Soft Skills Engineering - Episode 525: How do I save my company of 500 people and my new company is insane with AI

Episode Date: August 10, 2026

In this episode, Dave and Jamison answer these questions: Hey Dave and Jamison, First, I love your show and it has now been part of my Monday evening routine for many years to listen to your podcast while cooking dinner. I recently joined a mid-sized software company (500 people) as a senior data engineer. The salary is a significant bump from my previous job and I really like the team on a personal level. Three months after I joined, the company laid off half its employees and announced a strategic pivot away from their previous core business to become an “AI-first company” and develop new products in novel, more profitable sectors. The product that my team had been working on until that point was scrapped and there are no new projects coming from the business side. In subsequent communications, it was revealed why: Management has no concrete ideas for the future strategy, neither on what products to develop nor on which sectors to target. Instead, our team (among others) is now tasked with coming up with new product ideas to “save the company”. I’m a data engineer, not a business development manager, and while I have dealt with high ambiguity from the business side in previous roles, it was never this extreme. I was looking forward to working on interesting technical challenges within a great, experienced team, but now I feel like I’m expected to become an “idea guy”, a role that doesn’t really match my personality. My question is: Should I be patient and try to turn this situation into “an opportunity for growth”, or should I start applying elsewhere? I’m a bit worried that the short length of stay at this company will leave a stain on my CV. Many thanks to my favorite space lawyers, an anonymous listener. Hi! My current job is at a 10k+ large risk-adverse company that currently has nil, none, nada, nichts ai tools available for coding. I of course did the only reasonable thing. I’ve gotten a new job at a middle-sized fintech company that’s going ”AI first”. I’m really excited about finally using ai outside hobby projects. I worry however, that my future colleagues will go too quickly ahead while vibecoding, vibearchitecting, vibetesting… leaving a mountain of tech debt behind. A lazy colleague would actually be better than an intellectually lazy colleague in this scenario. (I sincerely hope that I won’t be that coworker) Are my worries overblown or sane? Any gotch’as to avoid? Maybe you can give a brief survival guide to the brave new AI world?

it takes more than getting into a good old-fashioned programming language fight to be a great software engineer this is episode 525 of the soft skills engineering podcast where i'm your host jameson dance i'm your host dave smith soft skills engineering is a weekly advice show about all the non-technical things that go into the technical field of software development like coming up with how to craft the perfect insult to somebody who likes a different programming language than you oh just it's been kicking off a little bit on twitter the last couple weeks and it's good to see it's back is this the rust thing what is this uh yeah probably yeah it's like systems languages and yeah we haven't moved past that programming languages

may not matter for like the software you write but they do matter for the drama and the fights you start still yeah that that becomes like this niche where programming language choice only matters for social purposes yeah it was already very tribal which one you picked but now it's maybe it's only tribal that's hilarious i don't know still it's good good to see some things never change yeah uh dave do you want to thank our page the only the only thing that changes is we used to think it mattered oh yes yeah now it's just recreational yeah yeah now now we know it's just for fun okay all right yeah let's uh let's thank our patrons who are contributing at a level where we say their patreon profile name every single week and i just have to say just about

all of you have changed your patreon name for this podcast and i just really appreciate it so freaking funny now i'm i'm starting to admire the stalwart people who just really are secure in themselves and say i see this group of people changing their names and no i am jenny kim yeah forever i appreciate you jenny kim is the only one with a prince with principles you can't be moved maybe just insulted most of our patrons shoot after years of encouraging this we we realized that you're all just going with the crowd anyway no we love it i love all varieties i wouldn't change the thing of it okay here we go they are i quit my job now i am broke but happy thanks for the advice

will how to wrap battle with space soft skills engineering podcast my em agent put my coding agent on a pip joel harks generate episode of the soft skills engineering podcast make no mistakes tim this week's enigma wheel setting is kxl pad skip six i love it who even knows what their actual names are my actual name on linkedin is yammy debugging in the dark candy token limit reached jacob chandling a tingle frozen barrel bees this will angel copy crab deeply appreciates will angel's decision to release his name into the public domain move over will angel we're doing space names now nick molyneux embedded engineers treat assembly the same way typescript engineers treat javascript javier gonzalez chewy had a baby soft skillet ng exclamation mark and then we have

a pregnant emoji i think i think yeah and then salted fries with an open paren way to go chewy congrats jameson's wait what is ng i don't know okay angular i don't know an angular baby yeah jameson's nordic country corner iceland is the one with the volcanoes is it what did we have last week i've already forgotten yeah this is a new one yeah i love this let's keep the nordic country corner going we have got to learn these countries okay next up won't angel dan from drone deploy never is not just a creator of mars flamingo emoji i like chicken i like liver miyamix miyamix please deliver i can see dodds kyle boss the person hearing this can do very hard things that's true kent c dodds oscar delgado listens

