Summary:
It’s been 130 episodes since Sidecar Sync last did a true deep dive on data—and a lot has changed. In this refreshed 2026 perspective, Amith Nagarajan and Mallory Mejias unpack what “association data” really means today, from structured CRM records to the massive untapped world of unstructured data like emails, community posts, and content libraries. They explore how AI—especially vectors and reasoning models—has flipped the script, making previously unusable data suddenly actionable. The conversation then tackles one of the biggest shifts in the AI era: data ownership. Even if you legally own your data, fragmented systems and vendor restrictions can limit how you actually use it. Finally, they break down what becomes possible once your data is unified and activated—from predictive insights to deeply personalized member experiences—and offer a realistic starting point for associations ready to take action.
Timestamps:
0:00 - Why Data Still Matters
👥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:
Member Junction ➔ https://www.memberjunction.org
Snowflake ➔ https://www.snowflake.com
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 🤖
[00:00:00:14 - 00:00:09:17]
Amith
Welcome to the Sidecar Sync Podcast, your home for all things innovation, artificial intelligence and associations.
[00:00:09:17 - 00:00:36:10]
Amith
This is your world of associations and AI, hyper focused on a topic that I don't think association folks ever get sick of, the data they have and how they don't get to use it the way they'd really like to. My name is Amith Nagarajan.
[00:00:36:10 - 00:00:38:04]
Mallory
And my name is Mallory Mejias.
[00:00:38:04 - 00:00:43:02]
Amith
And we are your hosts and we are swimming in a sea of data, aren't we Mallory?
[00:00:43:02 - 00:00:51:17]
Mallory
We really are. Amith, I don't know if you've skipped ahead in the outline I've prepared, but do you know the last time we had a data episode on this podcast? Can you guess?
[00:00:51:17 - 00:00:54:11]
Amith
It's been a minute, probably what, 12 months?
[00:00:55:12 - 00:01:11:23]
Mallory
Longer. I'm embarrassed to say the data episode, the official data episode that we put out was episode 12 of the podcast. Now we've done episodes that touched on different facets of data since then, but we have not done a complete refresh of the data episodes since episode 12.
[00:01:11:23 - 00:01:59:01]
Amith
Well, here we go. I mean, 130 episodes later, I think we've learned a couple things. I know I've learned a lot and the world of data and the world of associations has continued to move along quite quickly. So this is going to be a lot of fun. I think data is interesting because I've been working with data forever. You know, my whole career has been in one way or another tied to data. I started my first company in the early 1990s and it was a database oriented company and then that company actually eventually turned into Aptify, the AMS business, which is of course a data driven company. And basically everything we do at Blue Cypress is data centric. Our member junction AI data platform, all of our different SAS tools, a lot of the education we provide at sidecars is data related. So I think it's really good we're having a whole episode focused just on data.
[00:01:59:01 - 00:02:22:16]
Mallory
Yep, absolutely. I did have to go back and look at that episode 12 outline from way back when. And I've got to say, I didn't like it. It was pretty rudimentary. It was, I've got to give myself credit, right? This was the inception of the podcast. So I think we were very much figuring out how things would flow, but I'm excited to get a chance to redo the episode and put more oomph into it, more substance.
[00:02:22:16 - 00:02:24:12]
Amith
Let's put some oomph into it.
[00:02:24:12 - 00:03:33:08]
Mallory
Well, as Ami said, today is a single definitive episode on AI and association data in 2026. We did that other data episode all the way back in episode 12. And in some ways a lot has changed, but in some ways it hasn't. Back then the conversation we had was mostly about types of data you hold and getting your own house in order. Today it's a little bit of a different world. Agents are acting on your data in real time. Ownership has become a live question and the whole thing has higher stakes, or at least it seems like it does. So if you only listen to one of our data episodes, make it this one. We'll start with what your data actually is and why AI finally makes all of it usable, then we'll get into the question that matters most now. Who really owns your data once AI agents are in the picture? And we'll close on what becomes possible once your data is working for you and the realistic first steps to get there. So let's jump in. What your data actually is. The simplest way to make sense of your data is one split, structured versus unstructured. Structured data is the orderly stuff in databases, member records, and your CRM or AMS registration.
[00:03:33:08 - 00:03:35:07]
Amith
Everyone's AMS is super orderly, right, Mallory?
[00:03:35:07 - 00:03:41:19]
Mallory
You know, you would know better than me, I mean, it seems that way. When I think of an AMS, I'm like structured, very formatted.
[00:03:41:19 - 00:03:45:05]
Amith
You know, I've never actually been... Yes, definitely. It should be structured, yes.
[00:03:45:05 - 00:03:48:04]
Mallory
I have never actually been inside an AMS.
[00:03:48:04 - 00:03:56:12]
Amith
Oh, wow. You should, well, Sidecar Sync listeners, you should invite Mallory to check out your AMS sometime. That's an experience.
