When an AI agent replaces a human at the keyboard, who's accountable for what it touches? Adam Blue and Chris Struttmann of ALTR explore why identity—not just policy documents—is the foundation of trustworthy AI in banking, and why organizations without strong identity systems are already starting behind.
Q2's Helix and ALTR case study
Adam Blue
Hey everybody. Welcome to Cut to Context for this week. I'm Adam Blue and I'm here with Chris Struttmann, who is founder and CTO of ALTR. Thanks for joining us today, Chris.
Chris Struttmann
Well, thank you for having me, Adam. I saw some of the initial episodes come out and I really liked the format and I think you and I discussed, I know the customers really like hearing from you and you have a really special relationship with the Q2 customers. So I think this is a great little show you've got here.
Adam Blue
Thanks. Thanks. And thanks for being on today for sure. So we met in 2018, I think.
Chris Struttmann
2018.
Adam Blue
And you guys were just getting started at ALTR and we partnered with you on solving some pretty exciting compliance and data privacy and regulatory issues. It's really, really been great over the years. And so I'm super curious to see what you have to say and what you think about where things are going with AI.
Now interestingly enough, in another surprising dispatch from arsonists selling fire extinguishers, Dario Amodei then rotated back last week to saying, "Well, maybe we shouldn't go so fast with AI and we'd like to go slower at Anthropic for the good of all humankind, but only if everybody else slows down too, because I don't want to give up any advantages."
Chris Struttmann
Correct. Yeah. It's a little bit of wild times. It's almost like they're gaslighting an industry out there. It's hysterical to watch from the sidelines in my opinion. I mean, we're participants in the AI ecosystem and we enable customers to use AI all the time. I know you guys have some AI initiatives going on as well, but the public theatrics are getting to be a little bit spectacular.
Adam Blue
Yeah. I don't know. I can appreciate being a head of one of the two hottest, fastest-growing, largest-revenue, massively overinvested companies in the space, how stressful that must be. But man, pick a lane, bro, like figure it out.
Chris Struttmann
Massively overinvested with no clear path to how we're going to fulfill demand over the next 10 years. What are we doing here? The answer is literally unknown. It's kind of wild.
Adam Blue
It's fine. That is, as we say in our business, an implementation detail.
Chris Struttmann
Yes, exactly.
Adam Blue
So that'll get worked out. All right. So here's a question that I think is interesting that I think you'll have a fantastic take on. We've spent our entire careers in technology, working on tech, trying to use it to make things better, solve problems, but there was always a human being on the other end of the keyboard. And that's a double-edged sword, having dealt with human beings on the other ends of keyboards, but at least we knew who it was.
Chris Struttmann
Pros and cons. Yep.
Adam Blue
We knew what it was, right? So how does the world change? How do you guys think about it when maybe the thing on the other end is agentic now, an agent?
Chris Struttmann
Well, it's really interesting, and I just gave a talk to one of our customers, I guess it was about a month and a half ago now. They're down in South Florida. They're in the hospitality travel and leisure space, I guess. And I talked to them, I said, "Our customers that are doing the best at adopting these technologies are the ones that are able to govern the humans the best."
And that tracks. If you think about all of these model providers, Anthropic … I mean OpenAI originated this line of thinking, which is we want this thing, these bits and bytes being passed through this silicon. We want this to be as good as a human. So if you follow that premise and you embrace that premise now, I don't know if that's necessarily going to be true, but if you embrace the premise over time, then you're going to want to govern these things just like humans.
The only difference is it's no longer nice to have. It's more of like a business mandate and I'm sure over the infinite timeline it will become regulatory, as well.
Adam Blue
Yeah, interesting. So do you think that every agent has a person behind it who's accountable for what the agent does? Is that the way this maps out?
Chris Struttmann
Me, personally, I believe that for the way the technology is right now, for the current state of the technology, you can very easily not have that. You could very easily have an entire company run by basically one human being who's just engineering prompts. I don't think that's going to be good. I don't think anybody wants that. I don't think anybody's willing to accept that risk.