to soft skills engineering while will angel listens to 10 junk miles jenny kim the stochastic parrot ira chan jonathan king zanai beautiful functional user documentation adrian at kappa dot ai angelica cath or will angel official approves of all names but may raise an eyebrow at ragnar brayden canes john grant britney ellick cats are the best people close parenthesis followed by a literal close parenthesis close parenthesis followed by a literal close parenthesis oh the humor is multiple layers deep what a sweet treat what a delight all right you want to read our first question yes i do hey dave and jameson first i love your show and it has now been part of my monday evening routine for many years to listen to your podcast while cooking dinner oh

nice that's awesome i recently joined a mid-sized software company 500 people as a senior data engineer the salary is a significant bump from my previous job and i really like the team on a personal level three months after i joined the company laid off half its employees and announced a strategic pivot away from their previous core business to become quote an ai first company and develop new products in more profitable, novel sectors. The product that my team has been working on until that point was scrapped and there are no new projects coming from the business side.

In subsequent communications, it was revealed why. Management had no concrete ideas for the future strategy, neither on what products to develop nor which sectors to target. Instead, our team, among others, is now asked with coming up with new product ideas to, quote, save the company. I'm a data engineer, not a business development manager,

and while I have dealt with high ambiguity from the business side in previous roles, it was never this extreme i was looking forward to working on interesting technical challenges within a great experienced team but now i feel like i'm expected to become an idea guy a role which doesn't really match my personality my question is should i be patient and try and turn this situation into an opportunity for growth or should i start applying elsewhere i'm a bit worried that the short length of stay at this company will leave a stain on my cv thanks to my favorite space lawyers and anonymous listener wow that's amazing i was about to congratulate you on getting

a job for a much higher salary. And then either this company is utterly doomed because they have no idea what they're doing, or they're in a lot of trouble because maybe they feel like they have an idea what they're doing and they've completely failed to communicate that to the people on the ground. That people think that their job is to invent the next product that the company's going to build yeah i would be surprised if the ceo or i don't know some vp felt like we're just waiting for the senior data science person to come up with a product idea probably they think they have a different idea but that's also a bad solution that or sign oh my goodness oh man okay we need to pivot the business quick call the data scientists and the data engineers get me on the

phone with our data engineering team i want the one that we hired three months ago only that one yeah yeah we laid off the other ones we need new fresh eyes i think you it's on your obligation is to propose data science-based products say hey you know it'd be great if we had some kind of pipeline for managing different different transforms of the etl flow yeah i think that's a product we should build yeah and good news i know everything there is to know about it yeah yeah this is wild usually they if the if the ship is sinking they try and hide it a little bit better than this but this if if what they have said is true hey like come up with ideas to save us yeah well i mean it's hard to hide that things are going badly when you just did a 50% layoff

yeah that's uh that's rough it's also amazing that the person who joined three months ago is still employed after a 50% layoff yeah and for a significant salary bump either you were just outrageously underpaid and so you don't stick out as highly paid here or i don't know they just they got well the problem was all the old people we had here yeah that's the problem we just need new people new people new product that'll fix us i don't know would you be patient with this jameson or would you be like oh boy time to heck no i would 100 start to look around would you be worried about a stain on your cv as our question asker said also heck no this is a great story to tell about why it's fine i joined the company they laid off 50 of the people and said hey

come up with a new product idea and i don't want to do that i want to do great data engineering work but i did come up with some product ideas yeah like i i don't i don't know do you need a data engineer good news i would like to do that for you not be a product manager yeah exactly yeah i think this story this story flows very naturally it won't be a stain we've been we've actually been saying that now for the last five years um six years well we've just had excuse after excuse after excuse for why a short stay at a company is going to look normal for a long time Especially if it's one. If it's one, if you've got a long string of short stays,

then it looks more like a you problem. Maybe even one or two. I'm just thinking back on my mind. I'm like, okay, we had COVID hit. Tons of people changed jobs right when COVID hit because of layoffs and market conditions and things. And then even over the next couple of years, people would be crazy not to job hop a little

because salary increases were going off the charts. You're like, oh, I quit because I doubled my pay. Wouldn't you do the same? and they're like yeah i did the same thing you know it's like you're the person who's interviewing you is like oh why were you at this company so short well i doubled my salary to leave and they were like yeah me too yeah i also had a one-year stint there and and now it's like ai okay ai you know launched ai chat gpt came out in the in the public attention in november of 2022 four years ago almost but over the next few years now we're seeing tons of disruption from ai so it's like oh

