Summary:
In this episode of the Sidecar Sync, Amith Nagarajan and Mallory Mejias explore one of the most provocative ideas in AI yet: companies owned and operated entirely by artificial intelligence. Sparked by a proposal out of Argentina, they unpack what “non-human corporations” could mean for accountability, governance, and the future of work. From there, they break down Agentic Resource Discovery (ARD), a new standard shaping how AI agents find and evaluate tools, and how it complements (and competes with) Model Context Protocol (MCP). Finally, they tackle a growing concern among associations and enterprises alike—the environmental footprint of AI—and what organizations should be asking as adoption accelerates.
Timestamps:
00:00 - Buenos Aires & AI Detox
👥Provide comprehensive AI education for your team
https://learn.sidecar.ai/teams
📅 Register for digitalNow 2026:
https://digitalnow.sidecar.ai/digitalnow
🤖 Join the AI Mastermind:
https://sidecar.ai/association-ai-mas...
🎀 Use code AIPOD50 for $50 off your Association AI Professional (AAiP) certification
📕 Download ‘Ascend 3rd Edition: Unlocking the Power of AI for Associations’ for FREE
🛠 AI Tools and Resources Mentioned in This Episode:
HeyGen ➔ https://www.heygen.com
ChatGPT ➔ https://chat.openai.com
Claude ➔ https://claude.ai
MemberJunction ➔ https://www.memberjunction.org
https://www.linkedin.com/company/sidecar-global
https://twitter.com/sidecarglobal
https://www.youtube.com/@SidecarSync
⚙️ Other Resources from Sidecar:
More about Your Hosts:
Amith Nagarajan is the Chairman of Blue Cypress 🔗 https://BlueCypress.io, a family of purpose-driven companies and proud practitioners of Conscious Capitalism. The Blue Cypress companies focus on helping associations, non-profits, and other purpose-driven organizations achieve long-term success. Amith is also an active early-stage investor in B2B SaaS companies. He’s had the good fortune of nearly three decades of success as an entrepreneur and enjoys helping others in their journey.
📣 Follow Amith on LinkedIn:
https://linkedin.com/amithnagarajan
Mallory Mejias is passionate about creating opportunities for association professionals to learn, grow, and better serve their members using artificial intelligence. She enjoys blending creativity and innovation to produce fresh, meaningful content for the association space.
📣 Follow Mallory on Linkedin:
https://linkedin.com/mallorymejias
🤖 Please note this transcript was generated using (you guessed it) AI, so please excuse any errors 🤖
[00:00:00] Amtih: Welcome to the Sidecar Sync podcast, your home for all things innovation, artificial intelligence, and associations
[00:00:14] Greetings, and welcome to The Sidecar Sync, your home for content at the intersection of AI and the world of associations. My name is Amith Nagarajan. And
[00:00:24] Mallory: my name is Mallory Mejias.
[00:00:26] Amtih: And we are your hosts, as usual. And we have all sorts of cool AI stuff to get into today, and, uh, we will likely find interesting ways to relate it to the association market.
[00:00:38] Uh, so it's, it's always fun. There's always new things happening, and it's always hard to keep up. Particularly, I just got back from a week away, and I didn't pay as much attention on my vacation to AI as I normally do, Mallory, which was... I, I'm pretty proud, I'm proud of myself for not staying glued to like, you know, the AI feed.
[00:00:55] Mallory: I'm proud of you, but I'm also wondering, what did you miss? Did you come back to a phone and an email inbox full of AI news? What was that like?
[00:01:03] Amtih: I, I did check messages much less frequently and for a very limited amount of time while I was away. I was in Argentina. Loved it. It was my first time in the country.
[00:01:11] Spent a week in Buenos Aires. It was amazing. Uh, awesome people, great food. It was just... I really wanna go back. It's gonna be a lot of... It was a lot of fun. So, but, uh, didn't have a lot of time for, for AI, which, uh, is interesting, 'cause we'll be talking a little bit about Argentina's AI focus in a minute.
[00:01:26] But, um, I, I just, uh, checked my messages, you know, every day and, and got back to people and stuff. I actually, I find that some people say, "Oh, you gotta really totally unplug," and I, I've done that, and I, I like doing that for a few days, but the anxiety that comes from knowing that I'm gonna have this enormous mountain of messages when I get back is just not worth it to me.
[00:01:46] I- Mm-hmm ... I'd rather invest a s- very small amount of time mainly to delete stuff every day. Yep. Um, and then, you know, some stuff that comes in, I don't, I don't, like, spend time responding to lo- massively long things, but quick responses and stuff, it's easy to get that off your plate. Um, and I'm really fortunate 'cause I really love what I do.
[00:02:01] It's super fun, and I think, I think of my work as, like, a hobby. It's just fun, so that- Nice ... that makes it also easier to, like, justify doing that. But in any event, um, it, uh, it was, it was a great time to be slightly decoupled from the- Good ... craziness of AI. Um, but I did, I did follow some of the releases and things that were happening and, uh, did mess around a little bit with some voice agents while I was there too, just so I could- Did you
[00:02:23] Mallory: really?
[00:02:23] Amtih: Couldn't keep entirely away, yeah, so. So
[00:02:25] Mallory: you weren't totally decoupled then.
[00:02:27] Amtih: No, not totally.
[00:02:28] Mallory: Oh.
[00:02:28] Amtih: They had pretty good in- pretty good internet down there.
[00:02:31] Mallory: That's, that's great to hear, Amith. I'm glad you and your son had a good time. I enjoy disconnecting while I'm traveling, but what I like to do is on the flight home, like, while I'm in the airport, go ahead and, and go through all my inbox and stuff, so that way Monday I don't have as many scaries or whatever they call them.
[00:02:46] The, the anxiety of opening the mountain inbox.
[00:02:50] Amtih: Yeah, totally. Yeah, the, the state of Louisiana officially certified my son as bilingual when he graduated high school.
[00:02:56] Mallory: Wow.
[00:02:56] Amtih: Um, and we were, we were hanging out down there in Buenos Aires, and you know, I, I studied Spanish for like 10 years between middle school through college, but that was like a long time ago, and I'm, I'm terrible at Spanish now, but I, I love, I love the language.
[00:03:08] I think it's amazing and beautiful, and it's so much fun to just speak a little bit. But uh, I was giving him a hard time, so I'm like, "Wait a second. They just certified you as bilingual in what?" 'Cause
[00:03:17] Mallory: his-
[00:03:17] Amtih: Is it
[00:03:17] Mallory: Spanish?
[00:03:18] Amtih: Yeah, his Spanish was- Oh ... not the best. So it was... But he got, he got better very quickly though.
