Data + AI for Good: How Nonprofits Are Transforming Mission Impact with Databricks

Summary

This video introduces Databricks for Good, a three-pillar program offering technical services, compute discounts, and marketing support to help nonprofits unify fragmented data and build AI-ready foundations without large budgets.
Case studies from a care network serving 150,000 people and Crisis Text Line—with 1.5 million HIPAA-governed conversations—show how nonprofits are building unified client 360 views, enabling natural-language queries via Genie, and automating compliance monitoring.
The video presents a practical playbook for nonprofit partners: achieving a self-contained data and AI implementation within six months on the Databricks Data and AI platform, driving measurable outcomes like faster volunteer routing and population-level insights.

Data + AI for Good: How Nonprofits Are Transforming Mission Impact with Databricks

Nonprofits are tackling some of society’s most urgent challenges, but many are constrained by limited budgets, fragmented data, and limited technical capacity. In this video, we will share how Databricks for Good helps mission-driven organizations overcome these barriers through nonprofit pricing, technical enablement, and hands-on support from Databricks experts. Attendees will learn how nonprofits can use the Databricks Data and AI platform to unify data, strengthen analytics foundations, explore responsible AI, and turn information into insights that support better decisions and greater mission impact. The session will feature stories from Crisis Text Line and GO Project, highlighting their missions, data journeys, implementation lessons, and the opportunities they see to use data and AI more effectively. Attendees will leave with a practical playbook for how nonprofits can get started with Databricks and accelerate meaningful, measurable social impact.
Talk By: Joydeep Kamdar, Director, Field Engineering, Databricks ; Michael Berk, Staff AI FDE, Databricks ;

Chapters

FAQs

What is Databricks for Good and who is it for?

Databricks for Good is a program that provides nonprofit and mission-driven organizations with technical services, compute discounts and credits, and co-marketing support to help them build on the Databricks Data and AI platform. This video explains that the program prioritizes organizations with a clear, measurable mission and a realistic path to delivering impact within six months.

How is Crisis Text Line using Databricks for HIPAA compliance?

Crisis Text Line uses Databricks to manage and govern a dataset of 1.5 million conversations under HIPAA, implementing automated de-identification, compliance checks, and GDPR monitoring. This video shows how automated compliance monitoring on Databricks removes the manual overhead that previously constrained the team's ability to use sensitive data for improving volunteer routing.

How are nonprofits using Genie for natural-language data access?

This video shows how non-technical care staff at a nonprofit serving 150,000 people use Genie to query a unified client 360 data foundation in natural language, without requiring SQL knowledge or data engineering support. This capability is highlighted as a key enabler of self-service analytics in resource-constrained nonprofit environments.

What results have nonprofits achieved with the Databricks Data and AI platform?

This video reports outcomes including faster volunteer routing and the ability to generate population-level insights that were previously impossible due to fragmented data across states, systems, and spreadsheets. Crisis Text Line gained automated compliance monitoring that freed analysts to focus on impact measurement.