[00:03:56:12 - 00:05:50:07]
Mallory
I actually would be very willing to do that. I feel like I talk about AMS's all the time, but I've never actually been inside one. Unstructured data is everything that doesn't sit neatly in rows and columns, and it's the vast majority of what you hold. So unstructured breaks into a few familiar forms. Text is the most abundant. So think of your emails, community posts, articles, documents, and reports. Visual and audio is the mountain of recordings, images, and video you accumulate from conferences, webinars, and courses. And exhaust streams, which we haven't talked about in a while. But these are the byproducts of digital activity. So the opens and clicks on your emails, website traffic patterns, and community engagement that quietly pile up and almost never get used. So what changed? For years, unstructured data went unused because software couldn't make sense of it. You needed a human to read it. AI flips that. So it can read, summarize, and find meaning in all of that messy content at scale. No team ever could. Which is why the data you've been ignoring is suddenly your most valuable asset. The mechanism underneath that is worth at least 30 seconds. I hesitate to even talk about it this quickly, but we have a whole episode on vectors, if you would like to dive deeper on that. I believe it was episode 35. But when AI processes your content, it converts text, images, and audio into strings of numbers that capture meaning. So it can find things by what they're about rather than matching keywords. Vectors are why a modern assistant can answer a member's question from a document that never uses the member's exact words. That's the shift from keyword search to meaning search. Again, if you want to go deeper on that, check out episode 35. But Amith, of the types of data mentioned, text, audio, visual, structured, unstructured, exhaust streams, where do you think associations have the most value hiding that they're not touching?
[00:05:50:07 - 00:06:19:04]
Amith
I think, you know, there's, first of all, I love that description of vectors. It's such a clear way for people to understand. It's basically capturing the meaning of the data. And yeah, it happens to be a whole bunch of numbers that you never want to look at. And they mean nothing to anyone, anyone human that is. But they capture the meaning of the content and then you can compare them. Because we know computers can compare numbers really well, so they can say, how similar are these numbers or how different are these numbers? That's really what it is. It's kind of magical, actually. It's really cool.
[00:06:20:08 - 00:07:46:02]
Amith
I would say that of all the different types of data that you described, the unstructured data would be the category that I think is most exciting because it's what people have done the least with. And they have more of it than they have structured data. I was kidding earlier about AMS's somewhat, partly because I know our listeners will be laughing along with me about their AMS's not being particularly structured in the sense that yes, they are tables. Yes, there are rows and columns, but within those rows and columns, you have a lot of stuff to say it nicely. And that stuff, a lot of times, some of the most important information about your members and their relationships with your association was literally stuffed into a column like comments or description where because your AMS didn't support a field that is really important to you, you have adopted a structure or a technique or not a structure, but you just put this information in some other field. And so people do this all the time in databases, not just AMS's. And and that's definitely worth thinking about because that information all of a sudden can be converted to structured data quite fluidly with AI. But coming back to the truly unstructured data, I would point at something that all of us have more of than we might want as individual humans. But perhaps there can never be enough of it when it comes to AI. And what I'm talking about is email. I don't want more email, do you?
[00:07:46:02 - 00:07:47:16]
Mallory
No, I don't want more email.
[00:07:47:16 - 00:07:55:08]
Amith
Do not want more email. I don't want to send any more email. I don't want to receive any more email. I feel the same way about texts. I feel the same way about all these forms of communication.
[00:07:57:06 - 00:08:08:07]
Amith
And the thing is, is that we have a ton of that with our members. We have tons of email and it's in a system we have full access to. It's either Google or it's in Microsoft, most likely. That's probably 95% of you.
[00:08:09:10 - 00:10:29:05]
Amith
And that email can be used to power insights that you can put into a pipeline. So let me give you an example. You could say, take all the emails that I have received from my members and from those emails, let me infer their personality style. Let me infer their satisfaction level. Let me infer from that a whole bunch of other things. Now, if you had a really intelligent human, read all the emails to remember. Like if you are a member of an association and I was the member services rep and I had access to those emails, Mallory, and I read the whole chain of emails I've had with you for the last 10 years or whatever, and particularly focused, let's say, in the last three months, I could probably get a pretty good sense of your satisfaction level. I could probably get a pretty good sense of your personality style. Do you like quick, punchy summary messages or would you prefer long flowy pros when you're communicated with, right? Like I can get a pretty good sense of the person based on how they communicate with me. And actually, I think we all adapt to that. Like I know there's certain CEOs I communicate with regularly who have no patience for lots of detail. They don't want to hear the technical stuff. They just want to hear boom, boom, boom, one, two, three. And I get it. I'm like that too, when I'm on the receiving end. I have other people who are like, no, no, no, tell me the detail. Give me more. I want to know more and don't just give me the bullets. Give me supporting information, hyperlinks. And so I know these people. I know what they want. So like when I communicate with them, I do my best to tailor the way I communicate. Well, what if our systems could do the same thing, right? To know what it is that people prefer to understand their personality style and their various preferences, you could try to ask them. Problem is it's been pretty clearly shown through a lot of broad studies, not our studies, but studies that a lot of different social science fields have done that people are terrible predictors of their own preferences. They tend to tell you in some cases what they've been interested in in the past, but they tend to be terrible at telling you what they want in the future. Sometimes they can tell you some of these things that don't change frequently. Like I like summaries versus long things. But anyway, the whole point of what I'm saying is your email is actually this treasure trove of information that tells you a whole lot about people that may be used for improving the member experience the way I'm describing, or perhaps predicting if they're going to renew, maybe they're going to come to an event or not, all these types of things. And most people do absolutely nothing with email other than quickly file it away really as fast as they can.
