One thing I haven't named today, but I should, is that people hate AI. [laughter] Fears that students have about how their future jobs and careers might be affected are completely justified. The possibility of massive concentration of power and wealth in a small number of companies that are not effectively governed — we need to pay a lot of attention to that. [applause]
I am the sort of person who in my own life I generally am willing to share a fair amount of personal data in order to get like a better and more personalized experience. If I were a student at a university and I was sharing a bunch of information and in return I had an agent that I could use to let me navigate a vast bureaucracy without ever having to make a phone call, I would be very excited about that. I would just ask in the college not, you know, when I graduate sell that profile to Palantir so that I could later be targeted in a drone strike.
I think Silicon Valley knows that higher ed is a big like vertical for them to sell into. Anthropic is selling into higher ed. Google is selling into higher ed.
Yeah, absolutely. I mean like the reason that they're coming to you is they want to sell you the big contract because the money is great but also because they want to create lifelong customers and they feel like, well, we get a student using ChatGPT their freshman year of college, there's a great chance that they'll buy a paying subscription once they enter the adult world.
Do you think that the AI companies are actually going to be approaching more and more colleges for like better placement?
Hello Vineyard. Good morning. Hello humans. Good morning. You're here for the Vine Down Live from the Vineyard live podcast this morning. I'm joined by Casey Newton. Casey, why don't you come on out? We'll tell the people about you.
All right. Hey, everybody. [applause]
Casey [clears throat] is a well-known tech journalist and founder of Platformer. Um, Casey's like a pipe-hitting tech journalist. And I know we don't like to hear from like the outside in this room, but I'm going to force you to expand your minds. And not like in a psychedelics way, but like in a, uh, in a the news from the outside.
And um Casey has a podcast himself. It's called Hard Fork and a podcast called Platformer. Casey really is talking a lot about to tech CEOs and people who are far more well-paid than we are about the future of work, the future of technology and work. And I thought it'd be a really interesting conversation to have with Casey about students and higher education. Casey did a lot of cramming on higher education last night.
Good student. Good student. And I think that here at the Vineyard, the audience has spent the last couple of days like getting their minds around what they're going to go home and do next. Like they're making plans. They're like going to jump on their planes. I picture you all like getting into the seats of your planes and just being like, 'What do I do now?' and getting back to your campus and going like, 'Okay, I'm supposed to get into action on something.' And I'm curious, Casey, if you can kind of temper this energy with what we talk about as far as the AI productivity paradox, like how should they manage their energy in their next chapter?
Yeah. So, um, earlier this year, I wrote about this productivity paradox, which at its root is basically like executives tend to be much more excited about AI than rank and file workers. And that when you do surveys of CEOs, they will say, 'I had the most productive weekend of my entire life. I rebuilt every system from scratch using Claude Code. Let's go, go, go.' And then you talk to the average worker at a company or organization like yours and say, 'How's the AI going for you?' And they say like, 'It's fine.' You know, I'm tired of hearing about it, right? And so the paradox, like what is actually going on there? And there are a few things. One is that AI often gives you the feeling of being productive without actually making you more productive.
Oh, because you're just spending time being busy with it.
Yes. And so, you know, as you go home to your organizations, this is something that I would keep in mind. You know, I think it's cool to be excited about, you know, bringing technological innovation to your organization, but I think it's also good to have in the back of your mind. Uh, you know, you might actually just be spinning your wheels a bit.
So, what do you think? How much of that is like things that you need to do to manage sort of burnout and actually test for productivity? Like how do you know when you're in that space of like spinning your wheels versus being productive?
Yeah. So I always encourage people to try to boil things down to the level of the individual workflow, you know, like in Silicon Valley, we have again a lot of these CEOs who go nuts over the weekend like vibe coding uh new platforms. Um and this strategy of like trying to AI all the things I find just like doesn't generally give people great results. On the other hand, if you say, 'Hey, like we want to see if we can use an AI agent to help students uh fill out all of their financial aid forms, like answer all their questions or, you know, we're getting this huge volume of inquiries from students about this issue. Can we build an agent that reduces the number of times we have to connect them with a real person in the office?' Um, that's something that might actually save you time and money. So, that's where I encourage people to focus.