Full transcript

[00:09] How's everybody doing? >> All right. Yeah. Amazing. Um, so thank you, Joy. Uh, this is a pretty emotional moment for me personally. I've been working on data bricks for good for about three years. We got a logo, we got a booth, we got a keynote video as well
[00:22] with Virtue Foundation. Um, and if you want to learn more about how you can get involved, please reach out to forgood databicks.com. We'd love to chat, especially if you're a nonprofit.
[00:34] So, um, Joy alluded to the three pillars. A lot of stuff comes in threes in this talk. Um, but tangibly, the the deliverables that data bricks for good provides is three-fold. The first is
[00:46] services. So, that can be advisory or that can be technical hands-on keyboard. But in short, we'll bring in experts to help you for free build with data bricks. Second, um this is something
[00:59] that Joy has almost single-handedly driven, which is compute discounts and credits. So, as we noted earlier, nonprofits are very missionrich, but potentially bit resource constrainted. So, we look to augment that with some
[01:13] free DBUS or dollar DBUS. And then finally, uh marketing and PR is a necessary function of this. We do need to get executive buying within our organization and so collaboratively marketing these stories is an absolutely
[01:26] essential part of data bricks for good. All right. So, uh what makes a strong fit? Um we look to partner with a variety of different missions, different organizations or structures, sizes, you
[01:40] name it. Um but first and foremost, we work at data bicks. Got to use the data bicks platform at least for the majority of the implementation. Um, we also are in the business of doing good. So, it's important that it's a mission-driven
[01:52] organization. We've typically partnered with nonprofits, but that's not a hard requirement. We're willing to chat about that as well. And then, um, specificity and time to achievement is also very important. We want to make sure that we
[02:04] can achieve something within six months. We are willing to make those bigger bets. And for Virtue Foundation, it's been a over two-year bet that has paid off uh, dramatically. But, we do want to make sure that they're self-contained
[02:16] statements of work. Um, and then finally, as Joy alluded to earlier, we are in the business of doing good and driving impact. We want a long-term vision where when we build something, it
[02:28] will drive value for years to come. So, we want to make sure that all those capabilities and staffing are in place. If you generally meet these requirements, we can probably partner with you in some capacity.
[02:41] All right, so call to action. One more time, there's the email. And uh we also don't want to just like shout out all these general ideas and we're going to hear from two great
[02:53] nonprofits. But here are some common patterns that you can take away from this talk today. So the first is data bricks as I'm sure you've seen throughout this entire talk and throughout this entire conference uh
[03:07] is a platform. It's really important that we unify all of your data assets under one roof. And for nonprofits, that can take a variety of forms, but you guys probably have some sort of CRM system. Um, there's probably a bunch of
[03:20] semistructured or unstructured text in the form of surveys and intake forms. Um, and a variety of other things like financial data, which might be tabular. So, first you're going to be want you're going to want to unify it within the lakehouse platform.
[03:34] Um next once it's all in the unified lakehouse platform and governed with unity catalog securely um it's really important that we start getting value from it. So remember that resource
[03:46] conrain statement well maybe you don't have a bunch of coders and as you have seen throughout this conference you can chat with your data via G. So natural language interaction is really really important. Um, this helps you be agile
[03:58] and move quickly. And I'm not going to call out everything on this slide, but these are common use cases that nonprofits leverage within the lakehouse framework. All right, so baseline the data is in
[04:12] the lakehouse. On top of it, we have this governance layer and now we're starting to leverage these agentic and AI based capabilities as well as the traditional ML notebook driven stuff. uh SQL warehouses so you can write queries,
[04:24] dashboards. If you want to build a chart, you can do that as well. Um and from there it starts to actually drive value. So a couple things that I like to call out here. One thing that's near and dear to my heart, especially in the age
[04:36] of AI, is operational triage and routing. Super unsexy, but really, really easy to automate with LLMs. Also, in the top left, if you're writing
[04:48] emails to donors, maybe you have an LLM help you with that. or in the bottom right. If anybody's done grant writing, it's pretty time intensive. LLMs are a great way to augment this process.
[05:00] So, all of this unifies around the mission of staying lean. We know that nonprofits tackle the hardest problems often with the least amount of resources and AI and the data bricks platform helps augment this gap.
[05:13] So, I hand it over to Corey shortly. Um but would love to chat after if you want to discuss any data bricks for good opportunities and thank you for attending.
[05:33] Okay. Well, this is great to be here. Um as Michael said, my name is Corey VTO. I'm the chief data and information officer at an organization called the Global Orphan Project. Uh just really brief introduction. Professionally, I've