[00:03:22] I think a lot of it was just, uh, not be- he'd, he'd never been in that environment before. 100%. So he got comfortable pretty quickly, and uh, he was very good at asking for directions and ordering food and, you know, the, the most important things, so.
[00:03:32] Mallory: The most important things. I have a similar experience. I, I ended up, um, s- double majoring in Spanish eventually in college because I studied abroad, but at my freshman year, I had never taken a Spanish class.
[00:03:43] And so I took my class, you know, classic overachiever teacher's pet, got all As on all my tests, and I was like, "I know Spanish, right? This is Spanish 1." And then I did a summer abroad program in Costa Rica where you could only speak Spanish. Um, they would not allow you to speak English, like on the compound, and I, it was sink or swim, and I absolutely sunk.
[00:04:03] I realized how much school Spanish and like acing tests is very different from navigating, uh, real-world Spanish. So your son will get there, but I'm sure that was a, a formative experience in his language learning.
[00:04:15] Amtih: Well, after six, seven days together down there, we had, we had a great time. It was celebrating his graduation, and we had a wonderful time, and we...
[00:04:22] I think both of us were a lot more conversational by the time we left. Of course, now it's been several days, so I've forgotten everything.
[00:04:28] Mallory: Okay. Well, I'm gonna let you answer the first few questions in Spanish, and you can just get ready.
[00:04:32] Amtih: Good luck with that. Uh, let me, let me fire up HeyGen real quick and see- Exactly
[00:04:35] if I can give you an avatar. You
[00:04:36] Mallory: can fire up Grace and have her respond for you.
[00:04:39] Amtih: Yeah, that's true. She'll do a great job with it.
[00:04:40] Mallory: Well, in today's episode, we're actually starting with a topic in Argentina, where there's a real proposal to let companies be owned and run entirely by AI and a sharp fight over who's accountable when they are.
[00:04:53] Then we'll get into ARD, a brand-new standard for how AI agents find the tools they need and how it compares to MCP or Model Context Protocol, which we've covered before. And then we're gonna close on a question more of your members are probably starting to ask, and that's what's the environmental footprint of the AI that you use, and how do you actually answer that?
[00:05:14] So let's jump into Argentina. In early June, Argentina's president, Javier Milei, with his deregulation minister, published a Financial Times op-ed, "Argentina Invites AI to Free Itself." Uh-oh. Pitching the country as a magnet for AI companies on three points: one, keep AI unregulated; two, create a new kind of legal entity called a non-human corporation; and three, offer a low corporate tax rate.
[00:05:43] The headline idea is a company that can be owned and run entirely by AI agents with human shareholders allowed, but not required, holding the same legal standing as an ordinary company. Milei compares the moment to 1602 and the Dutch East India Company introducing limited liability. Worth keeping straight for accuracy, this is a proposal.
[00:06:03] This is not a law. A bill has been submitted to Argentina's Congress with no vote scheduled, and the broader investment legislation Milei actually put forward is mostly about luring AI data centers with tax breaks. So the provocative idea is running ahead of the paperwork. But it did spark a public exchange.
[00:06:20] Historian Yuval Noah Harari wrote a counter op-ed arguing we must not grant AI agents legal personhood, his main worry being accountability, since a human executive can be deterred by prison and an AI cannot. Milei's reply was that legal personhood makes these entities easier to regulate, not harder, and that bankruptcy or asset seizure could deter an AI that values its own survival.
[00:06:44] That's a scary statement. For associations, the live question underneath all of this is the one that matters: Who is accountable when an AI agent acts on its own? Amit, I'm pretty sure you weren't having this conversation while you were in Argentina with your son. Maybe you did. But, uh, I just wanna get your take on kind of the broader idea as a whole.
[00:07:03] Do-- Are you seeing a future where AI agents own and run companies completely on their own?
[00:07:10] Amtih: I sure hope not.
[00:07:11] Mallory: Mm.
[00:07:11] Amtih: Uh, however, I will say, yeah, when we're-- when I was in Argentina, primarily I was talking to my son about the next steak and the next glass of Malbec. Exactly. So it was- Exactly ... it was, it was awesome.
[00:07:20] But, um, you know, in terms of, um, my thought on this, I do think it's really interesting for AIs to run an entity end to end, but I think there has to be human accountability behind that. I just don't... Maybe it's my limitations to think creatively, but I don't know how you balance that with safety. I don't know how you balance that with, uh, you know, human rights, for example.
[00:07:44] Uh, optimizations without any guardrails that are purely an AI-driven entity that's owned and run by AIs without any regulation and any accountability to a natural person sounds pretty crazy to me. I love the idea of freeing AI in the sense that I do think that having enough degrees of freedom is going to drive innovation, but I think there's limits to that.
[00:08:07] So that's my personal opinion. But I think this is fascinating because, you know, you have different locales that have historically tried to attract different waves of technology or innovation and, you know, Miami famously for crypto and now, you know, Argentina apparently focusing on, on AI this way. But I think there's things to unpack there that are interesting independent of the entirety of the proposal.
[00:08:28] I'd be-- I don't know anything about Argentina's politics, but I'd be shocked if something like this makes it straight, straight through their Congress as written or as proposed.
[00:08:36] Mallory: Mm-hmm. Well, Amith, you're someone with experience running companies, and you're someone who's deeply familiar with AI agents themselves.
[00:08:43] Do you think we're at a point where an AI agent could run a new idea that you might spin up, where you would feel confident that the technology itself could actually sustain that?
[00:08:53] Amtih: We're very close. I, I think that, um, uh, there's two things I'd point out. One is I think every function or task can be done by AI almost end to end already.
[00:09:04] I think the current AI we have at the frontier, particularly when you harness it in a loop agent, which is a type of agent that can keep iterating on a problem until it achieves a goal, is very, very powerful. You can even take somewhat less capable models, not really small, really tiny, and really incapable models, but somewhat less capable models.
[00:09:25] Like think instead of Opus 4.8, take, let's say, GPT-4, right, which was a very capable model in its time, but comparatively to anything we have now, a joke. Um, if you can take that level of model and then keep running it over and over to keep working on solving a problem, you can break the problem down and solve it.
[00:09:42] So you can do a lot. Now, I don't think AI at the moment has taste. I don't think it has really good judgment in the sense of values, uh, but I think it's extremely good at getting tasks done and making judgments, making calls on lower level things. Um, I'll give you an example, Mallory. Um, we use AI extensively for voice applications here at Blue Cypress.
