 Well, this is exciting. This is our first ever online chat that we're gonna put out to the world So I'm really excited and my name is Tony Cappuccini. I'm one of the co-founders of Parsons TKO We have been having for four or five years now Fantastic internal conversations just like this over zoom and then we realized we should probably start recording these And getting them out there for the rest of the world to get a look at what we talk about sometimes You know, this is an informal setup, but informed discussions So that's really what we're shooting for here You can let us riff and then you know, hopefully if you like these you'll share it around and maybe we can start getting some Other participants in here with us too, but I'll let my colleagues introduce themselves as well start with Stefan Who's in my upper left corner. Hello, I'm Stefan bird Kruger. I'm the chief analytics officer for Parsons TKO Leading the data strategy team We spend a lot of time working with our clients thinking through the types of problems that they face and the ways in which they can use data to address those I think our team has a has a lot of sort of unique conversations around the unique circumstances that our teams encounter and We end up finding a lot of novel problems and novel solutions to those problems So I think having a format where we can talk more about that is as really exciting And hello everybody, I'm Nate Parsons the other co-founder of Parsons TKO and you know I echo my colleagues thoughts that you know We've had a lot of really interesting and thought-provoking discussions that if you know started often with our own problems or hearing about things that our clients are Interested in and you know taking those to really interesting places But a lot of that knowledge has been ended up ended up locked inside organization And we haven't had the chance to bring others into the conversation So I'm also really thrilled to be a part of this and to sort of open up You know some of our musings so the wider community so for our first show Stefan had just come back from being at the data X conference in San Francisco where he moderated Two days worth of panel discussions about machine learning So I thought that might be an interesting place to start You know what were the experience of the count and Nate also attended the conference too So he's got the attendee experience and Stefan has the looking out of the opposite direction experience So I don't know if there's something big that stuck out to y'all from the conference Maybe to start with you Stefan. Yeah, you know, I think it was it was it was very interesting So I mean this the data X San Francisco conference in particular. I was chairing the machine learning track You know a conference like that. You're talking about several hundred practitioners a lot of data scientists a lot of data engineers Coming together to talk about the latest and I know Nate got to see some of the other tracks But you know in particular the machine learning track I think a lot of people came to sort of hear what's new What's the latest in machine learning and how can they use machine learning and and what does it take to use machine learning? and a lot of the presenters themselves were also technical one of the things that really surprised me though about the presentations was how No matter how deep they got into the details looking at different algorithms Even you know specifically thinking about the formulas all of it came back to you have to understand your problem And you have to understand how you're going to use the results of machine learning even as the pieces change that sort of Central strategic foundation is unavoidable and and I think a lot of them You know when they see their companies or they see people in the field failing with machine learning It has little to do with the technology and it has a lot to do with misalignment with the problem I you know given how often we encounter that with our clients It was really nice to see that no matter how small or how big everyone sort of comes back to that same issue Yeah, in terms of scope size there, right? So we're we focus mostly in the mission-driven sector, you know We have some of you some bigger nonprofits, but I mean some of the folks you were speaking with were Uber Facebook LinkedIn so these are massive organizations, right? And it's interesting to see that scale to your point right from small to large You got to focus in on the alignment of the goal that I think is one of the reasons why You know why this this conference was really valuable and I got a lot out of it that I think our smaller clients can use Because the challenges are so human and it's you know big or small It's always humans that you're working with and so figuring out how you can empower your staff with these capabilities You know machine learning isn't a replacement for the work that you need to do It's a way to enhance that work. And so so I think yeah, I think it was it was really it was really nice to see How big organizations face a lot of the same problems and then hear how they address those how about you Nate? We're gonna take away. Do you have from your time at D to X? Yeah, I mean one of the things that really struck me that was really interesting was this the sort of concept of fairness and breadth of Discovery, you know, I think in a lot of the you know AI machine learning space There's this kind of intrinsic assumption at least one that I had where the tool kind of figures out the very best thing for you and gives you this One best example, right? And it saves you all this time or energy because you're just presented with the perfect answer, right? Or the best thing for you, but what you know, I was what we found or what I heard in a lot of the presentations was how You know getting the breadth of experience and finding the sort of Right area to expose with these algorithms was something that all these companies are struggling with so like for an example There was the the dating app hinge had a presentation there and they were saying that, you know when they first started working with machine learning their algorithm identified the sort of Prototypically most attractive people in any particular geography and they could get initial engagement with their application by just showing You know the most pretty people to the first people who signed up for the app But that actually wasn't servicing their business very well or the users because people weren't actually seeing the whole breadth of their Userbase people weren't really connecting with each other and the