Learning Library

← Back to Library

Decoding Company Strategy Through Job Posts

Key Points

  • The speaker demonstrates how large language models (LLMs) can transform the traditionally manual process of reading job postings into a strategic, automated analysis that reveals company direction, product focus, and hiring gaps.
  • By crafting strategic prompts, users can instruct an LLM to scan large sets of recent job listings, categorize themes, detect weak points, and infer broader business tactics without needing to manually review each posting.
  • A live example using an app built on the Lovable platform shows the LLM analyzing Anthropic’s job ads, deducing a focus on scaling core AI technology, a shortage of platform engineering hires, and a strong emphasis on alignment science and model welfare.
  • The speaker promises to share the exact prompts and API‑based tool (with a simple setup for engineers and a no‑code alternative for non‑technical users) so the audience can replicate the approach on any company’s job postings.
  • This methodology not only aids job seekers in targeting opportunities but also provides B2B sales teams and analysts with actionable intelligence on competitors’ strategies derived directly from hiring signals.

Full Transcript

# Decoding Company Strategy Through Job Posts **Source:** [https://www.youtube.com/watch?v=DHZgLzwWXfs](https://www.youtube.com/watch?v=DHZgLzwWXfs) **Duration:** 00:12:58 ## Summary - The speaker demonstrates how large language models (LLMs) can transform the traditionally manual process of reading job postings into a strategic, automated analysis that reveals company direction, product focus, and hiring gaps. - By crafting strategic prompts, users can instruct an LLM to scan large sets of recent job listings, categorize themes, detect weak points, and infer broader business tactics without needing to manually review each posting. - A live example using an app built on the Lovable platform shows the LLM analyzing Anthropic’s job ads, deducing a focus on scaling core AI technology, a shortage of platform engineering hires, and a strong emphasis on alignment science and model welfare. - The speaker promises to share the exact prompts and API‑based tool (with a simple setup for engineers and a no‑code alternative for non‑technical users) so the audience can replicate the approach on any company’s job postings. - This methodology not only aids job seekers in targeting opportunities but also provides B2B sales teams and analysts with actionable intelligence on competitors’ strategies derived directly from hiring signals. ## Sections - [00:00:00](https://www.youtube.com/watch?v=DHZgLzwWXfs&t=0s) **Decoding Company Strategy via Job Postings** - The speaker shows how modern LLMs can instantly analyze large sets of job listings to reveal hiring trends, infer corporate tactics, and empower both job seekers and businesses, providing ready‑to‑use prompts for immediate replication. - [00:03:28](https://www.youtube.com/watch?v=DHZgLzwWXfs&t=208s) **Identifying Hiring Gaps and Risks** - The speaker explains how analyzing Anthropic’s current job postings reveals a scarcity of internships, entry‑level, and platform‑engineering roles, suggesting looming technical‑debt issues, scaling challenges, and a possible need for extra capital investment. - [00:06:48](https://www.youtube.com/watch?v=DHZgLzwWXfs&t=408s) **From Trivial to Treasure: Data Reimagined** - It explains how AI models turn once‑ignored information—like job postings and public selfies—into valuable insights, highlighting privacy risks and new opportunities for job seekers, product managers, and salespeople. - [00:11:33](https://www.youtube.com/watch?v=DHZgLzwWXfs&t=693s) **LLM‑Powered 3D Company Insight** - The speaker explains how large language models can deliver rapid, multi‑angle analyses of a firm’s culture, strengths, weaknesses, and strategic signals—like engineering dynamics and scaling issues—and encourages sharing prompts so product teams, job seekers, investors, and salespeople can harness this new class of accessible data. ## Full Transcript
0:00In the next 10 minutes together, we are 0:02going to crack the code on reading job 0:05postings. It's a lost art and it's one 0:07that we can absolutely transform in the 0:10era of AI. And I don't just mean for job 0:12seekers, although that's obviously a 0:14huge benefit. You also can infer company 0:17strategy, B2B sales approaches, and all 0:20kinds of other things just from reading 0:23job postings. A year ago, I wouldn't 0:25have been able to recommend this. But 0:26now, with where LLMs are at, I can 0:29actually give you three different 0:30examples in the next few minutes that I 0:33was able to spin up that give you a 0:35comprehensive approach on how to read an 0:38entire company's strategy from just a 0:42set of job postings. And yes, I'll be 0:43dropping the prompts and everything else 0:45in the post, so