Okay. So you're kind of underlining something that I've had a bunch of conversations with in the last year is like something happened and this is like for the CIOs in the room like something happened in the last like 2 years to me where two years ago I would talk to anybody within like an IT organization at a school and they would say like yes we're going to have like a requirements gathering session and we're going to like solve a business problem by deploying a thing or like by buying a thing or whatever and like AI happened and people like forgot about requirements gathering. They're just like put AI on top of it. Like it just it feels like really like we forgot to like stop to be like what is our problem? Like what are we diagnosing?
That's the thing about that is so crazy about artificial intelligence is that it is this general purpose technology. And so I think for the past couple of years we have understandably been in this exploration space of exactly what can it do? And I think that's important honestly because its capabilities probably do exceed what it is doing for you at the moment. The only way that you're going to figure that out is by kind of like banging your head up against the wall. But to your point, it is good to gather those requirements and try to build really specific workflows, right? Instead of just like applying on top of it.
But I think so some of the something someone said to me that was interesting about what we were talking about yesterday and we had sort of a solutions showcase. We were talking through all of the different use cases of this. I was talking to a leader from an institution who said like, 'Look, I just watched this vendor go down, that vendor go down, this contract go down, that contract go down, and like in my CollegeVine sales heart, I was like, yeah.' And in my like I was like, 'Wait, what?' Like wait like maybe like so one of the things you talk about and have been talking about lately on your show is kind of the SaaS apocalypse which is sort of like the contracting of the constellation of the vendor sphere and like the different things that a company in your world, you know, this like high exposure to high tech in Silicon Valley, but like where should we think about the things that become sort of the point solutions, the SaaS contracts that we maintain versus the stuff that we build to be so bespoke and so small for us.
Yeah. I mean, I think there is real promise here that at some point uh your organization may be able to build some of its own things and spend less money on them. Like I do expect an outcome of this AI moment is that costs are going to go down. Um or just like at a minimum the structure of these contracts is going to change. Right now a lot of software as a service gets sold on this per seat basis. You know, we're going to, you know, buy a thousand seats of this. Um, what I'm seeing in Silicon Valley is more of these contracts are moving to like an outcome-based pricing. Um, you know, so we'll charge you, you know, some sort of money up front to deliver like this kind of service. And the overall effect of that is often to like drive these prices down. So, you know, because these agents can work 24/7, the prices are just maybe going to come down.
So, this is a group that actually negotiates contracts with their SaaS vendors. Like what do you think is coming in terms of if the pricing structure is going to more outcomes-based pricing like and that's something that we will see, you know, within the next year in this industry as contracts come up for renewal but like what should this group be prepared for in terms of like being able to negotiate against that or be in a favorable position to know that that's coming?
Sure. Well, I mean, I think you're just going to be able to draw on your own experience, right? Like I assume that most of the folks in this room have used something like a ChatGPT or a Claude. You know how quickly it can generate a document, a slide deck, a spreadsheet. And so to the extent that you're working with vendors who are doing those things on your behalf, you just know they can do them faster now. You know that they can do them with fewer people. And so that's a tool that you bring into your negotiation to say like, look, the world has changed a little bit here and that needs to show up in the contract. People are already doing this with great success in Silicon Valley.
So if you think about that sort of like contracting of the vendor sphere and those decisions you're starting to make about like what do we build versus what do we maintain on the SaaS contract, what do you contact side of things like what do you think companies are thinking about that we should learn from in this industry just as far as the things that you maintain like what are when you talk to folks in companies like what are the SaaS relationships that they will continue to renew versus the ones that they're just going to build like what's the break what's the water line?
I mean like the canonical example is like in Silicon Valley, everyone would love to no longer use Salesforce ever again.
I think you have a couple people in this room who no longer...
Extend here to the higher education.
The higher Salesforce did like a big higher ed move and some of you all are still like under that and that's we support you and that's when I talked yesterday about like the tears that I saw it was mostly in that chapter of my career. So yeah, you got some folks here who feel that.