people who are being shown the most were being overwhelmed and not really having good experience either And so they had this kind of initial min maxing problem where they maximize initial engagement with the app But it wasn't really working and they had to go back and figure out with their machine learning algorithms and their AI Suggesting algorithms how to not just show people, you know The very very tip of the top but find the right breadth of people to show in each geography And how to make sure everyone in their user base actually got a fair shake and got to be sort of exposed to each other So that they could make those connections and build a sort of deeper and more full Sustainment and that was really interesting uber on another similar case showed off their Process for helping uber drivers select the best response to uber riders when they're contacted So when you write an uber driver, what happens is that a machine learning algorithm Analyzes your text message and tries to figure out the intention of it And then it suggests what it thinks is the best Reply for that intention to the uber driver and they can just send that But if you don't have the right one you can kind of back out of it a little and see what it thinks are the mere neighbors Or the best closest other things to that kind of intent that you might want to say and I thought that was really interesting You know, they started out just wanting to make it like oh if they say X reply Y But then they realized well, maybe there's no nuance needed or the ability to kind of like shade it a little to the left or right That was, you know Pretty eye-opening and I think a lot of that gets back, you know sort of again at that principle You know machine learning algorithms are very good at finding answers to questions But they're not particularly good at figuring out what question you should ask and so I think you know in a lot of cases I mean what one of the One of the presenters that I think are especially relevant to a lot of the organizations we work with was Bloomberg sort of a big publishing outfit and You know, they use machine learning for a lot of the same things that our clients would figuring out what tags They should have figuring out how to use those tags figuring out how to personalize and recommend content to their users And one of the things they ran into as problems with overfitting And and so I think a lot of times they would overfit the solution to their their clients and one example of that is If they tried to recommend content based on Popular topics they quickly started to find that they were recommending Popular topics and those topics were popular because they had recommended it and so being able to sort of understand What you're asking your algorithm to do and understand what inputs are appropriate for that algorithm because the machine learning algorithm It's just going to run with it. It's not going to ask you whether or not, you know You're it's a doing what you meant for it to do So I think I think that sensitivity is important and you know, because there are so many ways And I mean one of the things we'll talk about is, you know How many different types of machine learning are out there now and different ways you can implement it So Bloomberg has actually developed a recommender system for their recommender systems So how can you use machine learning to decide which machine learning approach to use for a given person? So, you know, I think we're really seeing some a lot of growth in the field a lot of growth in what's possible and Yeah, yeah, nice to see how different organizations navigated that So, yeah, I'm wondering how we we take this down a level too, right? So this is these are groups that have already gone there experimented had Had findings of their own pulled them back in right like the dating app Which also reminds me what you said on Bloomberg, right? It's like if I keep showing this thing Well, then of course the popular thing stays the most popular thing because it's the only thing you're seeing, you know I remember Might have been like ten years ago when it was first like and we can have related content And we could do a trending piece of content on our website and it's like well And how do we determine which piece should show at the top and if we show that piece at the top? Well, of course, that's the popular piece But that doesn't actually help the audience right get down to like I think you always talk about innate Like what what's relevant might not be the most vibrant piece of content You have not the not the thing that happened today the thing that happened ten years ago, but that'll never be there, right? Is that is that something machine learning can help with was like and where would we even start but so one of the areas I see people tackling that kind of problem in particular are in these kind of guided discussion Experiences online that are assisted by machine learning and so like a one that's been in the popular news a lot is called Do not pay which is a thing that will help you contest your parking tickets or you know You're other kinds of not moving violations and in essence what somebody did was they said well the conversations So you'd have with like a traffic lawyer or someone else is going to help you get out of that ticket are pretty you know Well just understood right they'll pick different routes So they'll take different approaches depending on your answers and what the situation is but there's kind of a guided discussion format for figuring out like how to address your problem and this guy basically made a You know online machine learning system where you go through a wizard that kind of walks you through the interview about What's the deal with your parking ticket and then it suggests a way and in some cases actually automatically adjudicates that on your behalf It'll actually fill out the forms and go and do stuff to try and prevent you from having to pay that parking ticket And all that is based on this idea that there was a human experience That's being translated into the machine world and the human experience helps figure out some of that intent And I think that's where I see the biggest benefit for a lot of nonprofits that have an approach or have direct service Methodologies or things like that which is that you know AI is really good as you know Stefan is saying of figuring out the answers to questions once they're poised and you know if you can Develop a script or a system that says well Here's the kind of interview we want to get but then we need something that can quickly sift through the