Diana Alt [00:00:04]:
Hey y'. All. Diana Alt here on Worksheet Feel Good, the show where your career growth meets your real life. And I have a really important topic for you this week for my solo episode number two ever and that is about using Gen AI to review your resume. I didn't even know a couple days ago I was going to record record this podcast, but my friend Eric Recamp, who is an amazing C suite executive board level resume writer, released a report this week about Gen AI resume feedback for job seekers and I want to unpack it a little bit because it's really that important. People are burning tons of time and energy on trying to have Genai write or review their resumes and actually sometimes they're making it worse and I don't want you to be doing that. I want you to hear this episode, go and read the report, take some of the suggestions that I have, and then come back in a couple of weeks whenever I talk to Erica about this and learn more. So the study that Erica wrote or worked on was focused on Joe Jobseeker, Joe or Jane Jobseeker, not on fancy resume writers and also was focused on a person that doesn't really know about resumes very much.

Diana Alt [00:01:28]:
So I know some job seekers out there that are really savvy about resumes. That used to be me. That's part of the reason why I'm a career coach now. But for the sake of this thing, it was a, you know, basically geared towards understanding what kind of output A person that doesn't really know about resume best practices is just generally trying to get a job and doesn't necessarily have the skill, time or inclination to learn a ton of prompt engineering. So that's the setup. What they did with the study is they first off they were using eight generative AI tools. So the tools really quickly were ChatGPT, Claude Copilot, Gemini, Perplexity, Meta AI, Grok, and Deep Seq. I will confess that I've only used a few of those.

Diana Alt [00:02:24]:
I am mostly a chatgpt person because I've spent a lot of time training my ChatGPT on the important information and how I think about things and my best practices and checklists and all that kind of stuff. So that's my go to. And then I use Perplexity Sum. I've never used any of the other ones for anything related to resumes, but that doesn't mean that you shouldn't. The the reason Erica tested all of these is because she knows these are the ones people use, especially like co Pilot and Gemini which people, if they have jobs or if they like have their own Google Workspace account, they're, they're free or they're easy to access. There's not a big barrier to entry. Something that I ended up asking Erica whenever we were exchanging messages and comments on LinkedIn this week after she released the report is did you test any of the purpose built resume scanning tools like job scanner, teal or resume building tools? And there's a whole lot of them and she did not do that. And I think we're going to get into why whenever I have her on the show.

Diana Alt [00:03:28]:
So she had these eight tools and then she had three resumes that, that she created. And in the report are the resumes and our very, very detailed feedback. So, so it's a 26 page report. The first like 11 pages set up the study and the findings. And then the back part is you can look at the actual resumes that were put into the tools and you can see specific details of feedback and how that feedback was classified. So the three resumes were, number one, a good quality resume that was for a person in biotech. The second one was for a fair or adequate quality resume that was for someone in the nonprofit leadership sector. And the third one was a poor quality resume for someone in it.

Diana Alt [00:04:16]:
Now I am really, really hoping that since Erica said this is a preliminary study that she's going to dig deeper because one of the things I really want to see is what does the feedback look like if there's different quality resumes for the same person or targeting the same kind of role. So you know, what does good, fair and poor for the IT resume look like with the same kind of feedback? Because that kind of stabilizes some of the stuff. But hopefully we'll get something like that later or maybe I'll work with her to do it. I don't know. But there's a little bit of an apples and oranges thing in there that might concern some people. But broadly, based on my experience using AI for reviewing, scanning, creating, optimizing resumes, anecdotally, I'm not surprised by what's in the report. So the re. The way that this was conducted is that for each of the eight tools, the resumes would get uploaded and the prompt of give me feedback on my resume would be entered, which is real, real similar to the exact prompt I've heard people, other people say.

Diana Alt [00:05:28]:
So no, I don't know what research Erica did to decide that that was the right prompt to use, but I know that people are using that Prompt and the responses that were given back were categorized in one of four buckets. Positive feedback, which is just compliments on the resume. Constructive feedback, which is changes to make that are actually going to help based on standard resume practices. Unnecessary recommendations, which are things that like, yeah, you could do it, but it's a waste of time, you know, probably. Probably isn't going to help, but it might not. Probably shouldn't hurt too much either. And then there was the misleading or potentially damaging feedback category, which is of course the most concerning. Something that was done in the study is that Erica did not really classify feedback on things that were opinions that are not broadly held as top resume standards.