yeah i definitely have a good reason for why i've changed jobs so much so i'm not i'm not worried about the stain one bit yeah it does feel like every company's fear right now is that uh we have already been or will soon be disrupted by some ai magic so you're probably not the only one in this situation but i i wouldn't quit i would definitely stay and be looking and not consider this a long term thing. Yeah, me too. In fact, I'll tell, I'll go one further. I would say if I'm interviewing someone today and they've had the same job from like 2019 to today, and they're not an executive and they're, you know, they're like a individual contributor, data engineer, software engineer, I would be like, wow, why did you not change jobs in that seven year period?

Yeah. It's because the world has changed so much. yeah i'm just like i feel like you missed opportunities or you got crazy lucky and you just super value stability yeah i don't know it it tells a story either way and it's not always good to show that you've been at the same company for a long time during that period yeah i i just want to talk to this person i'm so fascinated what about what your day-to-day is like what the company does like i yeah maybe they've still got stuff to operate day-to-day you know like oh well got to keep the data pipelines running yeah i feel like the quote idea guys you have a lot of leeway now to be very eccentric eccentric however you say that word you got to take lots of walks

you got to do a lot of stuff that doesn't look like work yeah because that's where the ideas come from that's right yeah you don't get ideas churning out tickets no no no no turns out my best idea generation source is all of my favorite hobbies right golfing yeah hiking you know playing grinding league of legends turns out to be a great way to come up with sass b2b sass ideas just non-stop video gaming yeah creative energy yeah i i don't know this feels pretty straightforward that you might be part of a layoff i don't think this company sounds very stable i don't know maybe you just try your hand and see what it feels like to come up with ideas but i would not expect it to stick around and i would plan on finding something else pretty soon yeah i totally would

too and just remember the best business ideas always come immediately after a pentakill what if we could replicate this feeling when you got a spreadsheet really in a good state that's my idea okay dave do you want to read our next question yeah let's take a look here this comes from an anonymous listener who says hi my current job is at a 10 000 plus person large risk averse company that currently has nil none nada nicked ai tools available for coding i of course did the only reasonable thing i've gotten a new job at a middle-sized fintech company that's going, quote, AI first. I'm really excited about finally using AI

outside of hobby projects. I worry, however, that my current colleagues will go too quickly ahead while vibe coding, vibe architecting, vibe testing, leaving a mountain of tech debt behind. A lazy colleague would actually be better than an intellectually lazy colleague in this scenario. I sincerely hope that I won't be that coworker.

Are my worries overblown or sane? Any gotchas to avoid? Maybe you can give a brief survival guide to the brave new ai world was brave new world a phrase before that book oh man now you're really stretching my my knowledge huxley whatever his name is yeah because that book has some disturbing connotations the brave new ai world does sound pretty dystopian if i think a lot about that book are my worries overblown or sane no no your worries are very sane oh uh and okay hold up yes i just have to interrupt you what up brave new world is you're you're gonna be like

of course it is it's shakespeare huh at least it's in shakespeare okay now does that mean that it actually had anything to do with it i don't know someone's gonna have to write it and give us a full rundown on this and we're gonna have to add to our nordic country corner we're gonna have to add what's the word uh entomology etymology is it etymology or entomology one is bugs one is words i can never remember which one i think it's etymology but i now don't know and i choose not to look at it we're gonna need an etymology corner every week here but yeah anyway there you go that's one possible explanation yeah huh my colleagues will go too quickly ahead while vibe coding by protecting by testing yep absolutely they will if your company's going ai

first there will definitely be people there who are just firing as many tokens out of the slop canon as they possibly can yeah come what may and it is a brave new world you cannot deal with code at that volume like you would deal with human authored code it's just impossible you can't read it all you can't comprehend it all and do anything else with your time besides like reconstruct what wild stuff the machines have done yeah i don't think everyone naturally will be like that but it's certainly the the more the eye pill your company is the more likely it is to be there yeah so are the worries overblown you think uh leave a mountain of tech debt behind i mean i think the ai optimist position on that is and smarter models will solve the tech exactly

models better harnesses it's it's like like the uh the cowboy coder would say and and we'll have enough money to solve this problem later or someone else will clean up after this or now that's been replaced with like and the next i don't know anthropic release we'll be able to deal with this mess instead that's right that's an opus six problem yeah yeah is it true i don't know but it is a reasonable bet to make i think that it'll just keep getting better faster than you can destroy your productivity i'll tell you what i would say that as long as you are joining a team that is full of pretty skilled engineering practitioners, you're going to be okay. But if they have this idea that they can take AI and without a knowledge of what they're building