I think the human's going to be in the loop for quite some time. I think the models will continue to get better. The technology will continue to advance. I do think the technology's going to become more efficient, but I can't imagine a world where a human is not going to be in the loop if for no other reason than you need someone to be held accountable.
Adam Blue
Yeah. I think that's the key word, right? Because today if something happens at a bank, there's a building I can drive there. The building has a logo on it. There's people inside the building. And so if I'm unhappy and I can't get to them through a digital channel, I can do that. Amazon, Netflix, Apple, it's a little tougher. They have buildings with logos on them, but they don't want me to drive there. I'm not welcome there. Which is fine.
Chris Struttmann
They also don't want you to call them.
Adam Blue
No, they don't, but they'll interact with me. I've had issues with Amazon and Apple and Netflix, all three of them, and somebody will get back to you. So even though those technology companies sit on top of a mountain of technology, there's still a human being that's accountable.
And so I think a little bit of the paradox is we're going to lean into agents and agentic AI and we're going to get to a point where it's not going to be feasible for me maybe to approve every action. Maybe we can't read every line of code that's written. And still someone will have to be accountable. Even if it's a billion-dollar company with one founder and a hundred agents, if one of those agents does something that's not acceptable, violates the social contracts, laws, Uniform Commercial Code, that person is going to have to stand up and be accountable.
And so I think there's a yawning chasm between what we're capable of and driving that level of accountability. And so talk a little bit about maybe how you at ALTR are fitting into that or what you see the gaps are in that space.
Chris Struttmann
Yeah. So this is a really good segue. The first problem, we'll start with a gap. The first gap I see is there's a huge gap in identity. Right?
Adam Blue
Yeah.
Chris Struttmann
And so a lot of our customers, they're in the business, they're looking for ways to enable business intelligence across the organization, whether it's large bank, small banks, CPG company, it doesn't matter who it is. Everybody wants access to the data so they can gain insight. And obviously, and it's been like that since we've been in business. Tableau has a business because that's what people want.
Then regulation and other business requirements said, you can't just have it. You have to put some controls around it. And then AI comes along and says, "Well, we're just going to accelerate your thirst for that outcome." And along the way, technologists got a little sloppy. And so, you've got things like-
Adam Blue
Never.
Chris Struttmann
Yeah, no, that'll never happen.
You've got things like, and I'm just using broad strokes examples here, but Tableau talking to databases with service accounts and that obviates anybody who's using Tableau of any accountability of who's accessing the data. Right?
Adam Blue
Yeah. It really does.
Chris Struttmann
And I'm not trying to put Tableau on assault here. It's just an example that I see every day, but it obviates all of the accountability. So now it's like, OK, well, I want to hook my AI agent up to my data so I can get insight. Well, I've taken identity out of the loop. And when I don't have identity in the loop, how am I going to have accountability?
And so if we keep with our premise, which I think you and I agree on, which is humans are going to be in the loop, there's going to be some kind of vector for accountability for the future. It's going to be an identity. And now we're in the policy business, right? We're in the governance business. We're in the business of making sure that you can attest to who's accessing your data, how, and protecting your data when it's not being used or not being used by the right people at least.
And so identity is the predicate for all policy, right? It's the who. Who is accessing this data? And when somebody comes to access this data, what am I going to do about it? And so you ask for a gap and you think about how do we fit into this huge, huge, huge identity gap out there. And I think there's solutions. We're developing some solutions that could solve for that. There's other developing solutions in the industry, but I think identity, if your organization doesn't have a strong identity system, you're going to start from behind in a race to adopt AI. And I firmly believe that both personally and as the CTO of ALTR.
Adam Blue
Yeah. I think there's an interesting distinction to be drawn here between guardrails that you put in after the fact to say, OK, this is designed to prevent this kind of behavior. And then having a foundation under which the behavior you want to prevent is simply not possible. Those are two just radically different approaches.
Chris Struttmann
Correct. You can't isolate a single user through a service account. The service account example is just one example, but it's such a clear and obvious example for any technologist that it's just like, what are we doing guys?