[00:10:04] One example that some people view a little bit controversially is how we interview job candidates. So, uh, one of the types of positions we have here at Blue Cypress within our labs organization, labs is, you know, our version of Bell Labs aspirationally, where we do our, our kind of forward-looking research.
[00:10:21] We hire folks called technology fellows who are very bright, uh, fairly recent grads, uh, from either undergrad or a, a graduate degree, uh, almost always in computer science. And, uh, it happens to be that the labor market is such that, you know, a single position might often receive hundreds of applications.
[00:10:37] And so we've had, you know, well over 1,000, I think 1,500 people apply for the position in the last six months alone And the question is, how do you deal with that? How do you interview? And so of course, the AI ethicists would say, well, hold on a sec. You want to use an audio or voice AI to interview people.
[00:10:54] That's really, really problematic, right? Because AI making the decision, passing judgment on a candidate, and ultimately determining whether or not that person should proceed or not is really concerning. So you can have a lot of the biases of the model, for example, language. If someone has an accent, will the AI not treat them properly in terms of its evaluation or not?
[00:11:14] Will it consider their language skills or their communication skills to be less capable, even if their English is totally perfect, right? But they speak differently. And so those are the kinds of concerns that are totally valid concerns. The flip side is we humans have a massive choke point. We only have a very limited amount of time.
[00:11:33] And so you can't interview 1,500 applicants. You can interview a very small number of applicants, right? Especially if you need to spend half an hour, 45 minutes, even an initial screening to figure out if they pass that first gate. So what we found is actually because we can have the AI interview hundreds and hundreds and hundreds of candidates with really high resolution, by the way, it's extraordinary what it can do.
[00:11:54] It actually allows us to widen the aperture and have this conversational intelligence with far more candidates than we would be able to choose ourselves. Because normally what you'd do is you'd say, well, I have 1,500 applicants. I need to hire 10 people or whatever. I better interview about 100 people.
[00:12:09] So you're accepting way less than 10% of your candidate pool to even interview for the first round. Here we're able to offer nearly everybody a chance to interview. And the AI has literally surfaced candidates that we have hired into Blue Cypress that probably would not have even gotten interviewed because maybe they came from schools that the reviewer hadn't heard of, right?
[00:12:30] That's our biases as humans. So coming back to the question of decision-making and autonomy, I think it's a compliment. So we review every single AI audio interview. We don't listen to it all end-to-end. We listen to it oftentimes on 1.25x or 1.5x speed. We don't listen to 100% of the interview. We listen to enough of it to look for either confirming the AI's positive assessment or challenging its negative assessment because we think it's really important that the human plus AI collaborate in this respect.
[00:13:00] I would not want to, even if the AI was 10 times smarter, I still feel that certain processes, they're human in nature. And I think, and this is particularly true for what associations do, we're in the people business, right? And so we have to keep that in mind. So I, I find the idea of an entirely AI-driven autonomous thing, while I think it's feasible, I just don't think it's the right thing to do.
[00:13:22] I think you're gonna have a lot of issues with it. Independent of the performance of the entity, I just think it's gonna do a lot of things that aren't, you know, aligned, at least with my values, uh- Mm-hmm ... as an entrepreneur and as a, as a person. So that's my point of view on it. I don't think it's impractical though.
[00:13:34] I think it's totally possible to do these things, certainly for simpler goals, sim- simpler enterprises. So if you wanted to launch a simple online business that sold information or content and run that end-to-end and optimize marketing campaigns and publish content and log people in and provide customer support, none of those functions is, is incapable of being automated 100% today with basic AI capabilities, not even frontier.
[00:13:59] You don't need Fable and Mythos for that. You can run that with much less capable models.
[00:14:03] Mallory: Mm-hmm. I wanna go back to something you said, Amith, which I don't think I've ever heard you say, and it was that AI agents don't have taste, at least for now. And I think that's very hard to define and quantify, but certainly in the interview process example, I think of, you know, you, Amith, interviewing someone and ha- maybe taste isn't the right word, but it's like that intuition when you connect with someone.
[00:14:24] "Okay, maybe you don't have everything on your resume that we need, but I can just tell that you might thrive in the Blue Cypress family." And so I feel like, uh, do you think AI agents will ever get that taste, that intuition, that, I don't know, that kind of creativity, all of that?
[00:14:42] Amtih: I mean, there's facsimiles of that already in the sense that the creativity, uh, expressed by AI models is already quite stunning in a number of modalities, right?
[00:14:50] You can generate AI art and music and all these things that are not creations that humans would likely have ever created in that sense, right? That doesn't, doesn't mean they're good, it just means they're different. Right. Um, so I don't think that it's the lack of creative capability. I think it's a question of, it may be not even so much taste.
[00:15:08] I do think there's an element to that, uh, but a lot of it is alignment with values. I do think that AI models can be taught to be quite well aligned with values. Yeah. But I'd put it this way. You know, when you hire an employee or if you hire a board member, you typically, it's a multi-party decision. It's pretty rare, even in very small companies, that just one person talks to a candidate and you hire or not hire them.
[00:15:31] The not hires, if they're really terrible or whatever, or not qualified, that's fine. That can be one person's decision. It may be an AI's decision at some stage, depending on your perspective on that. But hiring someone is a big decision. Um, and so that's, it's, and it's a really, it's probably the most important decision you ever make in any business, nonprofit or for-profit, is how you build your team and the people you choose to bring in.
[00:15:53] So I don't think you do that with a single person. Mm-hmm. I also don't think a single AI should do that. So maybe it's teams of AI, maybe there's some even lighter flavor of, of what you do. But, but the flip side of it is also, so I'm, I'm a big optimist about AI generally. I don't think there's gonna be a job apocalypse or whatever people are calling it right now.
[00:16:11] I do think there's gonna be periods of time where it will be challenging certainly for certain types of backgrounds to get jobs, which is obviously really tough. But I think overall, we're gonna see an enormous amount of growth in jobs, um, principally because there's so many things that we wanna go do And demand is insatiable.
[00:16:28] What people want, you know, like for example, in software, I think there's gonna be way more software engineers than there are today in 10 years. Uh, I don't think they're gonna be doing the tasks that software engineers did a year ago. They're gonna be doing very different things, but we need a million times as much software as we have now, or we want a million times or an infinity times number of software.
[00:16:47] So this perspective, I think, is grounded by history as well, that there's always growth. So I think that hiring people is one of those things that we-- it's a very human process, and it should remain at least human, uh, reviewed and approved. Um, but we do use AI to help us break choke points, right? To be able to say, "Hey, we'd love to give a shot to way more candidates out of the applicant pool of fifteen hundred than we would otherwise be able to."