or without experience working in that domain, that they can get good results, they will be gravely mistaken. And I will tell you this because I recently ventured into a new domain and I allowed AI to guide me heavily in that domain. And that domain was some video camera and electronics and cooling systems and voltages and wiring and soldering and even some mechanical engineering for a little hobby project i was working on and i just have to tell you ai guided me wrong on every critical decision point and i just didn't know until i saw smoke coming out of a cooling fan or you know saw my video processing unit failed to power up for the seventh time you know and it was just it was horrible it was a horrible experience and i'm literally buying parts

online and having them shipped to me and then i'm frying them because ai told me the wrong thing like who knew you could fry a cooling fan by undervolting it huh well you can you know and uh ai didn't know that but then when i told it what happened it was like oh yeah of course that happened yeah like this is horrible so on the one hand it is totally a super effective tool for taking a skilled practitioner and just making them so much more productive they can crank out solution they can crank out so much code but on the other hand as an architecture guide it can be horrible and it can really mislead you very confidently and so i think the answer to your question is it will depend entirely on the skills of those who are wielding it it is a lot better at

known unknowns than unknown unknowns which is why i think it can be decent at code review and catch stuff that humans miss because it has stuff to look at it's it's yeah the space it is searching through is like something that exists already and it can check up on that. That's a very, very good distinction I've noticed as well. And I guess I don't want to turn this too much into on how do we use AI tools effectively, but I think there's a huge argument to be made that AI is very effective when you give it the entire context of the problem and ask it to interpret that context rather than saying, you know, use all the knowledge you already have baked into you to generate an answer or generate an idea that idea is very likely to be bad if you

do it that way hmm i think we just have found the solution to the first question though i mean obviously you must use ai to just be the most prolific ideas guy yeah there are some people that i've seen that how do i put it every every technological age feels like there's a shape of person who the strengths and weaknesses of that technology fits really well. And with AI being able to write code, that shape is different than, I don't know, cloud computing or mobile phone software or whatever, I don't know, whatever way you want to break it down. And there is a new shape of person that can do really cool things. Someone who can creatively come up with ideas and coherently express them, but maybe wouldn't have been able to do them now has a way to try a bunch

so it is cool to see that uh that shape is not often a really great software engineer they're kind of in the in the old world those are different shapes i think very meticulous kind of detail-oriented logical sort of uh abstract thinking systems thinking debugging a huge amount of patience and stamina for digging into a problem for hours yeah it is a tough transition and it can feel really bad what is my i don't know i just have a dumb a bunch of dumb generic advice i go insane if i try to keep up too much with the right way to do things there's just infinity new workflows popping up all the time people trying to become ai influencers by oh it's horrible usually using ai to write posts about how revolutionary this new thing is or whatever but i just feel

anxious like oh people are further ahead of me every second i'm not just absorbing this quickly and radically changing everything and then i decide not to and then i feel better yeah yeah so i think for me the healthiest thing is to just be i don't know at least a few weeks behind and let some of it shake out and then filter it a bit can't believe we're talking about how a few weeks makes a difference now yeah it's incredible that is true i don't know be okay with being behind the cutting edge but if you haven't done any of it at work then i guess you're already you're right behind the cutting edge yeah super fine so but you'll catch up fast yeah any gotchas to avoid like here's here's one gotcha to avoid the old concerns about code quality are very different

now you know and i've tried to ask some of my team members this question and i noticed that they get a little uncomfortable with it but they'll say things like oh we're we're not consistently using this pattern, or rather, we didn't keep the code dry. And so we've repeated some code blocks here and in a few places. And I have to ask them, does it matter? Well, it used to matter because we struggled to find those things and fix them. And so we would fix a bug once, but we didn't actually fully fix the bug, you know, and because it was actually, there was code repeated elsewhere. And I used to obsess about that. But that now I'm like, AI just kind of figures it out. Does it matter? And if you are just now jumping into AI coding, I think what you

will find is that a lot of the things that used to matter just kind of don't anymore, or they matter in new ways. So question all of your assumptions, like the stuff that's been baked into you for years, these habits and the things that served you so well. This is the first time in my 20 plus year career where I've really had to question those things and say, am I actually just raising concerns that just don't matter. I'll give you one example. We have we purchased a front end component that handles very sophisticated tables. And it's awesome. We use it in our front end. And we were approaching a redesign of a part of our app. And we said, hey, let's use this new newly purchased table thing. Well, it didn't have a couple of the features that we needed.