There the CEOs wake up every day in Silicon Valley asking, 'What if I never looked at Salesforce ever again?' And Salesforce, you know, for all the great things that I'm sure it can do is a fancy database. And turns out you can use a chatbot to build a database. And so you have lots of startups that are trying to, you know, build a better version of Salesforce, but you also just have some companies now, particularly the smaller ones, that are saying, could we build our own CRM? You know, could we build our own sort of fancy database? You know, these are things that we should be approached with caution, right? Like I'm much more comfortable with like a five-person uh startup trying to build its own version of Salesforce than some like amazing, you know, regional college that has like thousands of students and is, you know, is regulated in a way that a startup is not. So, you know, I don't want to like talk past that, but is this a good time to have a little bit of curiosity about like what elements of our might we be able to build ourselves over the next couple years? I think it's just the right time to start asking the question. I think there's, you know, sort of like little pilots that you could do, maybe even have some fun.
Can you educate this group a little bit on some of the thinking around like what happens when the UI on top of like a Salesforce because this is a group that like lives and dies by their CRM or their ERP systems and like what happens if you sort of just say like okay that's the filing cabinet like that's the system of record like we have to know where it is the data is in a filing cabinet but like what happens when you don't need the UI like necessarily need the UI on top of that anymore?
Yeah, I mean so people uh call these systems headless. Um, little disturbing. I didn't pick the name, but the basic idea [laughter] is that AI systems can just make the vast bureaucracy of higher education, government, you name it, more navigable, right? And so whereas before I did have to really understand the user face uh very well. I had to know where all those features were buried, now I just sort of type what I want into a box and I get it. And you know, this is really exciting. I think it comes with some of its own challenges, right? Like a lot of students if you present them with that sort of box aren't going to know exactly, you know, what features are buried within it. But in general, having systems that you can navigate without that super complicated UI, I think is pretty exciting.
Well, it allows you to do a lot like on top of it without the UI, right? Because you're not constrained by like table order or the appearance of things or you have to go here to get that thing but it like flattens it out so much that it opens you up to so much possibility that like you could also get frozen in the options.
Absolutely. And by the way, if you want to experiment with this, I encourage everyone to build a personal website. You don't have to put it on the internet if you don't want.
GeoCities, what are you talking about?
Yeah. I mean, no, I'm even going to say forget GeoCities. You know, it's been 30 years. We can move on.
I can't. You can't move on from GeoCities.
Here's the thing. So, you can use something like a Claude Code or a chatbot and you can have it build a personal website and you can say I want it to look like a GeoCities web page from 1996 and it can be hosted entirely within your browser. So, again, if you're shy, this doesn't have to be on the internet, but the reason that I think that you might want to consider doing this is it's going to uh illuminate everything that we've just been talking about where like you can see what it means to just type into a box and get a web page out of it. When I did this for myself, this is when I started to um feel like I was understanding the AI moment a little bit better because I am not a technical person. I do not know how to write code, but I do know how to type into a box. And the amount of things you can now accomplish by typing into a box is really crazy.
We bring you the experts here at CollegeVine.
Yeah, he knows how to type into a box.
That's right. So, okay. So when you think about sort of again the consequences of this work whether it's the consequences of this work sort of technically and functionally but also like personally and psychologically one of the things that we have talked about you and I is sort of around like how social content and you have written a lot about this over many years about how social content and platforms actually like influence and shape mental health particularly around young people. And when we talk about the ways in which sort of like a traditional age student population is going to be consuming more and more sort of AI content, AI generated things like what do you think the administrators in this room who really care about like student mental health crisis because like the kids are not okay right now like we know this and we want them to be but like what should we see coming in that vein?
Sure. Sure. So I mean the situation that really struck me here was that sort of already like uh the majority of children are interacting with AI companions on a regular basis, right? Like this is not a niche use case. This is not limited to the early adopters. And so I think that by the time that students are showing up to college over the next several years, we should just assume that this has been part of their lives for a really long time. And I think it's very much a mixed blessing. There are a lot of kids who are hurting and who need support and if they're able to find some of that support on ChatGPT I think that is a good thing. Certainly I prefer to the alternative of them getting no support whatsoever. At the same time if this is their primary means for sort of you know getting emotional support and moving through the world obviously I worry that they're not going to develop the skills that they need to just kind of you know navigate real world situations. Um, there are also just a lot of concerns I have with the way that the chat bots themselves are designed to be sort of sycophantic and to tell you, you know, that every idea that you have is a good one. So, it's a really kind of dangerous area. And yet, I do stop short of saying like get rid of all of it because again, there's just a lot of kids out there who need support and aren't getting it.