Adam Blue
Yeah, I agree. I'm pretty familiar with the OIDC and OpenID kind of the way OAuth evolved because we use it for open banking where I actually am still allowed to write a little bit of code, which I enjoy very much. And the notion of the scope is really interesting there. Because what happens is I grant access as my identity. I let someone else borrow it, but I determine, and this gets cryptographically signed in, right? I determine what they're allowed to have within the set of scopes. So as long as somebody with ordinary skill in the art of thoughtfully implementing data access does so in a reasonable way, I can say give a fintech scope to a certain set of accounts for a certain set of purposes. So I don't even have to give away the entirety of what I have access to. I can just give away that one little scope.
And so do you think this turns into maybe end users enabling agents on a task-by-task basis, like at a scoped level? Is that maybe how this is shaped? Because I'm not even sure I'm super comfortable saying to an agent, "Hey, just go crazy in my inbox." That makes me nervous.
Chris Struttmann
I am not. It makes me super, super nervous. James … James and I … James, the other co-founder of ALTR, we talk about this all the time. It's like these tools, they look so good and the value looks so promising, and I know it's going to make my day better. But when my email is signed, Chris Struttmann, Co-Founder and Chief Technology Officer of ALTR, that means something to a certain amount of people. Right?
Adam Blue
Yeah.
Chris Struttmann
And I can't just delegate that to anything where I can't predict the result. It's funny that you mention OIDC. I actually was thinking about this earlier. I was going back and forth on whether or not I was going to talk this technical. Have you seen OIDC-A?
Adam Blue
No, I don't think I've ... I've seen the term, but I haven't dug into it.
Chris Struttmann
So it's a bleeding standard. It's on archive. I want to say it was out of Cornell maybe or something like that, but it's an extension of OIDC that is specifically for agent workflows.
Adam Blue
Cool.
Chris Struttmann
So every agent in ... And that's ultimately where I believe this is going. That's why I'm mentioning it. It might not be OIDC-A.
Adam Blue
Sure.
Chris Struttmann
That might not be the thing that gets adopted, but there's going to be some standard like that. SAML democratize signing into enterprise apps and enterprise identity really OIDC, OIDC for consumer and open apps, open app ecosystems. There's going to be something like that that's going to enable the agents to provide a rich identity in an attestation and then the application or another agent downstream is going to say, "OK, here's your entitlement answer. Here's your scopes."
Adam Blue
Yeah, interesting. I'll definitely take a look because I think OIDC, in my opinion, works very well for open banking, right? It's got a good shape. The scopes are well thought-out.
Chris Struttmann
I would imagine so. Yeah.
Adam Blue
It fits very well. My guess is that agentic workflows are going to need probably more granularity and probably a new series of just axes to the way you think about building it out because that design has to be so thoughtful in order to avoid unanticipated consequences.
Chris Struttmann
It's got to be axes for control. And then also one of the other thing that gets forgotten about is the ability to attest to what happened. So it's got to be something that's going to be able to lend itself to that. It's like, "Hey, we asked for this credential, we asked for this ability to do this, and this is what happened along the way."
Adam Blue
Yeah. Yeah. No, that's interesting. Almost like there's an obligation attached to the delivered scope. That's interesting.
Chris Struttmann
Yes, yes. You asked if you have the ability or you gave me your credential to log in and you want to initiate a wire transfer, I'm expecting you to initiate a wire transfer.
Adam Blue
Yeah. Yeah. Interesting. Yeah. I came up on the security side and doing a lot of networking and every time I turn around that basic concept of firewall rules on an old, like an ASA, you would sit down and you would write …
Chris Struttmann
Or even just IP tables, right?
Adam Blue
Yeah, right IP import to IP import, established connections, not established connections. We keep coming back to this, it's almost like a filtering concept. We implemented the same thing in our commercial banking product and it's bottled on a firewall, right? And you can say, "This user can do this thing to this account with this payment type up to this amount of money under this set of circumstances."