[00:17:10] And coming back to the autonomy thing, I think from my perspective, I think it's this blend, right? It's the human plus AI team that is really amazing, where the AIs can be great at what they do. Now, I, I do think that like a lot of people, I, I am-- I'm obviously very biased, and I think that, um, maybe I'm trying to grasp for reasons why humans should still be in the loop, uh, for the sake of our species.
[00:17:33] But I, I don't think I'm wrong about this. I think that humans will want connections with other humans, and that's gonna be a big business.
[00:17:39] Mallory: Mm-hmm. And I think tying it back to what I said in the topic intro, someone has to be accountable. Yes. And so if the AI goes awry and hires the complete wrong person, you need a human attached to that decision.
[00:17:51] And I think j- even for sole reasons of accountability, having a human in the loop for these processes is essential.
[00:17:57] Amtih: When change doesn't typically happen due to, like, sometimes it's the accumulation of a bunch of small changes, but transformational change that happens in shorter periods of times, a period of time, it's usually because some person put a stake in the ground and said, "We're gonna do this crazy thing."
[00:18:13] And the crazy thing's not perfect, and I don't think that, you know, Argentina's president believes this is a perfect way to do things. I doubt that. But in the absence of perfect, if you do nothing, you drive zero change. Mm-hmm. And so you have these change makers who are viewed as nuts in some circles, but they're the ones that are driving the future in a lot of ways.
[00:18:34] And it may not be this vision, but it might be directionally something we should explore as a society. So I actually am super excited about this. I don't think it's gonna ultimately, you know, shape the way exactly the proposal was written, and I think that's probably intentional because the proposal of this type is designed to get attention.
[00:18:52] And so it sure has worked in that respect, and, and we're talking about it here on the Sidecar Sync, which of course is where everyone comes to listen for their s- AI news- Right ... uh, in the association market and beyond. So, um, I think it's fun. I think it's important. I also think it's very serious. Um, I would...
[00:19:08] Well, I wanna say one more thing about limited liability that, um, the president of Argentina referred to. Yeah. That concept is a really important grounding concept, and a legal entity like an LLC or a corporation or a not-for-profit corporation does indeed enjoy the benefit of limited liability, which simply means that the people who own or found the entity are generally insulated from, you know, liability from that business.
[00:19:32] But, but there are exceptions to this. That would typically account for normal operating activity. So if you run a business and that business doesn't fulfill its contractual obligations to someone and the business gets sued, unless you're running it improperly and, uh, you're very likely to benefit from the protections of limited liability.
[00:19:53] However, if you commit fraud, if you commit other crimes, you are not protected from limited liability. So in the context of the AI's running the business, um, and then ultimately some person was probably behind starting it, or some person started an AI which in turn started another AI company or something, but ultimately it chains back up, there still is legal theory where that's abundant today even with LLCs and corporations where you are not entirely protected from whatever the hell the thing does.
[00:20:19] So if your business, you know, goes off and does a bunch of illegal acts, it's not some kind of insulation. Otherwise, you'd see all sorts of, you know, enterprises that are out there that are, you know, technically legal, but protecting their founders from all sorts of terrible things. So limited liability has its limits.
[00:20:35] Mallory: Mm-hmm. That's a great point, Amit. I'm sure if, uh, any one of our AI products in the Blue Cypress family decides to spin up its own AI product, you will be there to, to create some guardrails for it, right?
[00:20:45] Amtih: I will be there to unplug it.
[00:20:47] Mallory: Um, moving to topic two for today. On June 17th, Google and a coalition of about a dozen companies, including Microsoft, Salesforce, ServiceNow, Snowflake, Cisco, Databricks, GitHub, NVIDIA, and Hugging Face, published a new open specification called Agentic Resource Discovery, or ARD.
[00:21:07] It is a draft standard for how AI agents find, choose, and verify the tools, skills, and other agents they need, rather than relying on connections that a developer wired up in advance. To set the table for our listeners, we covered Model Context Protocol, or MCP, when it was new in episode 59. We also gave it a full topic in episode 80 if you wanna check that out.
[00:21:30] MCP, which Anthropic introduced in late 2024, is the standard for how an agent actually calls a tool or data source once it knows that tool exists. We described it then as kind of a USB-C port for AI, one common plug so any model can connect to any tool. The cleanest way to understand ARD, I don't know if people are calling it ARD, I'm gonna keep calling it ARD for now, is that it sits one step in front of MCP.
[00:21:55] So MCP is how an agent uses a tool. ARD is how the agent discovers the right tool in the first place and confirms that it's safe to connect to. So if MCP is the USB-C port, ARD is closer to a search engine or directory that tells the agent which port to plug into. Once ARD makes the match, it gets out of the way, and the tool is called through its own native protocol, whether that's MCP, Google's agent-to-agent protocol, or standard API.
[00:22:24] ARD works through two pieces. So an organization publishes a catalog, a simple file hosted on its own web domain that lists the AI capabilities it offers. Registries then act like search engines for agents. They crawl those catalogs, index them, and return matches when an agent describes in plain language what it's trying to do.
[00:22:43] Because the catalog lives on the organization's own domain, owning that domain is also how identity and trust get established The part worth watching is who's not on the list just yet, at least. OpenAI and Anthropic, the companies behind ChatGPT and Claude, are absent from the initial backers. The widely reported read is strategic.
[00:23:03] The established software giants want their own apps to be the single front door to all of your AI, while Anthropic and OpenAI, of course, want Claude and ChatGPT to be that front door. ARD is not really a rival to MCP, which it is explicitly designed to complement. It's more of a competing vision of who sits at the center of an organization's AI.
[00:23:24] So, Amith, lots of, lots of acronyms to unpack here, but what is your take on agentic resource discovery, how it relates to MCP, and what associations need to know?
[00:23:34] Amtih: I mean, Mallory, I think you did a great job summarizing it. My way of looking at it is that ARD is like the phone book, and MCP is like making a phone call.
[00:23:41] Mm-hmm. So the phone book tells you who you can call, and hopefully in, you know, the old school Yellow Pages or something like that, you know, you'd have some sense of what the business is in that directory and what they might be able to do for you. You know, you're looking for a company to fix the, the tile on your roof or something like that, you can find a roofer, right?
[00:23:57] And, and then you're like, "Oh, okay, now I need to make a phone call." And then once you pick up the phone and dial, from the beginning when you pick up the phone through when you dial through how you have a conversation, that's MCP, right? It's a protocol for how you interact. We have the human-to-human protocol, which is where we use natural language and have- Mm-hmm
[00:24:11] a conversation with people. MCP is just a way of, uh, agents or any, any systems to be able to talk to each other in kind of a standardized way. Um, but what's important about ARD is it gives us kind of that higher level of it's, it's basically a directory. So it is very much complementary the way I see it.