you'll be able to follow 0:46up and do it yourself. So what are we 0:49talking about when we say a job search 0:51strategy informed by job postings? In 0:54the past, we would have said, "Hey Nate, 0:56go and get a thousand job postings or 0:59the last 100 from the last 90 days and 1:01do this yourself. Conduct a manual 1:03review. I want you to categorize 1:04everything. I want you to identify 1:06commonalities, spot weak points, notice 1:08what they didn't post. Now look at how 1:10the products they offer compared to the 1:12job postings." You see how it goes on 1:14and on. Not anymore. You can get all of 1:16that done in just a prompt. In fact, you 1:20can get more done than you could before 1:22because you not only have the volume 1:24game, which you can play with LLMs, you 1:27also have the strategy game. So, you can 1:29give the LLM a strategic prompt and you 1:32can tell it to reason and infer in a 1:34particular way over a set of job 1:36postings that it searches and it will do 1:38that and it will come back and it will 1:40give you a view. Sometimes I think just 1:43showing it is way easier. So, let me 1:45show you an actual response that I built 1:48about a real job posting situation at 1:50Enthropic, the major AI company. Check 1:53it out. All right. So, I built this 1:56handy little app in Lovable just to 1:58showcase what you can do with it. Don't 2:00worry about these initial fields here. 2:02If you're uh an engineer, it's really 2:05easy to put in an API key and use this 2:07yourself. I'll be sharing it. And if 2:08you're not, you can follow along and I'm 2:10going to give you some prompts that you 2:12can use in other search engines. So, if 2:14you don't know what an API is, you don't 2:16have to care. But look at what you get. 2:18So, this is analysis results generated 2:20today, 9:24 when I'm recording this. And 2:24it gives you so many different 2:26components to look at. It infers a 2:28product strategy, doubling down on their 2:30core AI build, uh, and suggests that 2:33they have a lack of fresh platform 2:35engineering hires, which would indicate 2:36that they're focused more on scaling 2:38existing tech right now. That aligns 2:40with what we see from Anthropic's recent 2:43moves. It seems like a solid insight. 2:45Meanwhile, alignment science and model 2:47welfare roles indicate a willingness to 2:48tackle unsolved safety and ethics 2:50problems. Again, aligns tightly with 2:51what we see from anthropic. We go down 2:54to inferred B2B sales approach. They're 2:56calling out that this signals a push for 2:58rapid enterprise adoption based on 2:59startup account management and B2B uh 3:02marketing. And there's little evidence 3:03of dedicated sales engineering. So, one 3:06of the things that's really interesting 3:07is you can start to infer a B2B strategy 3:10from this. You can start to look at this 3:11and say they don't have dedicated sales 3:13engineers yet. They don't have post- 3:14sales technical support. They're very 3:16early in their B2B startup account story 3:18here. There is an opportunity to come in 3:22and offer solutions for a sales team 3:24that is probably under stress right now. 3:26And you can read that from the job 3:28postings. You can infer that. Now, what 3:30if you're a job seeker? What does that 3:32look like? Well, they don't have 3:34internships or entry- level roles posted 3:36right now. and they have very few roles 3:38for platform engineering. And so what's 3:41interesting about that is that they are 3:43essentially 3:46essentially setting themselves up for a 3:49potential technical debt risk as they 3:51scale. And that is indeed what we see in 3:53some of the recent outages and the uh 3:56work that Anthropic has done to their 3:58credit to talk about why the outages 4:00occur. they are struggling to keep pace 4:03with scaling demand and they haven't yet 4:05invested in platform engineering. And so 4:07another another insight here that you 4:10can see as you read through this if 4:11you're putting another lens on this is 4:12that Anthropic may need an additional 4:15capital injection in order to start to 4:17scale some of these platform pieces out. 4:19It has inferred cultural insights which 4:21seem fair for what they're worth, but 4:23it's very easy to get them. It's trivial 4:24to tweak the prompt and get what you 4:26want. Inferred company weaknesses. It 4:28calls out platform engineering. It calls 4:30out a miss on PM, QA, and customer 4:32support. What's interesting here is that 4:36this underlines the sort of research 4:38bones of the company, where the company 4:40came from. This will not always show up. 4:41These are these are individual insights 4:44that you get per company. Now, this is 4:46the part that I love the most. You can 4:48actually see why the model did this, 4:50right? It will give you a table and it 4:52will say, "This is what I read. This is 4:54the link to it. This is the reasoning, 4:56and