It's a generalizable model because it ends up, man, I don't want to go off on a tangent here, but it ends up just being an expression of linear algebra effectively. So it's mathematically closed, it's easy to compute, it's easy to think your way through and yeah, it makes sense that we would get there.
I think the challenge a little bit to me just to bring this back to a little more human focus is, and how much time are we going to have to spend giving our agents permissions to things and managing those permissions? Because it feels like that could be a pretty big burden if you want to take advantage of the technology.
Chris Struttmann
I mean, when you first start using Claude Code, you spend the first ... You can't walk away from it because every two minutes it's asking you, "Can I modify this file? Can I modify this file? Can I start this Docker container?" And it's like, yes, yes, please, as a matter of fact, please do. What you cannot do is reach out to my AWS account and shut down an existing Docker container. Don't do that, but everything that's here on this machine, yeah, knock your socks off.
Adam Blue
Yeah. And it's fascinating because an LLM is an interesting construct because underneath it's just a long series of numbers and that's all it is. But you start to fall in love with it a little bit, not the creepy way, like the regular way. You start to relate to it like it's an embodied sentient kind of thing, even though you really know it's not. And so it's so easy to assume that it has just a little bit of common sense and it really doesn't.
Chris Struttmann
Not. Not.
Adam Blue
First off, there's no sense to it at all.
Chris Struttmann
No.
Adam Blue
It's essentially if it was a human mind, it would be insane just right off the bat.
Chris Struttmann
I think about a really old programming trend. You might remember extreme programming, two people, one keyboard.
Adam Blue
Yeah.
Chris Struttmann
Right? Of course.
Adam Blue
There was a sweater with four sleeves, right?
Chris Struttmann
Yes.
Adam Blue
And two people could wear the sweater at the same time.
Chris Struttmann
Yeah. Yeah. Exactly. Exactly. And so I think about coding a lot. Obviously we're in the software business. So using Claude Code or any of these coding agents, it's extreme programming with an intern.
Adam Blue
It is.
Chris Struttmann
They can keep their fingers moving, but somebody's got to be like, "Oh, you know what? Maybe don't use a for loop there. Maybe use a hash map instead because it's going to be direct access," or something like that.
Adam Blue
Yeah, but I'm always struck. I was coding, I had a code smell, Claude complained about it and I said, "Well, I don't have to worry about that because this code is always guaranteed to run exactly once per process lifetime." And it was like, "Oh yeah, absolutely. You're totally right." And then I just thought, "Man, you really should have pushed back harder on that. It's still a bad code smell."
Chris Struttmann
Yeah, but we still have an opportunity to fix it.
Adam Blue
Yeah. Yeah.
Chris Struttmann
Right?
Adam Blue
Yeah. So I said, "Let's go ahead and fix it. " And then it said something creepy. It was like, "I'm so glad you're going to fix it." And I was like, "Hey man, why don't you just keep that to yourself?"
Chris Struttmann
Hey we're not here to judge. We listen and we don't judge.
Adam Blue
That's right. That's right. No judging other people's code.
Chris Struttmann
Going back to what I believe one of the core tenets about identity being important in an AI world, one of the problems with that is, and I'm curious to see if you agree with this, is it's going to put pressure on the identity and identity verification, especially in these high-risk workloads. And making sure that who is starting off this process or who's interacting with this agent or who's giving them their orders or the on behalf of part of this equation is actually on behalf of. And so there's going to have to be an opportunity and a market coming up really soon to solve for that as well.
Adam Blue
Yeah. I think there's two fundamental challenges. One is very straightforward to solve, which is when we first turned up on the HashiStack, even before we finished the cloud migration, we were turning stuff up on the HashiCorp stack and we just melted our instance of Vault. And for those of you that don't spend your days configuring HashiCorp software and Nomad containers, Vault is the place that you put all of your keys and secrets so they stay safe.