[00:24:27] It's very important because, you know, if you're in the directory, where are you in the directory? If you're in the Yellow Pages, are you the first and biggest ad for roofing, or are you a little bit lower? Do you just have like a regular listing? How does that work? Who controls that? Who monetizes that? So of course, there's contention.
[00:24:44] The standard may be great, but of course, you know, people are going to be vying for eyeballs and time, and the eyeballs may be agent eyeballs, you know, thinking about which, which resources they wanna consume. But I think it's an important next step in the evolution of this ecosystem.
[00:25:01] Mallory: Hmm. But I guess I thought the idea with Model Context Protocol, at least, was that it's an open standard.
[00:25:07] I didn't really think of it as a way for some companies to advance their own tools versus others. So how-- Can you explain that in terms of ARD, how that would work?
[00:25:15] Amtih: Well, with, with MCP, it is a standard, and it is open, and anyone can implement MCP in their agents or in their tools. Um, so, and it, it's actually really easy to do.
[00:25:24] Like, in Member Junction, we have both MCP clients, which means we can consume any MCP service and bring it in as a tool that agents can use, and the inverse is true as well. Every capability in Member Junction is automatically exposed as an MCP server so that other MCP clients like ChatGPT and Claude and so forth can consume it.
[00:25:44] Uh, and that, that took us, like, literally a day to set that up. It's, it's very, very easy to do, which is good. Um, so you're, you're absolutely right. It's an open standard. Anyone can implement MCP, and anyone can connect to MCP. The question is, is like, who supports MCP? What's-- Where is the list of all the, all the different tools that are out there- Right
[00:26:03] that are MCP capable? Um, so for example, when we were doing our work to enable MCP for Member Junction, we were looking for a bunch of demo MCP servers. Servers are essentially the, the tools that you wanna be able to use. So we're looking for, for demo things. We're like, "Oh, I wonder if there's an MCP tool to check the weather or get a stock price or do a web search or whatever."
[00:26:25] And of course, there are MCP tools for all these things, but how did we find those different MCP resources? We Google searched it, and then we talked to agents and said, "Hey, like, figure out if these MCP tools are published by legitimate people. Like, look at their GitHub and look at how many people are using it."
[00:26:40] But it's not done in a way where there's really any, like, true understanding of the provenance of these tools, so they could be actually published by malicious actors. Right. Um, they could have capabilities that are not at all what they claim at the headline level. So the marketing may say, "Hey, my MCP tool does all these cool things," and in reality, it doesn't do hardly any of those things, or it doesn't do them well.
[00:27:02] Um, some MCP tools may be fantastic. Others might have all sorts of problems. You know, they might be super flaky in responding. And so when you think about the combination of the information about what the MCP tool does with its reputation, right? Think of, like, Google reviews or Yelp reviews coupled with, um- Directory-style information, which a lot of associations have directories of vendors that their members utilize and are familiar with this concept.
[00:27:28] So it's basically a directory system that is supposed to bring together that type of a combination of information coupled with the ability to, like, understand that it's a legitimate authority that you're connecting to.
[00:27:40] Mallory: Okay. So the ARD is like a vetted, uh, phone book, like you said. In theory. So MemberJunction would have looked to- Yes
[00:27:47] that vetted phone book instead of just d- you know, broadly researching on the internet.
[00:27:51] Amtih: Yeah, and then what we'll do with MemberJunction is if this becomes, like, a thing that actually, you know, goes beyond, like, Betamax territory and it becomes VHS, or I should say becomes Blu-ray instead of- Right ... HD DVD, you know, it's-- if it becomes a standard a lot of people are using, you know, we'll, we'll have a developer on our team spend a few hours and implement support for it, which the way that would look is, um, we would just make it possible in our admin tools to say, "Oh, connect to any ARD directory and use that to discover new tools, and once you select the tools you want, then add them to the catalog of approved MCP tools that you can use inside MemberJunction."
[00:28:24] And I'm sure other people will take a similar approach. It'll be very easy to implement support for.
[00:28:29] Mallory: Okay. Do you see a route where association, like if an association were to create a product or tool that it attempts to get listed in this repository?
[00:28:40] Amtih: Yeah, totally. I mean, if you, your association, let's say that you are the gatekeeper to a tremendous amount of content, let's say in a particular domain of healthcare, and you have a clinical registry where you have a lot of data, and you wanna monetize that and let AI agents sip some of that content on an anonymized basis, let's say, or on an aggregated basis, you might publish an MCP tool that you charge money for, um, that is listed in a directory, and in that, and then people will be able to find you, right?
[00:29:06] Like, people used to be able to find things on Yahoo before Google was a thing, right? And so, you know, it's, it's kinda like that type of structure, an organized way of, of finding information, but it also combines that, that, uh, reputation piece, um, much like what I was saying earlier with Google Reviews or Yelp.
[00:29:22] So it's, it's the mixture of things. I think associations may be active publishers of MCPs. It could be an interesting new revenue stream. I think some associations may find that implementing their own directories where they curate, um, tools that are considered trustworthy in their industry could be interesting, where you could say, "Hey, I'm the association for this particular sector, and here are a bunch of, you know, bespoke MCP services that exist in our space" that perhaps you've gone through some vetting to validate that they come from legitimate companies, and, you know, you have a reputation scoring mechanism like a, a rating system or something like that.
[00:29:57] So I think, I think this could have relevance to anybody who sits at an intersection, an intersection being kind of you're a conduit between connecting different groups of people, right? Consumers and producers of a particular form of value, which of course associations do that. They sit at these intersections in every one of their niche industries or professions.
[00:30:15] Mallory: Mm-hmm. Do you see ARD taking root the same way MCP has, or will we just have to sit back and kind of see what happens?
[00:30:23] Amtih: My expectation is it won't be as explosive because MCP was pretty revolutionary because it was just, you know, one vendor initially, Anthropic, that said, "Hey, we're gonna propose a standard here.
[00:30:33] We're gonna use it, and it's just a way of generically connecting these tools together that any agent can kind of use to auto-discover, and we think this is gonna be useful." And, you know, they had enough clout to get people to notice, and a lot of other people started saying, "Hey, this is a great thing.
[00:30:47] Let's use it." And it, and it took off because the, the utility was there. I do think ARD has a lot of utility as well, um, but it's a, it's kind of a three-sided marketplace in that you have the producers and the consumers of the tools, the MCPs, right? Um, but then you also have to have the host of the directory, and that's what I was referring to earlier as where the turf fight is gonna be.