Diana Alt [00:06:26]:
So, for example, there's a lot of different opinions on whether you should list the number of years of experience that you have on your resume. And she, because some people are concerned about age discrimination or think that results are more important. There's all kinds of reasons to say that, but she didn't. If some, if one of the tools said, hey, please, the years of experience on there, she didn't classify that as good or bad. There is a section that lays out some of the standard recommendations for resume writers that are globally thought to be correct. They are, you know, the standards that are taught in resume certification programs. They are standards that, when I read through the list, I'm like, yes, this is what I look for. This is what I use.

Diana Alt [00:07:13]:
A couple of examples are that you want to make sure that it's clear that the resume is targeted for a specific type of role. You want to make sure you list your work experience, which sounds like a duh, but who knows? Like, who knows what people are going to do. But you want to have the company, the title and the details of the work experience and then reinforcing qualifications with that information. So you want to make sure that you're building up your story. So at a high level, here's some of the findings. So on the strong original resume that was a biotech resume, that resume sadly got the most confusing result of all. It had more feedback that could hurt than help by a lot. So the report said something like nine pieces of constructive feedback, feedback that could help the resume versus 27 pieces of potentially damaging feedback.

Diana Alt [00:08:15]:
Anecdotally, when I have worked on final revisions to resumes that I'm writing or reviewing, particularly ones that I'm writing because I already know that they're good quality, I know how to write a resume. The ChatGPT does start to give advice that sounds weird and it starts to contradict itself and Things like that after a while. Each platform has its own quirks. Chat Copilot apparently swings to its extremes. It actually had the most pieces of feedback. Chat GPT like really loves to try to get creative with formatting. But basically Copilot, despite kind of being a little bit all over the map and having extremes in its feedback, was considered to have some of the most constructive critiques. So that was cool.

Diana Alt [00:09:08]:
I've never used, I've never used Copilot. I'm like a Google girly, so could be fun to play with that. On the fair resume, the results were also inconsistent. There also was more harmful feedback than helpful feedback, but it was not as stark as the good resume. The worst advice that was noted in the report was from Claude. So this was a nonprofit program leader type of person. And Claude suggested repositioning as just an organization leader. And that's way too broad.

Diana Alt [00:09:44]:
Like that's not enough context. That's not going to grab the eye. It potentially could be harmful for any sort of Boolean searches that are done in querying. So that was an example of tough feedback that is is not going to do anybody any good for the person receiving it. And then on the poor resume things look better. So the poor resume didn't have hardly any positive feedback, which you wouldn't expect. If it sucked and the resume is in the report, you can see it, it sucks. But it had so much negative feedback that if I was thinking about it as a coach, I would never just lay all that feedback at once on a client without any context.

Diana Alt [00:10:28]:
I would try because they might get discouraged and then just say I can't update it and then either never apply for a job or apply with the same bad resume. So it would be really cool if triaging could happen. And as a coach I would do that. I might use an AI tool to do an initial screen, but then I would go and say, okay, well if I can only have them fix three things, what are the first three things I'm going to have them fix? And then what are the next three things I would have them fix? There was a lot of good constructive feedback that was low hanging fruit, probably not too hard for a person to fix. And that's really good news because that could take you from like an F to a D or a C minus or something like that and then you can continue to improve. But there still was some misleading advice on there for poor resumes. The report said that meta AI did the less the least damage. I've never used meta AI for anything I Don't even like the AI summaries of conversations on Facebook.

Diana Alt [00:11:24]:
They drive me crazy. But maybe I'll have to rethink that. So for me, the most striking thing was what is called in the report the help and hinder gap. So when we think about help, help is positive feedback and constructive feedback that's in the help bucket. And unnecessary feedback and potentially damaging feedback are in the hinder. The good resume had the most gap between what's going to help you and what's going to hinder you. The fair resume still had more hindering advice, but it's not as bad as the good one. And then the poor resume was better.

Diana Alt [00:12:06]:
Like it had more feedback that would help than it did that would hurt. The problem I have is that for the most part, when my clients come to me, they don't have poor resumes. There are a lot of poor resumes out there, but most of the resumes I see actually fall in the fair bucket, especially if people are not drastically changing what they're doing. If you were a senior project manager somewhere and you got laid off and you just want another senior project manager job, usually the targeting is okay from the last time that you did your resume. And so, you know, the way people operate when that happens is that they'll take the resume and just add their last few years since the last time they got a job or updated it. And so it lands on fair because the positioning is okay, it's just not powerful, or they don't know how to write a resume to begin with. And they got a job during a boom time with a fair resume, which happened a lot for me personally. I use chat GPT to review and optimize resumes the most.