answers to that and suggest Something that's really relevant. That's a really good place for AI and similarly with search I mean I think that's another place You know AI is really good or machine learning is really good at classifying all of your content But you know that it needs a little help to help you find the right thing And I think that's where chatbots are really interesting where the chatbot can kind of ask you questions that actually uncover your intent And then it can use that content to find the best filtered contents to suggest to you once it has a little bit more of your Intents in place and you know in our search You know sort of customization and enablement service for our clients That's one of the first things we do is help them understand it You know Google's kind of taught everybody a bad way of searching Which is that there's just as far and you type whatever into it and then it does its best where you know The reality is for most people with structured content or more in knowledge about the content They're indexing you have a lot of intent You could figure out and filter on and use that to kind of provide really great results And I think we're seeing the marriage of like really powerful search and smart machine learning slash AI to kind of connect Those two and get the filters set down even better and also to find the content and describe it without you having to Do that manually, you know, there's tons of stuff using natural language processing right now Which helped figure out what the not just the words in a Documentar but the intense the sort of pseudo meanings the kind of meta knowledge that's in those documents And you know with tools like TensorFlow and other things you're seeing machine learning applications being built that really you know help people Marry up intent that you've discovered through like a chatbot or whatnot with indexed content that you could present to So it's that you're talking Kind of quick that it's TensorFlow. Yeah, so TensorFlow is just one of a variety of sort of machine learning Libraries basically that have been created. There's a variety of them out there that help you take things like, you know tags on content or parts of language within a natural language processing Sentence like you know word or sense that somebody says and actually tie that into models that say if we think the intent is that they are looking For the nearest clinic do X and so that's kind of where the rules-based engines come in is using things like TensorFlow that Connect to things that have figured out some of the information from the user And that's probably a little worth talking about one of the things I kept hearing over again Is there's this kind of workflow that these organizations are building where the first one is kind of you know understanding and Breaking down what somebody's asked to the machine learning or AI algorithm to do like what's your request like in essence Let me figure out what you mean by your request Then there's a second piece of it which behind the scenes has kind of Bucketed and libraryed and indexed all of the different things you have that could be answers to some of those requests or some of those Intents and then the middle piece the mapping piece are things like TensorFlow that kind of help figure out like okay The user has this intent and I have this amount of responses What's the best one and should I give them an exact response or a couple of options or say you need to help me refine down because there's Still too many and I couldn't figure it out and so there's a kind of the three pieces right there's this like well organized index of answers this Informational system that helps figure out the intent of the request and then there's a matching algorithm in the middle that kind of says Here's the very best thing for this situation You know and there's a bunch of different ways to solve that I think that's exactly right and Nate you settled on precisely the right word for a lot of the organizations We work with which is intent and I think intent is particularly hard for a lot of our clients You know some of these big Companies that were at the conference have it easy because they know exactly what their users are there to do You know you think about uber they have you know essentially two products And it's get people to take rides and get people to get food delivered I have that laser focus on what they're providing for their audiences But a lot of our clients especially you know you think about think tank clients Where each of these organizations can have dozens of centers And each of those centers can specialize in dozens of topics And so you have these hundreds of different combinations of things that they can provide to their audiences and different structures and formats for For each of those so understanding of all the things you can do for your constituents Which of those that particular constituent intends to do is is really hard And I think that's where machine learning can make a big difference for for not on profits Yeah, just to tie in that a little bit You know what the other interesting use cases that we're starting to see is the ability for AI to create Product variations or content variations based on the person's intent So like the most simple one of these are like summaries of YouTube videos or summary of some podcasts where the AI Tries to figure out what the podcast or YouTube videos about and provide a kind of synopsis to people up front You know one of the things that came up with is the idea of proxy metrics, you know, if something like mission impact is You know hard to measure and hard to connect back to individual experiences online What's something that you can measure? What's a digital footprint of mission impact? Maybe even before it occurs So rather than somebody found their way to a homeless shelter, maybe it's somebody filled out the form to apply for you know something And and so I think it's really really important that that we figure out What are those things that we want to optimize for so that machine learning algorithm doesn't run away and just sell everybody's shoes Regardless of what you're trying to do. So it's almost like the analytics are a gatekeeper on top of the algorithm Like is this the piece that I really want me. Oh, we've got to have a we have a frozen Nate Yeah, just to tie on that a little bit I mean one of the other things that I think is is interesting and related to that is this idea of using, you know You know machine learning and AI technologies to help kind of figure out what's happening in a landscape