this is the claims that I'm 4:57confirming here." here. And you can see 4:59that it's all recent stuff. This is not 5:00old postings it's working from. Is it 5:02perfect? No. Does it underline how much 5:06you can get out of just looking at job 5:09postings? Yes, it does. Yes, it does. 5:11Now, I want to give you a couple of 5:14other ways to look at this. This is not 5:16just something where you have to use a 5:18custom lovable app. You can do it 5:20directly in chat GPT. You can do it in a 5:22search engine like Perplexity. I have 5:24prompts for that. The key to take away 5:26is that the quality of this assessment 5:29depends on your ability to ask very 5:32clearly for exactly what is important to 5:36you. And that's why I built different 5:37versions. I built a version for job 5:39seekers that kind of lines in on 5:40available roles and what you can get. I 5:42also built a version for folks who are 5:45looking for competitive intelligence. 5:46And that's not something we've talked 5:48about yet in this video, but as someone 5:50who has had to run competitive 5:51intelligence in the past, this this 5:53would have been a lifesaver a year ago. 5:55Like it would have been huge because all 5:58you have to do is plug this thing in and 6:00you get a full competitive readout on 6:02your competitor just based on their job 6:04listings that they have publicly shared. 6:05You you're not doing anything 6:07inappropriate. You're just looking at 6:08their job listings. And here it is. And 6:10that is one of the larger lessons that I 6:12want to call out here. We are in a world 6:15where there is an entire new class of 6:20data that was previously considered 6:23trivial data. Data that wasn't worth 6:25hiding, wasn't worth securing because 6:28nobody had the time to analyze it. It's 6:30now open season. This data is now 6:33available for analysis. It's available 6:35for strategic understanding. If you're 6:36an investor and you're trying to invest 6:38in a company, why wouldn't you run a 6:40query like this on open job postings and 6:43cross-check that against what the 6:45company's principles are telling to you? 6:47Of course, you would. That makes just 6:48perfect sense. So, this is not just 6:50something that job seekers are 6:52interested in. This is something where 6:53if you need the tea on a company, it is 6:56now easy to get. And I want you to ask 6:58yourself, what other classes of data are 7:01like that? What other classes of data 7:03out there have been trivial for a long 7:06time and we're now thinking maybe that's 7:09not trivial anymore? I'm going to give 7:10you another example and this is actually 7:11a safety tip in the age of AI. Think 7:15about when you last posted a publicly 7:19available selfie outside. And the reason 7:22I say that is because with the advent of 7:25reasoning models, especially the Chad 7:26GPT image recognition models, they are 7:29extremely good at knowing where a 7:32photograph was taken in the world. And 7:34so if you have like your Instagram feed 7:36set to public and you're taking a bunch 7:38of selfies outside, even if you don't 7:40reveal your location, your location can 7:42be inferred from that information. There 7:45are other examples as well, but I think 7:46that gives you a picture. We we are 7:48entering a world where LLMs are making a 7:51whole new class of data that would 7:53previously have been like waste data or 7:55data that nobody cared about. It's now 7:57useful. I picked job postings because I 8:00think it's one of the most useful 8:01examples of this. There are, as I've 8:04been saying, a dozen and a half ways to 8:06use this, right? You can be a job 8:07seeker. You can be a PM who's looking at 8:09competitive intelligence. You can be a 8:10sales guide looking at how to approach 8:11this company. You can be a buyer looking 8:14at whether you want to buy based on the 8:15job postings. You can be an investor. 8:17Right? There's so many different roles 8:18you can take and still find this useful. 8:20Now, one thing I want to do is make sure 8:22that I share with you how this looks. If 8:24you are not a fancy pants engineer and 8:27you do not have a perplexity API key, 8:29what does that look like? Well, actually 8:31got a couple for you. So, let me just 8:32share that quickly and I will show you 8:34what it looks like. Here we are. Same 8:37exact company, by the way. So, this is 8:39Anthropic, right? It runs the query. You 8:41can actually see the query here. I'll be 8:43sharing it in the post. You can see how 8:44it works. Um, and it goes and runs it, 8:47right? It thinks for a bit and it goes 8:48and runs it. It gives you a sense of 8:50what it looked at. It gives you a sense 8:51of the signals it pulled out. So, this 8:53is a little bit different order from 8:55what you saw in the lovable app that I 8:57showed. This is a an order that 9:01emphasizes proving how it got there. And 9:04so, these are the grounds or the inputs 9:05that it's using and it wants to show you 9:06that first. So, if you want to just 9:08scroll, you can scroll down to insights 9:09and you can see where they're investing. 