview or have your cover letter say, I really align to your mission, and then just repeat the mission statement, or something like that. 

Think about what are the specific areas of that mission, or the values that align with your own personal mission statement or your own values. And think about where they align. And I think the same thing can apply when it comes to the AI tools. 

I think some people go into these interviews and they'll ask, what is your AI policy? I want candidates to think more about what is their personal AI policy. What are they going to use the AI tools for? What are they never going to use AI tools to do? What is your philosophy on this? 

And then when you're asking that question of what is the company's AI policy, think about how does it align with your own, because I think you'll most likely find that companies may not have a fully fleshed-out AI policy. They may want you to use it, but they don't quite have those guardrails in place. It's just we have to adopt. 

Jill Finlayson: This really levels up the conversation because it allows you to say, I've thought about these issues. I've thought about what responsible and ethical AI is. This is what it means to me. This is my practice and my process. Here's how I ensure that I'm getting accuracy. 

And I think that's really interesting because there's this double whammy today where they expect you to use AI, but if you use AI, they don't believe you can do it without AI. But if you can explain your AI policy, as you call it, that's going to show the critical thinking skills. To your point, they're looking for the skills, not the tasks. 

Sara Faradji: Exactly. And at the company I work for, there's a big emphasis on AI-native mindset. They want everyone at the company to be AI-native. And I've been thinking quite a bit about what that means, because AI-native doesn't mean I am going to use AI tools for everything I do, to the point where I'm just automating myself out of this role and the AI is taking control. 

It's really being thoughtful about what types of tasks AI is going to be able to help you with and continually questioning your current workflows, and where would it make sense to use an AI tool or automate this? For example, I've been using tools for scheduled tasks to be able to look through all of these different sources that I'm working with, whether it's email and tasks, all of these different tools that I use on a regular basis. Can it surface a really cohesive to-do list of what I need to do? 

When I finish day 2 week sprint or something, can it automatically look at all of those tickets and generate a summary that is tailored for a certain team's audience to be able to communicate that information to them? I mean, there are some things where the repetitive tasks or the things that are maybe on your task, but maybe they're not your favorite things to do, there might be some good ways to find automation. 

And then there's ways where you, thinking of your own personal-- your deep thinking and your skills, where your expertise is going to be best served, how can AI be that kind of coworker or copilot for you, but not just fully handing over the reins? 

I think that's really what AI-native is, is, how do you restructure your thinking? In a lot of ways, it's embracing and understanding yourself a bit more. It's a good opportunity to really be thinking more about what makes you unique and stand out, and how are you going to be in the driver's seat when it comes to AI, and not just this passive role and letting it take control? 

So being very clear and having a strong statement on how you're going to use these tools, what you're planning to learn from them, how is AI going to fit into your overall career strategy, these are all great things to be thinking about. 

Jill Finlayson: These are super helpful tips, and I love the fact that we're looking at applying this information and putting it into the workplace and using it to prioritize, using it to augment and support the work that we're doing so we can do it better with less of the annoying parts of our job. But AI is changing so quickly, so fast. What are the things they can do that will make staying current possible? 

Sara Faradji: That's a great question because I think it is overwhelming, in some sense, the notion of, everything is changing every day. We can't keep pace of all of these different tools. But I would just keep an ear open. It's kind of that social listening of looking what's happening on LinkedIn or social media or what your colleagues are doing. 

I do think that one big risk that I'm noticing is that everyone is working on these AI tools in their own individual silos in some ways. And they might be learning new things. But if you're learning it and you're helping with your own personal workflows, how can you help the broader team? Or maybe there's someone at a different company out there who's using the tool in a way that you hadn't thought about. 

So just be vocal. Keep your ears open, your eyes open, everything. And look for ways to collaborate. I think another thing that would be good to do is to make a plan, a schedule. And you could use an AI tool to even help you with this. 

Make sure that you're learning something specific every day or reaching a goal. And maybe one of those is, I want to spend 10 minutes a day training my resume builder AI tool. Or I want to spend 10 minutes a day building my domain expertise in this specific area. And I want to use this AI tool to help me learn something new every day about this. 

So I think just being consistent-- you might not have all the answers today, but trying to develop that plan of action and learn a little bit every single day. 

Jill Finlayson: Well, thank you for sharing your knowledge. 

Sara Faradji: Thank you so much. It was a great conversation. I appreciate your time, Jill. 

Jill Finlayson: Thank you so much. And with that, I hope you enjoyed this latest in a long series of podcasts that we'll be sending your way every month. Please share with friends and colleagues who may be interested in taking this Future of Work journey with us. And make sure to check out extension.berkeley.edu to find a variety of courses and certificates to help you thrive in this new working landscape. 

And to see what's coming up at EDGE in Tech, go ahead and visit edge.berkeley.edu. Thanks so much for listening, and I'll be back next month to continue our Future of Work journey. The Future of Work podcast is hosted by Jill Finlayson, produced by Sarah Benzuly, and edited by Matt DiPietro.

[MUSIC PLAYING]