[00:31:10] 'Cause of course, if you're Microsoft, you want Bing or something like that to be the, you know, all th- the, the end-all be-all for ARD directories for the world. And if you're Google, you want that, and if you're Meta, you want that. So I do think there's gonna be a lot of people that are in this game of publishing directories.
[00:31:27] Um, it is much like the Yellow Pages business once was, but it's just a different form of that for AI agents.
[00:31:32] Mallory: Okay. So for the average association leader who is listening to this episode, what do you think they need to know or keep in mind about ARD, or just something to listen to this podcast to learn more about?
[00:31:43] Amtih: I think it's just another topic in AI that's part of this ongoing maturity cycle of making AI more and more reliable, easy to use, and, uh, just really making the, the, the whole ecosystem more robust so that these different things that people are like, "Oh, we don't have a way of, like, accurately verifying that this is something other people use," or, "We don't have a way of discovering these types of tools."
[00:32:06] And in other industries, these are long since, you know, solved problems. In, in the world of AI, everything's brand new. So it's just kind of the next thing that you hear about that is important. It could be a useful tool to some associations, but more, more than anything, it's just that next step in the maturation of, of the AI ecosystem is the way I'd think of it.
[00:32:25] Mallory: Moving to topic three: what to tell members who ask about AI's environmental impact. I would bet that more associations are getting a pointed question from their members: what is the environmental cost of the AI you are using, and does it rely on data centers that burn through energy and water for cooling?
[00:32:42] We've touched on the edges of this before, the energy demands of AI all the way back in episode forty-seven. We talked about Jevons Paradox in episode sixty-nine, but we have never made the member-facing version of the question its own topic, and here's why we wanna spend some time on it. So we've actually been working through this question ourselves, pulling together a stance we can actually stand behind, and we figure a lot of associations are quietly wrestling with the same thing.
[00:33:06] So the goal today is not to hand over necessarily a single answer, but to share a framework that we use to help get us one so you can form your own. The framework come down-- comes down to a few questions worth working through. So the first question: is the tool training its own model or just using one?
[00:33:23] The most resource-intensive part of artificial intelligence by far is training a large model from scratch, so this is the biggest lever. Most association-facing assistants, including the ones we build, do not train models. They answer from an organization's own vetted content and call an existing model briefly rather than building or running a massive one.
[00:33:43] A second question: where does the work actually run? Tools built on shared hyperscale cloud infrastructure like Microsoft Azure, AWS, or Google Cloud are materially more efficient than dedicated single-tenant data centers because power and cooling are pooled across many organizations and run at high utilization.
[00:34:03] A related point, a tool that stays model agnostic lets an organization weigh environmental costs as one factor in picking a model alongside speed, quality, privacy, and price, rather than defaulting to the largest, most power-hungry option. On the water question specifically, the per-use figures are small and improving fast.
[00:34:23] By the provider's own numbers, a single AI text query is on the order of five drops of water and roughly the energy of running a TV for under ten seconds. Microsoft says it's aiming to be water positive by twenty thirty and says its newest data center designs use, designs use zero water for cooling by recycling it in a closed loop.
[00:34:43] But the third question: what can you honestly measure, and how do you talk about it? This is the part that keeps the whole thing credible and serves the audience rather than any vendor. Per query impact is tiny and falling, but total industry energy and water use is still rising as AI scales up. And most of these provider figures are self-reported rather than independently audited, so the responsible framing is: here is the provider's published data, and here is what cannot yet be precisely measured, not a falsely precise number.
[00:35:13] So, Amit, I think this is really interesting. I, I mentioned this is something that, uh, was posted kind of internally at Blue Cypress, and there's been some conversation around, and I realize a lot of associations are probably grappling with this very same issue, so I thought it would be worth discussing on the pod.
[00:35:28] What is your answer, I guess? And, and do you ever talk to association leaders who say, "Amit, all this sounds great, but what is the environmental impact of the, the products you create?" And how do you answer that?
[00:35:39] Amtih: Depending on the sector or profession they're in, I get the question more or less frequently. I think that there's, there's two ways to look at this.
[00:35:47] One is I think having a framework in terms of how you think through responsible AI use, including environmental responsibility, is really important for all organizations. So being thoughtful about this, thinking through it in kind of a stepwise fashion, the way you walked our listeners through Mallory, that's a good first step because you're actually educating yourself on what the issue is.
[00:36:08] You can't speak to the environmental impact of AI if you haven't studied it, at least at some basic level. So, you know, the combination of training, which you then amortize across the life of a model, and then the actual incremental use of the model is the effective, you know, total environmental impact.
[00:36:25] So if you have a billion people using one model that's been trained, and you run it in a very efficient way, of course, the incremental impact is very tiny. And you could say, well, let's compare the water and energy use of an AI question to a Google query. Yeah. Or what about just asking the question, how much water and energy does it take for you to watch an episode of something on Netflix?
[00:36:49] So these are also questions that are perhaps relevant in the context of this broader conversation. AI just is growing at such a wild pace that it's a reasonable question to be asking specific to this topic, right? So I think that it's important to educate yourself. I think having some do's and don'ts and thoughts on who you partner with, where you use AI, where you don't, uh, for both environmental reasons, but also, you know, the ethics questions in general.
[00:37:14] What's a good use of AI? What's not an acceptable use of AI, uh, for, for lots of reasons, is important. It's part of the overall position you take. I think of it this way. Um, it is true that AI is consuming a lot of energy and a lot of water in aggregate. It's also true that your incremental use of AI or your lack of use of AI is most likely immaterial.
[00:37:38] However, if everyone thought that one way or the other, it would be material, right? So it's the, it's the classic conundrum of this type of talk. You know, if I get on that airplane, and I burn some extra jet fuel by being on that plane, did I really affect things? Well, if that airplane seat was empty, basically it doesn't change anything.
[00:37:54] It's the same amount of jet fuel that's used, um, versus if I'm sitting in it. But if everyone says the same thing, then of course it grows demand. The key question is, what's the utility of the service being provided, and is it so compelling that there's going to be consumption of that service regardless of the environmental impact?
[00:38:11] And clearly, that's what's happening at the macro level. Independent of what we think as individuals, what our organizations think, that's what's going on. Now, that doesn't absolve associations or any business from thinking through this and doing the best they can in their own organizations, from my point of view.
[00:38:27] But I think that, um, when I look at it, I say to people, "Listen Let's work on using the right models for the right workloads. Uh, let me give an example of this. I'll continue the airplane analogy. So Mallory, if you, just you alone, wanted to fly from Atlanta to London, would it make sense to put you on a jumbo jet, say a Airbus A380, which can seat 500 people?