[00:10:30:05 - 00:10:47:04]
Mallory
I like that. So email being as a place is probably untapped for a lot of you associations. What about online communities? That's another place I'm interested in where posts and comments and stuff, I feel like probably not a lot of associations are doing much with that information, but it could be pretty insightful to ingest.
[00:10:47:04 - 00:11:59:15]
Amith
Totally agree. Yeah, it's bullseye. You know, I post a question to the online community about X, Y, and Z. First of all, that question should be immediately answered by an AI agent that has access to every other conversation that's ever occurred on that site and my entire knowledge base and to be able to expertly answer that question within seconds. That would be super useful. And then beyond that, if I've asked that question, it's a topic that I might be, I might maybe have evergreen interest in it, right? And I'll talk about what that means in a second, but I have long-term interest in this topic. Well, what if my newsletter starts to, started to be tailored? What if I started receiving offers about educational courses and seminars that are related to that topic? What if I started seeing sessions of the upcoming annual meeting that are relevant to me highlighted, right? Just because I posted something on the community. So these are all things that can be done. They're actually very straightforward to do. What I mean by evergreen, by the way, is, and this is one of the challenges of these interest profiles that many associations still send out, usually electronically these days, although I think some folks are still, still setting out paper forms. I've seen scantrons at some associations where people are asking their members to fill those out. Yes, they're still there. Still there. Yeah, still there.
[00:12:00:23 - 00:12:16:03]
Amith
So now this is not the norm anymore. This is like the laggards who are, you know, kind of giving up their scantrons and paper surveys out of their cold dead hands kind of thing. But they are out there. They're probably not listening to the sidecar syncs. So I hope if you are, I'm not trying to offend you, but it's time to go with digital.
[00:12:17:10 - 00:12:36:12]
Amith
But with digital surveys, it's the same thing, right? It's still people telling you what they usually what they used to like. So there are a lot of opportunities around that. But evergreen is a very simple concept. It simply means something that has durability. And so I'll give you an example. Mallory, if I call up a company and I say, hey,
[00:12:37:18 - 00:13:23:05]
Amith
I'm where's my where's my UberEats order or where's my DoorDash order? That doesn't mean that I have a durable interest in food delivery. I don't want to know, like, whether or not other people's food is delivered on time. And that's kind of a silly example. But another example would be you call up a company, you're using their software and you're like, I can't figure out how to log in. Does that mean that you want to get articles about how to log in forever? Probably not. But if you're like, hey, you know, I do podcasts on A.I. and associations and you're talking about something like that in the community and you start getting really interesting stuff about podcasting or really interesting stuff about associations and A.I., that would be really helpful. And that's because A.I. is now smart enough to be able to tell the difference between something that is probably a passing interest as something that has some evergreen nature to it.
[00:13:24:06 - 00:13:31:16]
Mallory
That makes sense. This episode is another example, hopefully, of something that's evergreen, maybe not as long as the couple of years like we did for the last round.
[00:13:31:16 - 00:13:34:10]
Amith
It's designed to last for at least 130 episodes.
[00:13:34:10 - 00:13:35:03]
Mallory
Exactly.
[00:13:36:10 - 00:13:56:10]
Mallory
Amith, do you have any concerns, though, with creating these, I don't know if profiles is the right word, but ingesting all of a member's emails and looking at their community posts and then having A.I. kind of formulate a profile of that person and what they prefer. Are there any concerns around that or not really because it's all information they've provided to the association?
[00:13:56:10 - 00:14:47:02]
Amith
Yeah, I mean, this is a privacy question in terms of use questions. So first, first rule of thumb is disclose. Make sure you're telling people what you're doing and let them opt out. So if I don't want you to retain that information on me, allow me to do that or tell me not to use that information to personalize my communication. Sure, we can turn it off. But most people will be fine with it and they'll stay opted in or you can you can opt them out by default if you want to be extra conservative. But they're definitely it's important to be open about that because some people will find it creepy if you don't tell them that you're going to do that. And then all of a sudden, you know, their next newsletter has something in it that relates to something that they asked you over email. A lot of people will be very pleased with that or maybe not even notice honestly. But I think that it is incredibly important from an ethics perspective to explain to people what your policy is and what you're going to do.