or community or an area That a nonprofit or mission of an organization wants to work in or is working in You know, I think, you know, there's I forget the name of it I saw that there's one organization that's working on pulling in all of the public sector data and Analyzing it to let nonprofits ask questions of that data and kind of say things like, you know What's the access to health care in this community or what's the at-risk factor? and there's another one that I saw that is developing a Likelihood of dropout for high schools and middle schools for students to help them identify the students who need the most support Algorithmically, you know and things like that to kind of help cut down on the amount of in-person triage They would need to know which students are at risk and whatnot You know, so I think that's another interesting area that's kind of tied into the analytics, right? Which is like, you know, there's an idea of like how much impact you're having or how much engagement with your attempts to change are happening and then there's also What can we tell about the world that used to be to, you know prohibitively staff time intensive or person intensive to figure out that AI and ML can help you figure out and I think that's a pretty Interesting and untapped area, you know this idea of what sensing tools are out there because of all this public information Or all this, you know aggregate information that this stuff could help you develop signal out of. Oh, yeah on that student one though, I mean how Personal is that is that are we getting in a PI territory with that? Is that you know what I mean? Is it actually? It's got to be rules around PII especially for students So you're saying they're like this thing can help me find out that Nate Parsons needs my help in San Francisco or it's just like no, no this area of the mission in San Francisco It's very personalized and it definitely gets into PI areas really quickly. I mean, that's one of the real challenges I think for you know this world going forward, right, which is that Each person's use case is relatively unique and especially when you get into things that are personal like health care medical care Education job applications anything like that, right? Like it has to be unique to you to be really valuable in a lot of ways But of course as soon as you do that you're creating Really risky area, you know and we see that already with things like Facebook where they have that information And they made it available to advertisers to buy and use against and so people were able to target their messages Against nearly, you know, very small groups, maybe not an individual level, but pretty close to it in terms of that kind of PII and it was like lightly anonymized But I mean, you know, you can see a world where you know If somebody hacked into the school database in the future, they could learn an astronomical amount more than they can today, you know Okay, we're getting on like there's like ten more conversations. We can start erupting from all these different threads here I'm trying to see if I could bring it back into a single conversation. I mean Yeah, the PII thing just gets me there, too. It's I mean, we're You know, there's two threads in my head right now one is getting your organization prep because If technology goes away, it always goes right machine learning and AI it's here. It's only gonna get faster It's gonna get quicker and it's gonna be here before you know it like how worried should people be now and You know, what are the pros and cons benefits for internal staff, too? You know, we always talk a lot about an audience coming and being able to complete a task faster or Have these tools that can sit for attentionality and get me to the right page But does that also help staff burden on the other side? You know, if yeah, I don't know if you guys want to rip on that a little bit Yeah, I'll just throw something out there Which is like one of the Altruistic or best the best things about you know AI in particular is its ability to help you change behavior and to do Coaching and do things like smart reminders and kind of help people Change the way that they behave and you know, that can be used for good or ill But generally within organizations like you'd be really good, right? Like I think a lot of organizations struggle with staff time and over commitment and like trying to focus on too many things And something that can help instantiate good cultural habits and help coach people and improve the way the organization works through that That is really valuable and you know, that can happen in really small-scale things like hey Make sure you send this thing to this volunteer is a thank you to really large-scale things, you know Which are like hey, you know the organization last month spent this much time on outreach and this much time on you know Under management and that's actually out of whack and it's gonna cause a problem You know like those kind of things are both very possible in the near term So yeah, and the thing I would say to that Tony is I think machine learning can change what people within these organizations spend their time on I think it can it can help change and it can also help People and organizations make their decisions with more confidence I mean, you know, how many of our clients have somebody whose job it is to decide what goes on the home page And you know, they like to control that that experience I think that's very good And I think we're gonna continue seeing that for a long time Unless you know, we can really show that machine learning does a perfect job of it You know up to the standards of the editorial team But how many other landing pages are there throughout any website that aren't being as curated quite as carefully And I think machine learning lets you take that judgment take that You know, you know good taste when it comes to curating those experiences and apply that to your entire site all at once And then your editorial staff start to spend more of their time thinking about what is the experience We want to create rather than going through all the clicks on the website trying to configure it one page at a time So it just sort of it takes it takes the best of what we do and it takes our best ideas and it helps us amplify those I think I think that's the potential So not not a lot to fear other than making our jobs more interesting and scaling our impact Yeah, and I wonder it's less about the fear or more about are we talking this is a Transformation that I can work into or is this like oh crap I got to stop the way I'm