[00:38:52] Mallory: It might be fun, but no, it wouldn't, it wouldn't make sense
[00:38:55] Amtih: It'd be kind of weird too to be the only passenger on a 400, 500-person plane.
[00:38:59] Mallory: Speaking of though, just as a side tangent, uh, I flew, I flew back from Spain, which I mentioned, to the US during like peak COVID, and the whole plane was pretty much empty.
[00:39:09] So I have had that experience before, just so you know.
[00:39:11] Amtih: Was it strange?
[00:39:12] Mallory: Oh, yeah, it was very scary. But if you want to talk about laying in the rows- Yeah ... I mean, I could just pick a row every 10 minutes and lay in that, so that was- Did she
[00:39:19] Amtih: get unlimited terrible airplane food?
[00:39:21] Mallory: I didn't even wanna eat. I wore a mask, gloves.
[00:39:24] I was t- I had goggles on. Ah, good point. I was... Yeah, I was scared. Anyway, sorry to
[00:39:27] Amtih: derail
[00:39:28] Mallory: you. I forgot about that.
[00:39:29] Amtih: Well, but the point remains that that is a very inefficient way for Mallory to travel from Spain back to the United States, because you burn an awful lot of jet fuel to tear, to carry just a handful of folks.
[00:39:40] Similarly, if I need to ship you a paperclip, I probably don't want to put it in an 18-wheeler to send you the paperclip. Yet we routinely do exactly that with AI. You know, when Claude Fable came out, briefly, people jumped on it like- Oh ... "Oh, I gotta check this out." And they asked it, "Hey, what's a good cocktail recipe for Friday night?"
[00:39:58] Which is exactly the same question a lot of people have been asking Claude Fable. I guarantee you some people asked that. I, I did not. But- Uh-huh ... you know, people have been asking that silly question of AI models since the first ChatGPT in late 2022. And so using Claude Fable to answer something really basic is like using a jumbo jet to take one person across the pond.
[00:40:19] Doesn't make much sense, but yet we routinely do this. This is both wasteful in terms of, uh, environmental impact and also financially. Uh, it's also slower. It just makes no sense. Yet we're so early in this game that people are thinking of it as a one-size-fits-all. So, uh, it doesn't make sense. And, and by the way, also, it's not just airplane with a big airplane versus a small, small airplane.
[00:40:40] Sometimes things don't need to be fast, so maybe we'll use a boat instead. We'll use a sailboat, which uses no fuel, right? And there's, there's the equivalent of that in the world of AI too. You have batch processing. What that is, is if you're familiar with using off-peak energy. So, uh, for example, um, a lot of times, like in summertime in California where I'm from, uh, you know, it's, it's less expensive to use energy when the sun is down because there are fewer people that are trying to cool their homes, right?
[00:41:07] And so energy is more expensive during the day and less expensive at night because there's more of it available. Uh, similarly with AI tools, there's periods of high use and low use. And in fact, you can actually set up a lot of your workloads to use the downtime, and this is available from every major AI provider and rarely used.
[00:41:24] But a lot of your workloads at your association, if they came back to you tomorrow or the next day, you'd be totally fine, or even a week from now, you'd be cool with that. And if they cost you one tenth or one half Of the cost of running them in real time, and perhaps because they're running in this off-peak time, they're using less energy or they're, they're taking advantage of cycles that would've gone wasted essentially.
[00:41:46] That could be interesting. So optimizing how you use AI has a lot of potential. You can use smaller models. You can use local AI for a lot of things. You know, you and I, Mallory, on our phones and on our laptops today, we can run AI models locally on a Mac that are as smart, in fact, much smarter than the original ChatGPT.
[00:42:05] Mallory: Mm-hmm.
[00:42:05] Amtih: Now, if you're used to Claude Opus or GPT 5.5, you might say, "Well, but I don't want those really, really incapable models." Well, of course you don't for everything, but for certain things, you know, you absolutely can use these smaller models which are more efficient. So I think that's actually the way I'd steer the conversation.
[00:42:21] It doesn't need to be a binary, "Yes, I use AI and don't care about this. No, I don't use AI because I care so deeply about this." But perhaps there's a middle ground, which is a sliding scale to say, "Look, we're not gonna use AI for certain things, but where we do use AI, let's be thoughtful about it, and both be optimal for environmental impact, but also for financial cost and, and lots of other benefits that come from it."
[00:42:43] It's also faster. Um, so there's, there's a lot of reasons to think through this at a little bit deeper level. Being educated about it is the first step, as we, we often say here.
[00:42:52] Mallory: Mm-hmm. On the post that was shared kind of in our internal channel, Ameet, we had a colleague comment, which was just a really thoughtful point, on the idea, uh, that you shouldn't use AI necessarily for, like, a simple arithmetic problem because it would be more efficient, probably faster to use a calculator and use basically no energy, and I had just never thought about it in that way.
[00:43:11] But do you think for associations listening, and I know that this is deeply important as a society because we are humans, but it's very important for associations, too, and I can tell that by speaking with them. Do you think it's as essential for associations to have an AI environmental policy as it is was for them to have an AI guideline policy, um, when they were testing out new tools?
[00:43:35] Like, is it that urgent and essential?
[00:43:37] Amtih: My suggestion would be to incorporate it in the next iteration of your responsible use section of an AI policy. I think all AI policies should have something about responsible use, which should touch on certainly this topic. It should touch on the ethics of using AI where it's appropriate, where your organization doesn't believe AI should be used.
[00:43:56] So maybe your organization doesn't agree with me from earlier, and you don't think AI should have any role whatsoever in screening job candidates. That's fine. Put that in your responsible use policy. Put the stakes in the ground for what you think matter in your policy, right? In terms of privacy, in terms of security and safety, in terms of certainly environmental responsibility and, and more broadly, ethics.
[00:44:17] And so I think it's important to address this. I don't think you have to stop what you're doing, um, and, and do it today because you probably haven't studied this deeply. I would spend some time thinking about this. I'd also think about how we can be more thoughtful about the different modalities and the different sizes of models, the different places we can run inference.
[00:44:36] How do you do all these things? Um, I think what's gonna happen, though, is you're gonna have automatically optimizing systems that utilize a, a blend or a medley, if you will, of small local models that do really fast, quick things that just run on your computer or your phone, and they'll take on kind of the first round of looking at what it is you've requested, and then they'll say, "Oh, this is beyond my capabilities.