[00:14:48:03 - 00:16:27:15]
Mallory
Just felt like that was helpful to touch on. Well, everybody, this first part that we just covered, that was basically the bulk of episode 12. So you're going to get a lot more in this episode today. We're moving on to who actually owns your data. So you probably think you own your data and legally you do. But here's the catch. It's scattered across a dozen different systems that don't talk to each other like Amith was talking about your AMS, your LMS, your CRM, your financials, your email tool, each one holding a piece and none of them sharing. So when you actually want to do something with it, ask it a question that spans all of it. You can't. And any practical sense data you can't reach and can't combine isn't really yours to use. The fix is to get it out of these silos and into one place you control. The current term we like to use for this is an AI data platform, a single environment where your structured and unstructured data can live together unified owned by you and ready for AI to work on top of. The idea isn't new. What's new is that it's finally realistic to do. So why is that? Amith, feel free to jump in if I get any of this wrong. But the old way of unifying data was a process called ETL. Extract, transform, load. Extract and load are easy, relatively. You pull data out of a system and you put it somewhere. The transform step in the middle is where the money and the pain live because every system stores data differently and someone had to manually map your AMS, your LMS and your CRM into one clean matching structure. That's fragile, high maintenance work. And it's why these projects used to run a lot of time and a lot of money. Amith, do you agree with all of that for ETL?
[00:16:27:15 - 00:17:09:03]
Amith
Yes, that is a great description and puts it in just basic terms. It's when you kind of take your data and make it play a game of Twister or something like that, where you're trying to basically get this thing to go way out of its natural shape and turn into something that it was never designed to be. And that also you lose resolution there sometimes. So you're kind of compressing the data and you're getting like summary data in and out of systems. So it's a big, big problem. And I wouldn't even say that it's what people used to do. It's what people are actively doing now. Most folks, not just associations, most large enterprises think that the only way to unify their data is through classical techniques, which includes a transform process.
[00:17:10:05 - 00:17:13:01]
Amith
And that's why it's exciting to be talking about how we don't need to do that anymore.
[00:17:13:01 - 00:17:44:09]
Mallory
So it sounds like the AI data platform drops that transform step entirely and it just copies each system's data over exactly as it is. And an AI layer sits on top and does the reasoning that used to take months of manual mapping, recognizing that a member in your AMS and a contact in your CRM are the same person and letting you ask questions in plain language across all of it. That shift from a year or years and a fortune to days and a fraction of the cost is what puts this in reach for associations that have never considered it before.
[00:17:45:11 - 00:18:16:21]
Mallory
Here's kind of another point why owning your data matters. It always has, but why it matters now. The software companies you rely on are starting to control how AI reaches your data. HubSpot has talked about metering and monetizing AI agent access. SAP is moving to lock third party agents out of its systems. The data you assumed was yours is becoming something your vendors may gatekeep and even charge you to access. So the associations that consolidate their own data have options and the ones who don't are stuck with whatever their vendors decide.
[00:18:18:02 - 00:18:28:05]
Mallory
So, Amith, do you think owning your data means something different in 2026 than it did even a couple of years ago? We could say pre-AI or even back in, you know, 2023. Is it different now?
[00:18:28:05 - 00:21:13:23]
Amith
It's a much more practical consideration now because there's things you can do with it. So if I were to say, hey, you know, like, Malry, you need to get all of your emails, you know, make it so that they're not just stuck in Gmail. You need to get all those emails. You'd go, OK, but like, why? Like, what am I going to do with all those emails or what am I going to do with like all the raw data out of my financial management system? And so it's not a backup strategy, right? It's a way of like, how do I activate that thing? How do I turn it from a latent asset to an activated asset and actually generate value of some sort from it? So that's what makes it more practical now is 2026, we have a toolset that's broadly available, that's inexpensive, that we can use across all of these different kinds of data to, of course, compare them the way you described earlier with vectors, but more broadly to actually do different things with this data. A good example of this is predicting the future. Now, part of what AI has been doing that's been so stunning for most people over the last few years is what we call generative AI. And generative AI is actually a type of predictive AI. When we talk about how language models communicate with you, they do something called next token prediction and they do. There's lots of different techniques for this, but the basic idea is they're predicting how they should respond to you based on their training data. There's also a reasoning process where they're kind of like looking at what they generated and then editing it. That's what's really driven a lot of the intelligence boom in the last 18 months is reasoning models, but essentially they are forms of prediction machines. But classical machine learning and AI before large language models and image models really entered the fold in most people's imaginations in late 2022 with chat GPT were specifically built prediction tools that predicted one thing. And so you would build a machine learning model, for example, to predict which movie Mallory might like to watch tonight when she logs in to Netflix or which product Mallory might be most likely to purchase when she's on a particular product page on Amazon. These are classical machine learning workloads and they actually haven't decreased in value. They've increased in value in many ways, but they complement what people do with language models and what people do with image models. But they've very different. They've been very difficult to build until now. You know, it's been it's much, much easier now to do the data science, to do the work, to prepare the data, to train these machine learning models and to put them into production. In association land, the examples that I'd give you are how do you predict someone who's been a longtime event attendee, let's say they've been to at least three of the last five annual meetings for your association. How do you predict what the probability is of them attending this next annual meeting? And if for some reason you think the probability is lower than a certain threshold, what do you do about it? So if Mallory has been a Digital Now attendee, Digital Now is coming up when Mallory?