doing everything now because in two years things gonna happen and like I don't have anything Correctly put together and just from our experiences. We know this right like mission-driven sector Big budgets come for that once every three-year wet build Right, which becomes the impetus for the change of everything and then by the time they're done Technologies already so far past when we've seen this with marketing automation like nobody's still really doing it She's super reportable and it's been there No, that's a really a really good point and actually that gets out another one of my big takeaways from the conference which was the importance of modular systems all of these companies have moved themselves towards a more modular setup where Each piece can stand alone on itself and it has you know as a you know Nate Nate describes very well The data contracts between each of these systems so that when a new piece comes out You can swap the the benefits of a new system in cleanly simply without having to redesign the whole website And you know, I think we already have that thinking and a lot of organizations already have that thinking sort of between You know the website and the email marketing those can be separate. You can have a separate donation system But increasingly we're starting to see that within platforms And and you know systems CMS is like WordPress are already well built for that You can have plugins and so you can build a custom plug-in to help manage just one component And really leaning into that trying to use those capabilities In order to manage how your whole organization works is so much better than just hard coding everything in Or or trying to build one system to rule them all And you know, I think that sort of future future-proof nothing can future-proof you better than modularizing your setup Even within the the machine learning workflows that these organizations have they're talking about Modularizing within that so if you have a process and you have a particular algorithm That's done an okay job at you know personalizing a feature on a website For example, you can in the background test 10 others and continue testing them and see if you can keep tuning them And as soon as you have a winner you just swap it in and even you know from a sort of technical perspective There's a lot of work on on making that platform and language as agnostic So if your whole you know technology stack is written in Python for some reason You can swap in a piece that runs in Java or you can swap in a piece that runs in C And so making it so that you can leverage all of those different strengths and Compensate for the weaknesses of whatever system you have. Yeah, I mean, I do think there's a There's another you know, we see this across the order, you know the industry But there's a sort of consolidation movement that seems to be picking up steam within the nonprofit space and I do think You know organizations that you know start investing and figuring out how to do these things now I have a much better chance of being the sort of winners in that consolidation space Because you do see that once people build a good self-service tool or some other way for the AI or ML systems to help people online Those become very popular and instantly kind of elevate those organizations towards the top of the pack, you know You know, there's one that helps, you know People who have been victims of like police profiling and things like that, you know report that online And it's become the de facto standard for that simply because they made it for one city You know, and I think that shows the power of these things and so it's not necessarily if you don't do it You won't be able to succeed But I do think that you know, these are certainly an area where it's going to make a lot of winners and losers because it's going to Change the way that people perceive how services are offered, right? Like if before you had to call someone and they call you back and they contact you and you had to work your way through a Process and hopefully their expertise was available then, you know depending on your time zone or where you're at Or you can go online and have a really curated and you know Powerful experience anytime any way that it seems to get better each time I mean those organizations are going to get more of the share of people who need those services or trying to find those services And so I really do think it's another place where you know, the people who make an early investment in this are going to Be way out in front of the people who don't So from having a way through Small business affairs, but the DC government's website is there any way we could pitch this to them so they can get this because their Websites are possible to find information on even when you're within certain contexts. It would be super helpful I can't imagine. They're the only government site. That's like that Let's hope the government adopts those teeth And would you be able to pull to get that link to the what you just mentioned about the police profiling system? Yeah, totally. Um, make sure to get that the show notes Yeah, it's called Ricky.org and I'll make sure to throw it in show notes. Cool. So I think we need another Talk and video probably really just getting into this democratizing data and how to do it, you know that It's interesting right like has to be woven into the DNA of the organization And I'm just thinking of well if I had access but I didn't really like stuff this program And I wanted money from my program and I can go in and pull my own insights out to make stuff This program that looks so good You know, we'd like to think that everybody's pulling in the same direction But you know, I think we have seen and it's something we like to take head-on It's why people hire us, you know that there's ego in these things that has to be shaken out and understood You know people are working in these fields because they want to get something out of it and they're driven You know can work and data drive you to do Can it can it actually be you know because I would think on the other side could it be an undermining force? If too many people were in there playing with it like then what's the actual insight that I'm trying to take away Well, I you know, I think well first and foremost I always have the highest faith in humanity and so I think the situations where we get Nefarious use of data are going to be fewer, but I think that democratization of data is a defense against that And you know to Nate's point before about what are the defined processes? How do we actually do