And so we were like, dang, that's going to be horrible. We still have to use the table because obviously we don't want to re-implement the whole table thing but the engineer who got in charge got put in on the project he just went ahead and re-implemented the whole table thing and it's been fine we're like whoa it actually does all the things we originally wanted and it handles all the weird table stuff that we were worried about that we thought we couldn't re-implement and it's fine and that i think that was a moment for me where i said the things i used to care about like we could never spend the time to re-implement our own advanced table system. Well, maybe you can, and maybe it's just going to be okay. I think that's my biggest struggle right now is

I feel like my sense of what's going to be hard to change later is still not attuned to reality right now. And I think generally that feels like I'm spending time on stuff that leaves the design quote better leaves it easier to change easier to reason about for humans and i i think that still pays off for ai maybe but the ratios are different right yeah i do think i don't know a horrible code base with tons of spooky action at a distance that's really hard to follow i think it's going to be easier for ai to break that code yes but how much easier i i don't know and what if you just have ai surround it with unit tests and integration tests and then you're like all right go nuts inside there as long as the tests all pass you know yeah yeah it does

it just feels different and i don't think i'm comfortable with where i am right now yeah i for sure i'm not and i think that's in part that's a that's a wise instinct i think that you have jameson because we actually don't know how good ai is going to be at handling these problems you know like for example like years and years ago once even even i am young enough to not really remember the days where we questioned the output of our compilers. But the generation before me certainly did. They cared about that. And my generation stopped caring about it completely, and we never looked back. You know, once in a very, very rare moment, even when I was working on C++ 20 years ago, I would look into the compiler output code, the assembly code. Very,

very rare. And I think that the generation before me would have been like, wow, you definitely shouldn't write your C++ like that because that's going to lead to unoptimized assembly output from the compiler. You really have to hint it this way or that way. And they were just wrong. They just were wrong almost all the time. And so I'm a little hesitant to say that we're in a moment like that now because we haven't had enough time. This is why I'm saying you're wise to feel that way because we just aren't quite there in terms of time elapsed from having made these decisions to see what the real consequences are going to be. It does feel similarly shaped. I mean, obviously compilers are much more deterministic than whatever stuff you put in to get ai written

code out but i i've heard that feeling before and i think you can just like there are people who care a lot about assembly still and care a lot about really high performance but it's a niche maybe maybe that's where it's going i hope not but man i don't know anybody who says they know is probably wrong oh yeah they're wrong um or has a billion dollars because there's some they've got a conflict of interest? Lab super genius. Yeah, I don't know. That's the thing, right? Nobody knows. I mean, look at how much our world, the world of day-to-day software development has changed over the last, even just the last year. I mean, 12 months ago, most people were not doing coding harnesses like Cloud Code, literally 12 months ago. And now most people are. So we've changed

the way we write code completely in 12 months. So how could anyone be so arrogant as to say they understand what things are going to be like in a few years time from now, let alone another 12 months i do remember this very when when context windows were starting to get big we didn't realize that search and kind of command line stuff could help a model get context there was this moment where somebody said hey you can just put your whole code base into context if it's small enough and then it'll just know and it'll give you good results like we uh it didn't spoiler alert because models do worse as context fills up but even that was i don't know a year and a half ago maybe somewhere around there which feels like an eternity so things like very core critical

parts of the workflow are still wildly up for grabs yeah i just feel so dumb saying all this generic stuff verification is really hard now when you have when production is easy verification becomes a new bottleneck because it's you don't have to wait for an engineer to carefully build the thing you just can have the option to say robot do this for me and then you got to test it more and it's a more surface area to test because you got more stuff being built and then there's more combinatorial complexity of how these things interact that feels like one of the big new bottlenecks is just how do you know it's good now when there's so much that you cannot just manually carefully review it all yep very true and the answer is oh look at the time

we're at the end of the show have we answered the question i mean my my general answer is you're going to have a great time i think it's going to be one of the coolest and weirdest overnight switches you've ever experienced in your life in your professional life you're going to go from writing code painstakingly one character at a time with the occasional tab complete to ai generated code to a fully ai built company and it's going to be interesting seeing how things go down yeah and there will definitely be some downsides but it'll be really really cool the times i've felt like i've learned and grown the most have all been me being in the least familiar circumstances and this is certainly one axis that will make a lot of things unfamiliar so totally it'll be cool

good luck good luck good luck to all of us yeah i know what's gonna happen in the next year yeah what can people do if they want their own questions answered go to softskills.audio and click the ask a question button where you can fill out our little form tell us as much or as little about yourself as you'd like and thank you for sending in your questions we read them we love them we answer a very small percentage of them but we commit to answering all of them eventually yep thank you for listening we will catch you next week