[00:21:14:23 - 00:21:20:05]
Mallory
October 25th through the 28th in Washington, D.C. or in Roslyn?
[00:21:21:08 - 00:24:08:11]
Amith
Yes, just across the bridge from Georgetown. So we are so excited, by the way, about Digital Now. We think it's going to be well, we know it's going to be the biggest one ever. We think it's going to be the best one ever. It's going to be at a beautiful new Hilton Hotel that just opened up over there. Really, really cool. But if I want to know whether or not Mallory is likely to attend Digital Now, having been let's say she wasn't a prior speaker and deeply involved with the event, but just an attendee and I want to know, hey, what's the likelihood of Mallory attending Digital Now 2026 in October? It'd be great if I had a really high quality machine learning model that could give me a pretty good predictor of that. And if for some reason it flags that, oh, actually this year Mallory is only 40 percent likely to attend, whereas at this prior point in time ahead of the event in prior years, she was always at least 70 percent. Well, something's up. It'd be great to have this red light glowing on her record, so to speak, digitally. And then it drives some kind of business process to reach out to Mallory and say, hey, are you planning on coming to Digital Now? Don't bother her, like if she's already registered, obviously, but certainly if it looks like she's going to register. But if we could predict that, it sounds like sci-fi, but there's a lot of signals that we get in our interactions with our members. So going back to the email conversation, going back to what you started with Mallory in terms of text and audio and video and then also the concept of other modalities of data. If we can pick up some signals from all of this stuff we have and then use it to then predict downstream what may or may not happen, that can be very powerful for that intervening, providing more value, figuring out what's going on. Of course, these models are not perfect. They don't literally predict the future, but they are able to give you good probability assessments based on the signal that's out there. They're very sophisticated and up until now you would need a team of data scientists, people with graduate degrees, sometimes PhDs to spend months or years building these models. And first of all, the models have gotten better, just like the models we talk about all the time. This is a different class of AI models, but these are models that you train with your own data. And so there's all sorts of work that's involved in that. And now by the time this episode airs, I'm really excited to say that Member Junction has a new predictive studio that's launching, which is an agentic data scientist, essentially. Predictive studio ships with an agent that will literally do everything that a data scientist would have done for you. And data scientists, by the way, is still important. But like the basic role of what a data scientist would do, it's fully automated now in the box and totally free. So you can go as a business user and say, hey, I want to predict whether or not my long time attendees are going to come to my next year's event. And it will work on that problem sometimes for hours, sometimes for days. And it will figure that out. It will come up with models. It will test them against your data and it will give you a toolset that you can start to deploy. But you have to have the data feeding in both structured and unstructured is the point.
[00:24:09:11 - 00:24:23:05]
Mallory
And I want to clarify here because I feel like we've said it a few times. So we've talked about an AI data platform. That's what we're focusing on in this topic. Member Junction is an AI data platform that is free and open source and available to the association community. That's right, Amith.
[00:24:23:05 - 00:25:22:16]
Amith
That's right. Yeah, it's we've been building it for years. We've open sourced it because the goal for Member Junction is to bring the best and the secure and the most secure AI data platform possible for this market to every association and every nonprofit on the planet at no cost. So the idea is you can host it yourself. We do offer a hosting service called Member Junction Central, which starts out at a couple hundred bucks a month. But you can host it yourself for free. And there's there's tons of possibilities with this tool. It's a great platform that allows you to do all sorts of different things. There are other ways to do this as well. You can bring your data into lots of other platforms. You can use commercial tools like Snowflake and Databricks at the high end. You can use all sorts of other databases and get cloud code or cursor or one of these other tools to build tools for you. We just wanted to have a turnkey solution that was still super flexible, available for this market. So that's why we talked about it a lot. It is it's it's one of the things we do, but it's a totally free open source tool.
[00:25:22:16 - 00:26:06:13]
Mallory
Mm hmm. One point I always go back to it. I think you've said it on the podcast a few times, Amith, but it just really stuck with me is asking an association leader, you know, can you give me a full picture of Member A, every event they've ever been to, the emails they've ever sent, the courses they've taken, the certifications. And it would probably be really difficult for most people listening to this podcast to have a full picture of a single member that we pick out out of the whole population because the information on that member, the data is dispersed across all these siloed systems. I do want to ask you because I don't think we have touched on it yet. The idea with the AI data platform is not that it replaces your systems, right? It just exists in addition to them with all of that data flowing in.