this? I mean, this is something we actually already delivered to a lot of our analytics clients is our data catalog our data inventory and our data governance Documentations and and I think within that if you say a part of the process of Reporting and sharing insights means you also have to share your methodology So actually link us back to the report that you use so you sort of show your math And and I actually think making sure that you have it opened up like that So that anyone can go in and double-check the numbers and sort of you know Even not not to challenge, but just out of curiosity say I want to understand how you did that How you came up with that result it gives you more opportunities for people to catch one another not in something You know aggressive, but even just in very honest mistakes, which are very common in analytics So being able to say aha well you did it this way But if we do it this way then we'll actually get a more consistent number or a more global number or now we've yeah We've we've made the analysis you did more generalizable So that we can apply it to all of our content and not just the one case study that you wrote So I think I think it really helps data grow and it keeps people honest and it keeps the insights growing Yeah, one thing I might just tag on to that is that you know that question I think Tony is sort of hinting at something else Which is that there is a need for grand strategy in every organization, you know And it's funny how many organizations look at data to help them come up with their strategy versus to manage or run their strategy And I think that's where the data fudging and the kind of you know potential and accuracies that people want to introduce the data come from you know like If your organization is using something like ok ours or some other system where there are hope you know Objectives that are publicly known for both the organization and departments and maybe individuals and there are you know key results like numeric Goals that are attached to those that we people say our signs of progress towards the objective That's a really good way to kind of narrow the argument in the focus about what people are doing with numbers and what they mean And where they go, but you know the key there isn't like the ok our process although that's a nice one It's it's about The organization having a defined strategy and using data to support that strategy Not hoping they will find their strategy looking in the data, and you know We find lots of organizations when they first start coming analytics They're using what I call vanity metrics, you know Which are things that that are like downloads or things like that But don't actually measure or attached to any of their strategic goals or mission and you know That's where the road to ruin and data and automation both lie You know you don't want to you know look at data or automate anything that isn't focused on a mission or a strategic goal Yeah, I mean all that seems right to me And it's why is this such a ripe conversation because it's I actually think you call it creating a culture of analytic stuff And but I think there's a deep culture change and shift that has to happen industry-wide And then get understood within the mission-driven space, you know the for-profit larger organizations that were speaking at your conference They actually have the ease of having a bottom line Where you have a mission right and then you got funding from X different sources, and they didn't really tell you what the Metric you're supposed to give a back was so you're trying to figure that out So I think it gets a little harder to do that, but then it also takes the shift of You know I and in the space for 20 years, and it's the same three metrics that have gone to every single board when I worked in house right or two really page views and as a page views and unique visitors What's that telling anybody that to do anything for the strategy organization that makes everybody feel good? so I think there's Challenge to the status quo that has to happen like the status quo is going to do nothing to help these organizations evolve and move forward You know in my takeaway from this conversation is we're talking about machine learning and AI At least we started that and we got into a whole bunch different directions as the size conversations go That sounds big and scary to a lot of organizations and how the heck do I get there? And I'm trying to just make sure my site's mobile responsive But it's one of those things where if you're not ready, and you don't embrace it And you don't look out it will just happen And so you know change is either going to happen to you or you could be an active part of it in the transformation So I was that's like this bigger moment, right? I could you could see what the data could do And then it's how do we train and teach people and get them on board to tell them it's going to be okay? And some of the roles might need to change in these organizations, but I think you know We got really big on like a lot of different topics and machine learning AI it comes up and we hear it in Conversations, you know one of our big clients. They talked a lot about their all staff and we're working on a web build with them Like is this a setup to get there? I mean, I guess if I was an organization listening to this or Where would I start like okay? This sounds cool. I'd like to be in that future But here I am what? Where do you begin? What do you do with this? Well, I mean I'll start on that. I mean I'll say to how intimidating machine learning is That is changing in a in a big way, you know machine learning used to be a multi-million dollar proposition for a lot of organizations And I think many things are changing in a big part The the experts at a lot of these for-profit companies have been as they work contributing to open source libraries So a lot of the capabilities are themselves getting democratized So if you are a very experienced or even moderately experienced developer There are libraries that you can go get now and you can you can play with them you can use them and so it makes it a lot easier to incorporate machine learning into Existing tools existing platforms. I think in addition to that there's a new generation of commercial products that use machine learning either machine learning built into software packages like email marketing systems and CRMs or standalone machine learning platforms that can be you know, we were talking about modularization before you can just have your site And you can push your data out to this third-party platform