Diana Alt [00:13:15]:
And I do it piece by piece. I kind of have a way of doing it that I can unpack for people that really want to geek out about it. But I use chat GPT mostly because I have trained it based on my expertise, the niche that I serve, which is mostly tech marketing, like tech adjacent people, leaders, senior people, up to VPs. Like that's my universe of people. And I've trained on that data. And the average resume I actually write is really good. So I've mostly looked at fair resumes that people send me to review or as a starting point for writing a resume, and then good resumes, which I sometimes get from clients and I always write for myself. So if I was just dropping in, give me feedback on this resume, I would get mostly advice that didn't help.

Diana Alt [00:14:07]:
This is really dangerous for new Resume writers because people get enthusiastic and they love the idea of writing resumes. Maybe they've taken some training, but perhaps just need a little bit of a confidence boost from their AI tools of choice. And if they take that thing that they worked on using their knowledge, but then go try to get what is the advice that Gen AI is going to give me, they might make the resume that they just created using their knowledge worse based on the combined intelligence of the Internet, which I've been on the Internet and it's not that smart. So what to do about it? Here's my suggestions. If you are a job seeker that cannot afford or does not want to hire a coach or a resume writer to review or write your resume, there's three things that I want you to tweak. They do not require you to become a huge resume expert. They do not require you to become like the fanciest person with AI prompt engineering. The first thing I want you to do is instead of just saying give me feedback on this resume, I want you to give it something to compare your resume to.

Diana Alt [00:15:12]:
The most useful things are a job description that is, you know, in line with the target that you wrote the resume for, or just a tight targeting statement. So if you don't have a job description handy that you think works, you can say, I want to be a senior technical project manager in a fintech SaaS company. Series B, series C. That should be enough to try to get people going. The second thing I want you to do is make sure that you ask your AI to play a role. I do this frequently. I don't do it 100% of the time. I need, I'm getting better and better at it.

Diana Alt [00:15:49]:
But if you ask the, if you ask your chat GPT or your clot or whatever, I need you to put this hat on. I need you to put on the hat of a director of a PMO that might hire that senior technical project manager. What would they think of this resume? I want you to look at it like a recruiter would look at it, or I want you to look at it like a peer in this type of role would look at it. Ask it to play a role before it gives you feedback. And then the third thing I want you to do is consider asking the tool to rate your resume in terms of likelihood to get called in for an interview. The goal of a resume is to get an interview. It's not to get a job. You can't get a job without the interview.

Diana Alt [00:16:39]:
But once you're in the door doing interviews Especially once you pass a recruiter screen or like a, a one way interview screen kind of thing, you're dealing with people at that point, you're already qualified and they have seen what they need to see from your resume and now they want to talk to you in more detail. So if you ask it to rate on a scale of 1 to 5 stars, 1 to 10, how likely is this resume to secure an interview call? For me, that will be more constructive. I do this regularly whenever I'm working with resumes. But the trick with this is that you're never going to get a 10 because AI tools are designed to continually iterate on tasks like this, you're never going to get a 10. So if you start with your resume and it's a three because you have a poor IT resume like we see in the study, but, but you implement feedback and you get it up to a six, a seven, like that's showing improvement. But when you have a good resume, yeah, you know, maybe it's an eight. Whenever you do it, if you make some changes to get it up to a nine and then you also quickly check to make sure none of the required skills or desired skills are missing, you probably can stop. Instead of making yourself crazy.

Diana Alt [00:18:05]:
I would rather have you get to an eight or an eight and a half and make sure you're not missing a major skill and then start networking with people than to continue trying to go over and over and get the 10 that none of these tools are ever going to give you. So that's kind of my take on it as I kind of started earlier. What I would love to see Erica do next is or somebody study this with the same, same role and target at different levels of quality. And it would be really interesting to see for tech resumes and non tech resumes because those are different and there's a lot more like keyword richness sometimes for tech stuff. And then also looking at like executive, you know, manager, senior level, individual contributor and entry level. Like does this calculus change at all for those different levels as well? I'd love to see what happens with a second pass because what I found is when I'm iterating getting feedback on a resume, it starts to step all over itself. So you know, what does this look like on a second or third pass would be really cool to know about. And number three is more detail on the purpose built resume tools.

Diana Alt [00:19:17]:
The spoiler alert that she shared online is that they're not great and most people don't want to pay for another thing. So if they already have a Claude premium thing for 20 bucks a month or whatever clog costs, they might not want to pay for a resume tool as well if they're watching their pennies. So that's my take. Bottom line, don't ever just ask the Gen AI to just give you feedback on a resume. Make sure you're giving it something to target to give it a role to play and consider asking for a measurement. And good luck. Anybody that's in the job market. I know it's tough right now.

Diana Alt [00:20:00]:
I work with people every day. If I can help you, please let me know. And I hope everyone has a good, wonderful week and I'll see you next time.