And when your container starts, it has access to Vault to go get what it needs so that your secrets aren't in source code, which is a huge, huge problem, people leaving keys and access points and source code. But if you don't plan for it and Vault is underscaled, it's a huge problem. And so we may increase the number of authentication events that happen in an enterprise by tenfold, a hundredfold, a thousandfold. I'm guessing a lot of people's authentication infrastructures are probably not adequately scaled for that level of engagement. So that's one problem.
Chris Struttmann
No, definitely not.
Adam Blue
And then the second problem is, I have written software for a long time and I think about software all the time and it is not always easy to answer the question of should this particular process be given access to this particular chunk of data? That's a hard question to answer without really thinking about it. And so the other scale problem is, how effective can end users be, especially when they don't think about things this way all the time and effectively granting all of those entitlements and permissions without first off, just getting exhausted with it and mashing the yes button over and over and over or making tiny mistakes that have significant consequences. I think it's going to be a challenging problem in a lot of ways.
Chris Struttmann
I agree. And I just went to see, it was the Gartner Security and Risk Management Summit in Washington DC, I guess it was two weeks ago today. And I went to go see an analyst there. I went to his talk and he did have some interesting perspective.
We're doing work, I'm not going to turn your show into a venue to plug my own product, but we are doing work in the data classification space. And so I went to his talk, which was all about classification in the AI era and how to enable value out of that classification data, which was really good. And one of the insights that I gained from it as a product leader here at the company is tying that classification and the necessity for data access back to business context, that's something that we're working on, that's something that I know a lot of the people in our industry are working on and that's something that's going to be important. Because yeah, it's not really a meaningful data point to say so-and-so or so-and-so's agent is asking for data, right?
Adam Blue
Yeah.
Chris Struttmann
Or asking to do something or same thing applies to banking, asking to facilitate a wire transfer or asking whatever process needs to happen, to your point, it's not enough to say, "Here's a data point saying somebody's asking to do something." There's got to be some context. And in our world, that's a business context.
Adam Blue
If you think about social engineering as a security phenomenon, a lot of that is I think you could put it in a bucket that I'll call context forgery, right?
Chris Struttmann
Oh, yeah.
Adam Blue
So I send an email that's not really from the CEO, but it looks like it's from the CEO and it says, "I need you to go buy $10,000 in Visa gift cards and put them in a mailbox to here." And people do that because they look at the …
Chris Struttmann
I'm at a convention center in Las Vegas for this tradeshow.
Adam Blue
Yeah. Whatever it is, right?
Chris Struttmann
It makes it real, right?
Adam Blue
Yeah. And so there's enough manufactured context that it kind of passes the baseline sniff test. And then for an employee that has pretty high-power distance to the CEO, the risk to them is like, there's a very real risk of getting yelled at or not getting a good review or whatever it is. And so leveraging that scenario. And I think we're about to bombard people with a lot of asks for, "Hey, very rapidly evaluate whether or not this is a reasonable context for the exposure of this data, make a decision and then try and go back to your regular job." So it's an interesting notion.
Chris Struttmann
Yeah. And just even more broadly outside of security, these AI agents and all of these workflows and tools like ServiceNow, which I love, they're turning us from thinking machines, mainline code processors as humans into interrupt processors, right? The interrupt pen goes high. It's like we go to this, we put the instruction pointer here and we process this line of code and then we get, like you said, we go back to mainline code.
Adam Blue
Yeah. I don't want to spend my work life feeling like all I do is fetch items out of some Kafka queue and then put them back on the queue after doing something. That doesn't feel ... It's not fulfilling work either.
Chris Struttmann
Correct. Precisely.
Adam Blue
So the flip side of this topic then, which I want to spend a little time on, is I keep seeing people, especially on LinkedIn, which I read due to professional reasons, but it always makes me sad.
Chris Struttmann
They make you cringe.
Adam Blue
There's so much slop on LinkedIn now. Yeah. I mean, there's plenty of smart people who post great things on LinkedIn. And then the other 90% of the content is like, it's clearly AI-written. It's full of terrible AI pictures.
Chris Struttmann
It's Instagram for enterprise.