that applies to machine learning to it And then it sends you back the results and helps you run your personalization engine So I think there are lots of ways that you can start incorporating machine learning that are you know now five figure instead of six or seven figure Problems and maybe even cheaper in some select cases. I do think focusing on that modularization pieces is often going to be best for Non-profit organizations because nonprofit problems are pretty unique And a lot of nonprofits have their own, you know, particular pipeline their own particular structure of their content And so you need to have control over the data you put in to machine learning process Garbage in garbage out is a famous saying in there. So you need to make sure you're putting in good data If you're going to get out good insights That guy go exactly right And so so yeah, I think that's right and then also for the the investment in machine learning to be worthwhile I think it's important to figure out where your work can have the most impact So I would say your tools and you know the parts of your ecosystem that either have a lot of use Are getting a lot of traffic and so the website's going to be a logical place to start or The part of your ecosystem where it can have the most impact and so I think email marketing Is really powerful because there's a lot of impact that goes through a people's email systems Or on the on the program side, I'm figuring out how you can use machine learning to actually conduct the research actually design your services So I think those would be the places where I would want to start Yeah, you know, I had a slightly different angle on it, which is like You know to even get prepared to do those those things I think one of the things organizations can do and they're lucky in this respect is kind of double down on existing Best practices for things like content strategy, you know one of the most important things for any organization that wants to use ML or AI or anything like that is to have Content or pieces of data that the system can interpret and look at and you know what that really means from a you know Organization standpoint is making sure that you figure out what fields you want attached to your content like oh instead of putting the author in the Body field of a blog post put the author in an author field so that you can easily identify that That's the author of a blog post, you know There's simple things like that that make the machine learning hill a lot shallower and you know The one that we always propose to organizations is to come up with a unified and central taxonomy for the organization, right? So that you know article X can match up with video Y which can you know match up with podcast Z, right? Like it's so common for organizations Especially in different departments to use variants or similar but not quite the same taxonomies for things or to just Have different taxonomies or to have the other problem, which is they don't manage the taxonomy They let lots and lots of duplicates and aliases and synonyms for the same terms up here You know and just solving that problem, which is well within the way with all of every Organization is a huge step towards being machine ready or machine learning ready or AI ready And so I totally agree with everything seven said but I would just start even a level lower Which is like just get your content strategy humming and you'll be way better off than if you didn't do that It's that's a huge point and and the good news is everything you do to prepare yourself machine learning is just good for your organization Regardless, so even if you don't get there You know having that as your goal is gonna is gonna help you put in some some best practices that will improve your content overall So, you know, that's a really good point All right, so we've we're gonna we got a meeting with our top Prospect finally got him into the executive office And they're like hey AI ML, why do I care? What's this supposed to do for me? Well, you know the easiest one is you know Do you want your dollar to buy you 50 cents or a dollar 50 cents worth of goods, you know And I think that's the difference between AI and ML having orgs and not Which is that every dollar spent on that is force multiplied or magnified by the Successful use of those tools and the organizations that use those are going to have a much greater impact for Staff hour spent or per dollar spent than the ones that don't and you know It really is going to be a winners and losers situation where if you are competing at someone in the thought space or marketing space or Services space or even just trying to have it better impact than the negative impact of other organizations You need to have this kind of you know tooling at your disposal to compete and to you know Be competitive and so I'd say it's all about how far do you want your investment dollar to stretch of course I want my investment date, but I was what you're telling me I need to do is a five hundred thousand dollar project today, and then I still not there How many years does this take what kind of sounds like there is a starting point? Maybe it's making sure I got tax on them your questions. I can ask but what does this look like over the course of three years? Yeah, well, that's a good. That's a great question, you know What we always recommend is that you you know don't invest a lot early so that you're you know You don't have anything to invest later once you've really figured out your problem There's a lot of learning in the doing and there's a lot of culture change and change management in your organization that you need to Adopt in order to be successful if these it's not just install a system and go and that's a huge mistake that we see a lot of Organizations do where they get a lot of money together, and they get something But they didn't change the culture or their way of doing business and the tool just sits there And they have you know Salesforce marketing cloud or Marquette or part odd or any of these other tools just sitting around And they're not using it effectively and the reason is somebody convinced them They should get the capability without the culture change So I'd say no don't spend half a million first spend You know 50k or whatever your budget will support getting some of these things working Showing how people need to change their workflows and their processes figure out how the organization really needs to adapt Itself in order to use these tools and keep making incremental investments that keep improving that I mean I could see you're