So, like some is better than none or something is better than nothing with sort of be aware caveats.
Yeah. Yes. Absolutely. And like, you know, I'm sure lots of people in this room have read the stories about kids, you know, with violent impulses who find themselves like encouraged all the way down the path to doing something terrible with these chat bots. And like that's a very real dynamic here as well. So, you know, there isn't sort of one clear answer about what to do here. But I do think that it has to begin with just acknowledging how shockingly common this is for students already.
Yeah, I think this audience definitely cares about and like thinks a lot about like the mindset that we or a question I often get when I'm talking to our partners is just around like well what happens when an agent gets a call from a student at 10:00 at night talking about like self harm or like a mental health crisis or you know violent impulses like how do we actually root that and how do we link together the systems and the people such that like attention is given with the right urgency like in that moment. So like how do you think organizations should think about those structures in that plumbing in order to like maintain that safety?
Yeah. So I mean my hope would be that that student has immediate access to a person, right? Like this doesn't feel like a situation where we want them navigating a phone tree. Um at the same time as we do move to a world where students are learning the behavior that I can type into a box and get anything I want, that might be one of the things that they are seeking. And so I think that just becomes a case where the moment that becomes available and you're telling students, hey, if you are in crisis, like use this system, that needs to be like 100% accurate and that needs to deliver them to a human being as close to instantly as you can get them.
How do you think companies think about like high stakes? Because I think we in a very insular way in higher education think about like the things that are high stakes, right? Like the student health mental crisis things um and the sort of safety things, right? Because everybody who is sort of selling higher education really has to make this promise to students and families that they're going to be like healthy, happy, safe at their institutions. And I think companies have different requirements of them depending on the industry. And I'm curious from the like corporate lens like what do you think like the corporate analog of that is like the things that feel high stakes to companies that we like on the outside of corporate should be aware of?
I mean, at the risk of overstatement, like what feels high stakes to a company is like, how much money am I making? How many users do I have? And like everything else is secondary, right? So, like unfortunately, we cannot trust them to take care of your students. Like, you care way more about your students than they ever will. And the things that I spend a lot of time writing and podcasting about on the safety front like around mental health within the companies these are perceived as sort of unfortunate edge cases um it's something that you know we're going to like hire people to work on so the CEO never has to think about it and so you know while certainly their representatives will tell you oh of course you know we take this very seriously this is like a high priority for us in my experience they're often not willing to make the compromise of making the product safer if it would trade off against how fast uh the product might grow or how much money they might make.
Okay, I love that. That's like the second key takeaway here. The first one was like don't forget to gather your requirements. Second is like watch your own ass because like nobody else is going to do it.
Yeah. Yeah. Absolutely. Like and I hate to sound that cynical, but I...
No, I mean I think that's great because I think like I think part of the point of this event is for people to like come and understand like how to think about this stuff. And I think there is such like a curiosity about like our company and I think a view that like CollegeVine is like the leading edge of this technology but also like a really high requirement at this moment to be like educated about like everybody else who's coming at them for you know across all of the vendor sphere across all of the tech sphere. Like I think Silicon Valley knows that like higher ed is a big like vertical for them to sell into. Yeah. And I think like this audience definitely has had over the years like it used to be you just would get sales outreach from like the edtech companies and now you're getting outreach from like all types of tech and like you know Anthropic is selling into higher ed, Google is selling into higher ed and I think like that again sort of like that you have to maintain like hold your own frame there for your institution like matters.
Absolutely. I mean like they're and you guys all know this, but the reason that they're coming to you is they want to sell you the big contract because the money is great, but also because they want to create lifelong customers and they feel like, well, if you know, we get a student using ChatGPT their freshman year of college, there's a great chance that they'll buy a paying subscription once they enter the adult world. So that is in my experience is where the focus tends to be.
Yeah. I actually don't know that we focus on that long tail stuff. Like I think we just sort of feel like the ways in which companies sell into higher education but I don't know that like we know that that's their like motivation is to like for them to create users like I think it's implicit but not overt.