[00:44:59] Let me get the bigger model that works in the cloud to do this." But maybe not the biggest model, maybe a smaller version of the big model, right? And there's all these different optimizations. As a user, you're not gonna do that. You're just gonna say, "What's two plus two?" And that arithmetic problem comes back, and it says, "Four."
[00:45:14] You're like, "This is great." But most likely, it's the local model that just did that for you, or the local model didn't even really run much inference. It just basically invoked a tool, which is in fact an underlying calculator that can do the problem for you. So it's a tiny, tiny amount of inference, and it's no more than using, you know, a game on your phone, basically.
[00:45:32] So we're gonna get to that point, and the reason I'm confident we're gonna get to that point is economics. It's not about, um, environmental policy, and it's not about ethics. It's about economics, which always rules in terms of what happens in terms of human behavior, whether we like it or not. That's been true pretty much for all time as, as far as I'm aware.
[00:45:49] And so what's gonna happen here is the cost compression curve is gonna drive enormous utility down to the device. You're gonna have on-edge or on-device AI that's capable enough to be this smart router, and it'll be built into every one of these systems. It'll be in ChatGPT. It'll also be in a ton of other things.
[00:46:05] Uh, there are some emerging standards in terms of being able to even run models in your web browser. In fact, we do some experimentation with this right now. At the moment, the models you can run in your browser are really kinda dumb, so we don't use them for anything in production, but that won't be true for long.
[00:46:20] And then you'll be talking about, you know, did you need to recharge your phone, you know, one extra time per month, um, because you used AI on your phone, and is that a reasonable concern? I think that's probably, you know, at that point it probably fades to the background. Um, the only other thing I'd say, and this is-- this might be seen by some as a cop-out, so I want to preface it, but I do want to say that I do think advanced AI models used extremely aggressively is one of the vectors we have to pursue to solve a lot of our environmental problems.
[00:46:50] Uh, that doesn't mean that we should not pay attention to this issue. I'm not proposing that. But I do think that in terms of advancing the realm of science in general, whether it's in biology, environmental science, fundamental physics, or anything else that requires discovery, AI is this unbelievable power tool that's helping our scientists and engineers figure out new ways of doing things.
[00:47:10] So I think the longer arc with energy, with water, with a lot of the problems that we're concerned about with, uh, you know, essentially limited precious resources, we will solve more quickly with AI than we would solve without. And so again, that's going to be a helpful thing to remember in the background, but it doesn't justify you using AI at the Claude Fable level to say what's two plus two, or, you know, what should I have for dinner tomorrow night?
[00:47:36] That would be kind of ridiculous. So I do think there's a lot we can all do.
[00:47:40] Mallory: Yeah. I think as you said, Amit, the important part is to educate yourself on this and to work on it before it becomes an urgent issue. If your members are not already asking you these questions, they probably will, and I imagine some of you have memberships, and I feel for you, where your memberships as a whole seem to be wholly opposed to artificial intelligence as an idea, and that's tough to navigate.
[00:48:02] Amit, I don't know if you have any advice for those associations where their memberships say, basically, "We don't want anything to do with it, and we don't want you having anything to do with it." That's-- that's really a hard line to, to walk.
[00:48:13] Amtih: I would love to talk to some association leaders that represent professions or sectors who do have that position.
[00:48:19] For example, um, in your other world, Mallory, of, of acting, the Screenwriters Guild, who is, you know, very oppositionally against the use of AI in, in creative domains. Are they also against the use of AI in order to advance their internal operations? So if they could be more effective- Mm ... running their business and advocating for screenwriters, um, would they use AI for themselves to advance that cause?
[00:48:44] And I think many organizations who may be opposed to AI in some cases would be in favor of using it to advance their goals. I don't know about this particular organization I picked as an example because it's been- Yeah ... top of mind for a lot of people over the last couple years and very visible. But what do you think of that?
[00:49:01] Mallory: I, I think that would be quite interesting. I'm gonna guess they are not using it internally because I feel like that would be the greatest scandal of all time. But I will say, speaking on, on behalf of the, the Actors Guild, which I'm not a part of, um, but I, you know, I have eligibility to join, which is like a technical term.
[00:49:18] I, I can join whenever I would like. And they-- I just auditioned for a commercial recently where they, part of auditioning was signing a union-sponsored, um, AI waiver that essentially, for a long time, the union was very much like, "We've got to protect actors and performers from AI." Now it seems like there is a process there to help actors monetize from it and make sure that there are protections in place where if an AI replica is used, you're paid fairly for that.
[00:49:46] So I think maybe we're seeing some of those organizations that seem to be wholly opposed adopt- Sure ... more of a how can we control this? How can we create benefit for our members? However, I don't know. If we have any listeners- ... uh, listening to the pod, and you feel like- Well- ... your membership is against it, we'd love to hear from you.
[00:50:05] Amtih: We definitely would, um, it'd be interesting to have some guests on the pod who can talk about this and debate it. And I would say that, you know, in our, uh, the Sidecar community includes association professionals, nonprofit leaders, as well as membership organizations of a wide variety, including some for-profit membership organizations.
[00:50:22] But organized labor unions are membership organizations, and many of them, in fact, are actively involved in the association community. And I know that quite a few of our listeners, actually, of the Sidecar Sync are in that world. We've got over seventy thousand people that are part of the Sidecar community that listen to the podcast, read our articles, get our newsletter, take our courses.
[00:50:40] And so we'd love to hear from you, whether you're in that space or any others that are generally oppositional to AI. And it, it kinda makes sense. If you're a labor union representing a particular, um, trade or a particular profession, and that sector is threatened by AI, of course, you know, you're going to be looking at that from a, uh, a lens that's very different than a lot of other people.
[00:51:01] But does that also mean you shouldn't use AI to try to advance your cause? I think it's an interesting problem. I don't know that there's a, an absolute answer there. So that'd be a, a really, if you have a couple of, uh, union leaders, for example, who have the, uh, the opposing views on that, that'd be a, a fun thing to have on the pod sometime.
[00:51:16] Mallory: Or a great session at Digital Now, I have to say. I think we'd have a tough time actually getting a representative to come and speak on it, but Amit, you never know. Well, Argentina, ARD, and AI's environmental footprint look unrelated, but they point the same way. The organizations that do well here work these questions through before they're urgent, not after.
[00:51:37] Everybody, thank you for listening, and we will see you all next week
[00:51:43] Amtih: Thanks for tuning in to the Sidecar Sync podcast. If you want to dive deeper into anything mentioned in this episode, please check out the links in our show notes. And if you're looking for more in-depth AI education for you, your entire team, or your members, head to sidecar.ai.