[00:26:06:13 - 00:27:18:20]
Amith
That's right. Yeah, it's definitely complementary to all the existing infrastructure that you have in place. It is it is not going to replace your AMS or your LMS. It's designed to take the data from those systems and then bring the AI toolset that you need to go beyond what those tools do really into the forefront. So you will still run dues renewals in your AMS. You will still run event registrations in your event system. The AI data platform doesn't do any of that. What it does is it brings the data from all those systems together in one place. So to your point, at the most the most basic level, which is incredibly valuable, is just see it all together. So if I want to look at Mallory's record and see everything she's ever done with Sidecar, I can look at that in one place and I'll have all my data together in a nice, organized, cleanly displayed way. That's all part of the AI automation that's in the data platform. And to your earlier point, you don't need to do transformations anymore. There are natural ways to detect who Mallory is across these system boundaries that AI is smart enough to figure out now. So we don't have to spend gobs of money and lots of pain in getting that done. And then from there, we layer up and we can do so many other very sophisticated, very exciting things. But it's not a replacement for existing infrastructure.
[00:27:19:24 - 00:28:48:21]
Mallory
Well, you talked about those very exciting things that we can do with it. And that's a really nice segue. Thank you, Ami, to the last topic for this episode, which is what is possible when your data is activated and where to get started? So once your data is unified and AI can work across it, it stops being something you store and something that actually is working for you. The shift worth naming is data at rest to data activated. And the value shows up in two directions at once inside your organization and also out to your members. So internally, the obvious when is speed. The cross system question you could never easily answer. How many members attended an event last quarter and also completed a certification? Used to mean exporting three spreadsheets and stitching them together by hand. But with your data activated, you can just ask in plain language and get the answer across every system at once. Faster analysis, less manual grunt work and decisions that don't wait on someone building a custom report. But the external value may be the bigger prize. And this is where it gets strategic. Activated data is what makes real member personalization possible. Communications and recommendations tuned into who each member actually is and not that one size fits all blast. It's what powers member service that can give instant answers from your own vetted content instead of making people wait. Put those together and you're building something more important than efficiency. You're making your association genuinely indispensable to the member, the kind of tailored responsive experience that makes you hard to replace.
[00:28:50:04 - 00:29:11:20]
Mallory
So where do you start? Not with a year long project necessarily. The realistic first move we think is to pick maybe three to five of your most important systems, the ones holding your most valuable member and operational data and get them replicated into one place that you own. That alone, seeing your data unified for the first time tends to surface possibilities people didn't know they had.
[00:29:12:20 - 00:29:31:02]
Mallory
So I mean, I want to talk a little bit about that external value piece. We talked about prediction and we've kind of talked about data analysis. I feel like all of those are helpful to run the association, but they're still focused on internal running, internal operations. What does activated data actually let an association do for a member?
[00:29:32:04 - 00:30:00:09]
Amith
Well, the common challenge that I hear from association leaders, large associations and small alike, is that members can't get the stuff that they need. Members cannot find the content that they need. And I've asked this question to so many groups of CEOs, again, from small associations all the way to very large associations and said, hey, do you ever get call from your board president or your board chair saying, hey, I know you guys have this article or this
[00:30:01:10 - 00:30:04:21]
Amith
resource about blah, blah, blah topic. I can't find it.
[00:30:06:01 - 00:31:56:16]
Amith
And basically universally, they all shake their heads and it's frustrating and it's something that they really deeply want to solve for. So that's an example is like just finding resources. But how do you know which resource is most relevant to that person? Because the reason it hasn't been solved isn't because people are like, yeah, whatever. Like that's not the reaction association people have had. They've been trying to solve this problem diligently for years and years. They've implemented search tools. They redesigned their website countless times, invested crazy amounts of money in that over the years. Yet it's still a problem. And part of it is this generic search tools are not good enough. Federated search tools, which bring together data from lots of places, still not good enough. AI understands the person that's making the request. It's one of the reasons Google is better and better every year because it knows you better. So the predictions it makes from Mallory are different than the predictions it makes from Amith and so on. And so activating it first and foremost in terms of external value creation is let's take some of the pain away from our members daily lives when they want to engage with us. Let's get them the stuff they want and get them on their way. And so to me, that's a big, big part of it. It's this resource access conversation. Then the next piece on top of that that I think is a natural extension is predicting what the member might want and making sure that that is what you're sending them proactively. So even if they didn't think to ask, does your association have a ABC resource? You're sending it to them saying, hey, we think this may be useful to you. That's the idea behind predictive intelligence that powers tools like RASA and many others. So it really changes the game because it turns your association from kind of an obligatory part of the profession where people feel like they need to be part of the association and maybe they engage with you once a year to renew or to keep their credential up to date to something that they want to engage with, that they need you on a daily basis to your earlier point.