Adam Blue
Yeah. But I swear to you, if I read the word “governance” one more time without somebody specifying what they mean when they say “governance,” I'm going to lose my mind up in here.
Chris Struttmann
Yeah.
Adam Blue
So why don't you tell me to you guys, what does governance mean in two sentences or less? And you can have a couple independent clauses in there if you want, but what is governance? How should we think about it?
Chris Struttmann
Governance is knowing what the data is, where it's living, and being able to attest to how it's being used and by who.
Adam Blue
All right, that's fantastic. So what, who, and to some extent, why, right?
Chris Struttmann
Yeah.
Adam Blue
It's those questions. And so I see a lot of talk now about we're going to bring you a governance framework. You have to have a governance framework for AI. Is that a piece of software you implement? Is it a policy you write? Is it a mindset? Is it like a Zen koan where we talk about it while we rake the pedals? What does it mean?
Chris Struttmann
You asking what I think it means?
Adam Blue
Yeah. I'm asking what you think it means.
Chris Struttmann
Or you asking what does the industry think it means? Because the industry thinks it means ... Thinks it's going to be …
Adam Blue
Buy my product.
Chris Struttmann
Well, yeah, but I think the industry thinks it's going to be some new SOC 2 Type 2 process. Right?
Adam Blue
OK, interesting.
Chris Struttmann
It's like, "Hey, here's the risks, here's the controls, here's what we do to mitigate the controls. We bring it together. We'll see you next year, pay us our however many tens of thousands or hundreds of thousands of dollars. You can throw a logo on your website."
And so I think that's what the industry thinks it is. I think below that, there's going to have to be some kind of practicality. And much like I'm sure in the banking world, our customers, they've got so many disparate systems and so many different places the data can live and so many different places the data can go. There's not going to be a way to make attestations to that without having technology behind it.
Adam Blue
Yeah. So from a practical perspective, if governance is what data, who's accessing it and for what purpose, then there's an integration of some set of technology that helps you answer those questions and implement your policy, enforcement. And then there's also probably a role for the thing you said, which is the, we're going to write down these controls or we're going to test whether or not they're actually accomplished and then we're going to give you a report at the end.
And so I think you're right. I think it's probably going to feel like another SOC 2 or another kind of compliance exercise, but it's going to be much broader and much richer than it is for when you only have humans operating because there's just limitations on what a human can do over a relatively short period of time that are not true for agentic software.
Chris Struttmann
It's probably going to be broader in number. I think I also get to take part in our SOC 2 Type 2 practice here. And so it's probably going to be broader in terms of number of controls because these AI agents are unbounded and humans are also unbounded, but like you said, there's only so much that can be done in a day, and also there's only so many things that humans want to do and there's proper tension in the system as well, but it's probably going to be broader in number of controls that are just going to be very, very, very specific
Adam Blue
Yeah, that makes sense. That makes sense. All right, last, let's touch on a little bit of a paradox that I think is really interesting. So as a consumer, not as a CTO or a person that works in software, as a consumer, there's a whole lot of places I want my data to go securely. And I went and got a new car and so they emailed me a PDF, which I then printed out with a printer and then I took a pen and I made marks on the paper, and then I took a picture of it with my phone and I set it back and there were, I think, three forms for me to fill out and they all had the same information on them.
And that just felt ridiculous. It just felt absolutely absurd that I couldn't just point them to some like, if I want to sign up for Venmo, I can go link my bank account. If I want to do something else, I can probably electronically transmit some of that data, but there's so many processes not just in banking or finance but across the world that still have this very primitive relay. And on one hand you think, "Man, why isn't all that just electronic? Why can't I just transmit some QR code that contains my fundamental identity?" And then on the other hand, I'm like, "Oh my God, I'm so glad we don't have that because ..."
Chris Struttmann
Right. Take the Social Security number problem and multiply it by a million.