obviously choked up about these organizations that you know don't use those tools the way they should I mean we have passion for all of our client projects, and that's why we talk about this Stefan do you have anything to add there? I mean that's I couldn't have said it better I mean I think that being able to start with something focused in mind This is a problem. I want to solve. This is the capability. I want to add To to my ecosystem and then go with that and get it in place and figure out how you can learn from that Figure out if the approach you use to solve your first problem can be applied to your second third eighth tenth And so on and and I think there's a there's a lot that can be learned each time you go through this process And starting small that's you let's you have a win And then lets you figure out how to get the staff aligned around it help them learn from the process so they can do a Better job the next time I you know like You know you've never built your last website And and you'll never have built your last machine learning algorithm You know there's there's always room for improvement and as situations change you need to be able to change it and update it So I think planning for this as a a new line of capability That you want your staff to be able to understand to own to maintain and to evolve Is is really important. I think you know to Nate's other point You know why do it to scale your impact if you want to have more impact. This is a way to do that I think quite quite simply Most organizations we talk to Are not running perfectly There's there's always something about the way they manage things something they can optimize. There's always an opportunity For organizations to improve the experience of their constituents in one way or another And I think machine learning can be at the heart of a lot of those solutions Made any final thoughts Yeah, I just think one other thing to think about for organizations is that you know The doing changes the organization and that's usually for in a good way, right? And if you successfully adopt machine learning and artificial intelligence techniques You're probably going to want to do different things with your staff and with what you're doing even in your mission Because you'll solve some of the basic problems that are capacity blockers for doing more advanced work And that's another reason you want to roll into this incrementally, which is that you don't really know what you want to be doing Organizationally once until you have reached a new plateau, right? Like I think every organization is reaching new maturity new sophistication and how they're addressing their mission And one of the things AI and ML let you do is jump a maturity level by taking a lot of things You already know how to do well and taking them off the plate if your staff to let them learn how to do things They don't do well yet or don't have the time to do well yet And I think that's another key part of this like organizations that adopt this aren't just doing what other organizations are doing faster You're actually learning new deeper more interesting things to do on top of the things other organizations are doing Like it Stefan was at your closing thought before did you have one more you wanted to I have as many closing thoughts as you need Yeah, I know we probably went really long. This is our first one So anyone who's made it this far if you're listening, you know, maybe we cut this down Maybe it's a little shorter than the hour we just spent but needless to say we could rip for a while You know, there's show notes below Leave us a comment here on our blog or if you're seeing this on LinkedIn leave us a comment there like it share it Telefriend about it set them our way if you got ideas for topics you want to hear folks from our company talk about or You're an outside guest and you want to come in and talk with us. We'd love that too, you know Some of my key takeaways. I think no matter the size of the organization. It all starts with alignment You know, what is the goal, right? Don't I like Nate when you had said, you know, don't let the data tell you what strategy is You got to have strategy and then use the data to help you get to where you need to go and amplify those tactics Think Stephanie Lisa said it in our company But we always like to use it right as that answers get all the credit for questions do all the work So what are those questions we can really be pushing towards here? You know, I think some of the other pieces are it sounds big and scary in the aggregate But this is happening and it's happening now and there are steps you could start taking to get there You know, they're the status quo is not going to jump you to where you need to be You know, we talk about my other takeaways. We talk a lot about Change management and you know, we we as a company backed away from marketing and talking about transformation And I feel like I'm ready to steer back into it really hard just on this, right? And I love what you guys are saying there. It's it's in the doing of this That that change and transformation starts to happen, right? And it's Maybe there's another one of these where we get into just agile But how do you remain and have the agility within your organization to then take a learning and change and do something different And enhance it so it's sound the machine learning sounds like it could Benefit my audience for intentionality quicker serving them to complete a task to get what they want Which hopefully will make them want to come back and engage even more deeply with me because they're serving my needs in an easy way Can help me amplify my staff, you know, we get asked this question a lot when we're on transformation projects of like Well, now who do I hire? What do I need to do? Well, it's That's probably not the right question exactly, right? Like what do you want to be doing? What can the machines do to take care of a lot of that more repetitious task? And let's get you to the next higher level of something that will always remain high touch human focused And let's find the right people to do those jobs I think we've identified at least An insights officer and a chief automation officer has to do potential roles coming out Uh, so this was exciting. Um Thanks for your time today gentlemen Oh my pleasure Yeah, thanks for having us All right. Well signing off for our first official We don't have a name for it yet, but I like the constantly Parsons TKO If y'all got if you got ideas you can leave that for us too. Maybe we'll call it that All right, all right, thanks guys