[00:31:58:00 - 00:32:39:20]
Mallory
We talked about agents a few times on this episode. I'm thinking a member service agent that can help handle the inquiries that come into the association or a data analyst agent or the predictive analysis agent that you mentioned to me. I'm assuming for all of those to work well, right? We want to make sure our data is in order and that we have good data to provide to the agent in that scenario, though, if someone's listening to this and they're on the path of creating an agent for their association, but they're hearing this episode and they think, oh, we don't have an AI data platform yet. Maybe we should stop everything, get the AI data platform, then get the agent. How do you recommend thinking about kind of all these different paths someone could take and as how they pertain to data? I guess.
[00:32:39:20 - 00:34:39:05]
Amith
I mean, you can build agents without a data platform. You can have these agents directly connect to your source system. So I can build an agent that directly connects to Salesforce or HubSpot or my AMS. You sure can do that. The problem is, is it becomes this really unmanageable web of connections and then the agent has a really hard time knowing what's good and what's bad and how to make these things make sense. So what you really need to solve for, in my mind first, is to unify the data to be able to put kind of a layer of intelligence on top of the data itself. It's part of what an AI data platform does is it brings a higher level of understanding of what the data means, not just through the vector work, but also through this concept of a data dictionary, which is to basically have what a dictionary would do for a language. Right. It's essentially definitions for everything, saying this is what this piece of data means. This is what this piece of data means. That's also AI generated, by the way. You don't have to go build that manually, which is what you used to have to do and why it never got done. So it's something that makes agents smarter. So if you try to run an AI agent just directly against your raw sources of data, it might work and you might have some success. But how do you know it's correct? How do you know that it's actually checked it against multiple different sources? You need to have a foundational layer, in my mind, that really highly increases the probability of the correct answer coming from the agent and also make it so the agent is more consistent, more reliable. There's also a security question because you really want your agent, most likely in my mind, to kind of live in a box. I don't want my agent to kind of troll around my entire information systems across every system I've got. I'd rather have my agents kind of in a safe place where they have what they need and they can do the job that I hired them to do, but to not have access to all these other systems in kind of a more ad hoc way. So I think it is possible to do it without a platform. But I think you'll run into quality problems very quickly and you'll run into scalability problems, as well as some potential information security concerns.
[00:34:39:05 - 00:35:01:13]
Mallory
Mm-hmm. So someone is still with us listening to this nice, refreshed, updated data episode and they're thinking, you know, we really probably need to get an AI data platform and get our data unified and in order. Can you be realistic with that listener on how long does something like this take? How many people need to be working on it? What goes into that process?
[00:35:02:13 - 00:36:03:23]
Amith
Well, I can speak to it most effectively by talking about Member Junction, which is our AI data platform that's open source and you can implement yourself if you want. You can hire a consulting company to do this for you. You can do it yourself. We've got a team that does this type of work as well, but there's a lot of people who can help you. And the essence of it is you set up the software, which is like installing any other program. It's a little bit more involved because it is a database. It's got some infrastructure. We do have a hosting option where you can go to Member Junction Central and it'll like within a few clicks of the mouse, you have an environment up and running in the cloud. So you don't have to be technical to do it. And then you have to figure out how to get your data in there. And Member Junction itself has an integration architecture, which has connectors for many dozens of systems that you regularly use in the association market. That's part of what makes it an association AI data platform is it's got connectors for a bunch of AMS's and Ls. is stuff that you actually use and you can bring that data in.
[00:36:05:00 - 00:36:51:03]
Amith
And then once the data is in there, you need to keep it up to date, which is part of what the integration engine does is it incrementally sinks the data for you. So that's the basic process. If you know your earlier advice of picking a handful of systems to get started is what people want to do and they don't want to boil the ocean and try to bring in every data source. I agree with that. You're talking about a matter of weeks. It's not months or years. Many people can get up and go up and going with a useful initial implementation in say six weeks, sometimes a lot less to actually physically get the data in is pretty trivial. But then you have to go through this process of looking at the data. You will initially be shocked at how bad your data is if you've never actually seen it. It's it's quite eye opening, but that's actually good because it's helping you get a realistic sense of what the data looks like.
[00:36:52:07 - 00:37:46:23]
Amith
And by the way, to mention earlier what you were talking about with unstructured data in MJ's case, it also has connectors for all of your unstructured data sources be that, you know, whether it's a SharePoint site that you have or a website or if you have data in Google Drive or in box.com or Dropbox, MJS connectors, all of that so that the agents can reason across both structured and unstructured data together. And, you know, in this environment that I'm describing, it's designed for this industry to make it easy. You're literally talking about a handful of weeks, Mallory. It's not a long arduous process. Now, if you want to build something custom or use something that's, you know, from outside the sector, it'll take probably a bit longer. But it's it's still totally doable. It's not an AMS implementation. It's not like a data warehouse is what a lot of people try to compare this to. It's not at all like a data warehouse. It's so much more straightforward and so much more natural, actually, for people in this industry to understand what's going on.
[00:37:48:04 - 00:37:52:11]
Mallory
I was going to ask you, is that weeks in a meath time or in an association time?