Adam Blue
That's right. I often believe that these technologies, when they arise, first they create a whole bunch of new problems, but oftentimes they bring with them the capability to solve those problems in much more interesting, much more efficient ways. And so is it possible to tell a story where the paradox of open data sharing and making data more available and the challenges associated with agentic AI, where's the good story in this where agentic AI makes it more possible to share more data more securely than we do today?
Chris Struttmann
Well, as you know, especially due to some of our work, we're huge proponents of data sharing. I think your example about buying the car is a great example. You go to the car dealership, you've got your title document, its warranty document, whatever else. And they all have the same information as a perfect example of the problem. And these organizations, they don't talk to each other. And so the fundamental premise of data sharing is to let the organizations talk to each other. There's of course hurdles to overcome just like with everything. In the world of data sharing, most of it's regulation, of course. Don't let any good idea go unregulated, that's for certain. But the ability for these disparate systems, organizations, people to just have access to the data is a very promising world.
Now, how can AI be a part of that? I think it's got to be done, first of all, it's got to be done responsibly. But if you empower the AI with the context of who it's working on or what case it's working on, regardless of whether that AI was at the car manufacturer or at the car dealership, or they say that they know that Adam wants to take delivery of the car on Wednesday, so we're going to send them an email on Tuesday or whatever it is, giving them the context benefit of data sharing is certainly going to make things far less high friction for the consumer.
Now, in terms of a data privacy aspect, we're both aware, ALTR is working on some of the technologies that make that safe. And so things like format preserving encryption and tokenization, again, not to plug my own product on your show, but those are just enablers to be able to do things like that and to get those attestations that you need in order to feel good about AI.
If we assume that, and this isn't going to happen, but let's just say we live in a perfect world. If we assume that everybody goes by the same adoption principles, which is we need to have identity, we need to have policy at the data, and we need to have attestations, those are my three principles. Let's just say everybody adopts that. That means that everybody needs the same ability to have all of those controls. That means that when you share the data, you have to assume that that person that you're sharing it with has those controls and their agent can do part of their part of the job perfectly, just like your agent can do your part of the job perfectly.
And so it's going to have to take a universal approach for AI to be able to do that. But I have a tremendous amount of faith in our industry, otherwise I probably wouldn't do it, but I do. I think the promise is so good. Sorry, that was a really long-winded answer.
Adam Blue
No, it's great. That's a great answer. I like the way you're thinking about that. I think it's interesting.
All right. Anything you want to say and wrap up before we get to our kind of cultural artifacts to share for the day?
Chris Struttmann
I’ve just got to tell you, this cultural artifact has been absolutely anxiety-inducing. So let's just get to that.
Adam Blue
All right. All right. So Andy Warhol famously said “Art is what you can get away with,” but even that was a copy. So today's reference for culture to accompany the episode is Marcel Duchamp, a French-American artist who was one of the Surrealist movement way before Warhol. And he very famously presented a urinal as a piece of art and kind of sparked a revolution. And you look at some of his works. Also, this is not a pipe painting, that's a painting of a pipe.
Chris Struttmann
I've seen it.
Adam Blue
That's another very famous one of his. And he just had this interesting perspective on what was art and what could be art and what he wanted to say. And I'm not sure if he was an entirely positive force or an entirely negative force, but there was a certain suspension of the rules when he decided to make work. And I think the tech industry sometimes can operate at that same kind of level. And I think there's very interesting lessons in Marcel Duchamp's work. Hugely influential artist, fascinating.
And you can go to a gallery and see the work, or I mean, you could just use Google. You could literally just Google Marcel Duchamp and fall into an amazing internet black hole of fascinating stuff. So I encourage everyone to do that.
Chris Struttmann
I am going to look it up right after this.
Adam Blue
All right. Well, thanks for being on today, Chris. This was really fun. Really appreciate your insight.
Chris Struttmann
Yes.
Adam Blue
I appreciate your participating.
Chris Struttmann
Of course.
Adam Blue
And so that's it for today's Cut to Context and you can find us on whichever purveyor of fine, high-quality podcasts that you choose to patronize. Thanks everybody.
Chris Struttmann
Thanks everyone.