9:12Uh you can see career opportunities. Uh 9:15and this one is absolutely aiming at the 9:18career side. So it it brings out more of 9:21the career piece than I have in the 9:23Lovable app. Although Lovable makes it 9:25really trivial to remix these. So when I 9:27publish this, anybody is going to be 9:29able to just remix it and make it what 9:31you want. So you can make it a career 9:32one really easily that's just about 9:33careers, not just about company 9:35intelligence. And I'll include this 9:36prompt so the career folks are going to 9:38have plenty to work with. Uh, so it has 9:40a Seattle office that's growing rapidly 9:42and an NYC hub. It talks about the comp, 9:45which is of course insane AI comp. And 9:47then it gives you the receipts to show 9:49you kind of how it's thinking about 9:50about it. And it's also talking about 9:52sort of competition, which of course 9:54like that's not surprising, but it's 9:56nice that it pulled out, right? It's 9:58nice that it showed it. Um, and it calls 10:01out automation risk. It calls out less 10:03emphasis on consumer features, so you 10:05sort of know where they're at, which 10:06aligns with what the lovable prompt 10:08found. And so this is sort of like a 10:10lens on the same company from a 10:12different camera angle where you're 10:13looking just at careers and obsessed 10:15about it. And as you can see, it's not a 10:17fancy web page, but it's lots of 10:19information you can use. And you don't 10:21have to have an API key or anything. You 10:22just run the prompt. And by the way, if 10:24you don't have Perplexity or use it, 10:26Chad GPT has its own search engine. It 10:28will also run this prompt. Let me give 10:30you one more peak. I love this one. Um, 10:34this is a company radar that's more sort 10:38of like for the product manager or 10:39someone who wants to do like an overall 10:41analysis and I think it's really cool. I 10:44think it it sort of gives you a sense of 10:46what's in the box. Let me just share it 10:48with you here. All right. So, it's going 10:50to go through and it's going to look at 10:52all the signals. It's going to prove its 10:54way forward and then it's going to get 10:55into product strategy, right? is going 10:57to talk about how it's investing in 10:59claude code, what MCP looks like as far 11:01as a moat goes, which is a really 11:03something I've been calling out is like 11:04it's sort of an engineering mode for 11:05them to build that ecosystem. Um, this 11:08one talks a little bit about how they're 11:10doing sort of B2B sales. And this one 11:12does catch sort of a sales position 11:15piece around healthcare and financial 11:17services, which the lovable prompt 11:18didn't get. If you're looking at sort of 11:20reconciling that out, what I would 11:22suggest you do is you pull all three and 11:26then you start to hybridize them and 11:28harmonize them a little bit and pull out 11:30specific insights you're looking for. 11:32It's almost like getting 3D vision, 11:33right? You get different perspectives on 11:35the same job and they're they're roughly 11:37aligned, but you get different nuances 11:39that come out. Uh you have a call out on 11:42how engineers work, which I really love. 11:45Uh you have a call out on anti- 11:47hierarchy signals which is another great 11:49one. Um and you have some interesting 11:51inferred weaknesses, right? Are there 11:53too many engineering manager positions 11:54with no teams built? Is there euro chaos 11:57because they're aqua hiring teams? Uh 11:59scaling fractures. This feels like it's 12:01really big and really fast. TPU 12:03dependency, which is frankly a really 12:05interesting piece of intelligence. Um 12:08and so I think I think this is a 12:10phenomenal overall perspective on the 12:12company. if I were in any kind of 12:14competitive intelligence, this would be 12:16really exciting for me. So, one of the 12:18things I want you to take away as you 12:19look through this is that this is not 12:21hard to do. Like, I'm going to share the 12:23prompts. I'm going to share how I worked 12:24through it, but what you should be 12:26thinking is, where is there data that I 12:28want to get a hold of that would 12:30previously have just been really hard to 12:32do? How can I get a hold of that data 12:34and make use of it? LLMs make whole new 12:37classes of data accessible and they give 12:39all of us an easier time as a result. 12:41And so if you're in product, if you're a 12:43job seeker, if you're an investor, if 12:45you're a buyer, if you're in sales, I 12:47hope this helps you imagine differently 12:49what you can do. And obviously make use 12:51of the prompts. Obviously go use the 12:53Lovable app and have fun with it. Love 12:55to see what you built. Cheers.