[00:37:52:11 - 00:38:36:17]
Amith
No, this is in real time. This is this is not like in my brain. This is this is what people are actually doing out there. We have a lot of people that we're talking to that we're helping. We've got people outside that are kind of self hosting and rolling it themselves. But it is a realistic thing to do. This particular piece of software has been around for years now. It's it's pretty well refined and it's improving at a rapid pace as well. So the goal here isn't to pitch member junction, although obviously I'm talking about that a lot. Because that's what I know. At least that's what I've been working on for years with this project. But there are other ways to do this. The point is bring your data together, put it under one roof and then build your A.I. agent strategy and your machine learning strategy and all these other things on top of that unified data architecture because you now own your data operationally, not just legally.
[00:38:38:14 - 00:38:58:13]
Mallory
Well, we're talking about at the top of this episode how your whole career has basically been built around data and specifically focusing on association problems and opportunities with data. Is there one common misconception that you hear when you're speaking to association leaders about data in general that you want to set the record straight on right here on the podcast?
[00:38:59:13 - 00:39:25:00]
Amith
Yeah, I'm glad you asked it that way because it triggered a memory for me that I do see a lot of patterns. And the number one pattern is people acknowledge that this is a problem and that they need to solve it. And people generally like this direction. What I often hear as something that's not an objection to the idea, but something that slows people down as they said, well, our data is so bad, so, so bad that I can't bring it in and do anything with it. I got to clean it up first.
[00:39:26:08 - 00:41:51:11]
Amith
To which I say, is that a new problem or has your data been bad for a while? Usually the answer to which is, oh, no, no, it's been bad for a long time. Like we've been struggling with this for years. And then I usually say, well, and you presumably you've made prior attempts to clean it up that you're describing doing now in order to get ready for A.I., right? And they say, yeah, we've tried a bunch of times. We've hired consultants. We've spent money on it. We've had staff to work on it. It's usually exasperation and exhaustion, you know, as in these folks voices. And I'm, I'm deeply empathetic to that. It's such a big problem. What I would say to you though, is that now is different. And the reason is, is that previously you were using hand tools. If you had any tools at all, you probably really, the best analogy is that you're using your bare hands to try to solve your data problem. And maybe you thought, OK, I'm going to pick up some hand tools to try to solve the problem. But if you're trying to dig the Panama Canal, you're not going to get too far with your bare hands or with hand tools. You need new tools. You need power tools. You need tools that haven't been invented until recently. And those tools are here and they're all in MJ and they exist in general. They're A.I. tools. These are power tools that, you know, make your hand shovel look extremely primitive. Right. These are tools that are like super powered lasers that came from. Some sci-fi novel in comparison. So the tools exist in the A.I. data platform to help you clean your data, to help you get your data ready to do all these other things. If you try to clean your data somehow before you get it to the A.I. platform, you know, you're inverting the process from my point of view, because the whole the whole tool set you need to actually solve the problem this time, to actually not feel exhausted and exasperated by this issue actually do exist. But you have to drop your data in as it is. It can, you know, come as you are, basically, is what I would say to your data. Don't worry about, you know, getting dressed up for the occasion. Just just get yourself over here into the data platform and we'll help you out from there. And the tools are in there. I'll give you one concrete example, Mallory, and that's duplicate management. This is the bane of existence of a lot of membership managers. They that Mallory calls up or Mallory emails and they look up Mallory and there's eight Mallory records. They're like, oh, which one of these is Mallory? And it's such a pain. It's a massive internal friction. But it also breaks a lot of websites because they're we're not sure who you're logging in as. So how do you detect that these are these different Mallory's are really the same people and not incorrectly identify that, oh, wait, this other person's actually a different human.
[00:41:52:14 - 00:42:45:18]
Amith
And that's something he is extraordinary at. It's really, really good at detecting whether or not these people are the same or whatever duplicates are, are the same or different things. Vectors help with that. And then LLM reasoning on top of those vectors makes it really, really accurate. And so don't try to do that manually. Use AI to do it. And then the AI will help you kind of send signals back to your source systems to say, hey, this is a duplicate record. Some source systems, by the way, can't do anything about it. You just have to kind of like market as a duplicate. Some source systems like some AMS's dual item merge records and stuff. So the first part of the process is detecting these duplicates. And that's just one example. There's tons of other stuff like is this person does this person work for this association anymore? Does this person have an incomplete address and their solutions for all these things in this AI data platform and with AI in general? So don't worry about coming with your data all dressed up for the ball. Come as you are.
[00:42:45:18 - 00:43:05:09]
Mallory
I love that. I love that. Come as you are with your messy data and all. Hopefully from this data episode refresh, you've walked away with a few things. Knowing what data you actually have. Understanding that owning and unifying it is real leverage now that agents are in the mix and taking the first step to get it working for you instead of sitting idle.
[00:43:05:09 - 00:43:10:24]
(Music Playing)
[00:43:21:17 - 00:43:38:16]
Mallory
Thanks for tuning into 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.
[00:43:38:16 - 00:43:41:22]
(Music Playing)