[05:46] spent about 25 years doing two things. One, helping build a predictive analytics firm. Uh luckily, that firm had the opportunity to join Mastercard where I spent about 10 years as a senior vice president of engineering in their
[05:58] data and services division. But personally, my wife and I, we've you our oldest is adopted. We've been foster parents. And so I always knew at some point I wanted to make that jump to some
[06:11] sort of nonprofit. And so about 10 months ago I did that and went to Go Project, Global Orphan Project. And so what is Go Project? Go Project seeks to break the orphan cycle
[06:24] by utilizing preventive platforming to stabilize families before breakdown occurs. And so what does that mean? How do we do that? We have these three different platforms. Care Portal, Strong Family, and Goax Apparel.
[06:38] Just to explain those briefly, each one of them, Care Portal is a care network. And so what that means is it's kind of like a marketplace. On one side, you have social workers. Social workers who
[06:50] work in the child welfare space, social workers who might work at a school, anywhere a professional social worker would be employed. On the other side of the marketplace, who have local faith communities, small businesses, you know, people that want to help. And so that
[07:03] social worker can put that there's a family in need and it might be something as trivial as trivial as they need a they need a bed. They need uh help pay a bill and they put
[07:17] those requests on our platform. This was sort of the active request at the time the screenshot was taken and that fam then those people in the local community have the chance to help. And it's not just like mail a car seat. It's show up,
[07:29] build a relationship, form a connection. In terms of data, you it's not a spectacular big data set. You know, we we serve about uh 60,000 requests a year. We're up and running though in 40
[07:42] states, you know, almost every county in some of those states. Um serving about 150,000 people each year through that. So, that's one of our platforms. Second is strong family which operates in some
[07:55] of the you most vulnerable parts of the world in rural communities where we work with about 70,000 people and strong family platform in addition to program that is management
[08:09] pulling in data for each of the people that we're serving. And then lastly is GOEX which is a fair trade apparel company and we run a factory in Haiti
[08:21] and provide jobs and dignified wages and dignified income for uh those people in that community. And so all these have a lot of data involved and when I joined
[08:35] 10 months ago a lot of what Michael just said was true. It was definitely a sort of this isn't Mastercard anymore. This is sort of a different world and many of you said you work at a nonprofit can probably relate
[08:47] to a lot of this. We had fragmented data across multiple systems, way too many spreadsheets. We had manual reporting draining our data analyst staff of one
[08:59] who spent a lot of time writing SQL queries so that people could put that data in spreadsheets. We get lack of shared impact definitions. Um places where data that needs to talk didn't talk. So, a person on our advancement
[09:13] team going to speak to a donor and not knowing that donor volunteered last week. You know, these are things that the data just needs to be able to do. And we also had a lot of aspirations and visions for AI and what's next. And the
[09:27] reality is we weren't ready. You know, the reality is we need to build the foundation first. So, what I'm about to go through is not going to wow you with the agentic workforce we've built. Yeah.
[09:39] That's that's not where we are and that's probably not where you know a lot of you are. This is how did data bricks help us with the basic block and tackling we needed to do to go from not really ready to ready and sort of at
[09:53] that bottom of that hockey stick. Yeah. And ready to take off with a lot of the uh things that are part of that last row. So how did we do that? What did we do? What did we build? Um, we're really
[10:07] focused on the left and the middle all about just building a data foundation and bricks is spectacular because you can do all this in one place. One of the things I was very worried about before I
[10:20] took this role is gosh with such a short small team how am I going to piece together a bunch of different tools and I didn't have to do that u which was wonderful and so we really focused on what are the different data sources
[10:34] building pipelines that we you know were very excited about automating getting those into one place and in terms of capacity and skills we don't have a lot of Python expertise on on our team um
[10:47] But we had some SQL experience. So we didn't use Python. We use SQL. And if you don't have SQL, you know, the new lakeflow designer tool seems quite helpful for you teams that don't have all the you sort of necessary skills
[11:00] because it helps you do that without some of that underlying technology skills. And then we had to deal with people that were the prior owners of these data sets sort of wondering out, okay, now it's all in one place, who has access to what? and Unity catalog sort
[11:14] of just, you know, quieted those conversations very quickly. Uh, and then working through sort of our medallion architecture, but really focused on getting a really good set of gold tables
[11:27] that we were very happy with. Um, which really meant for us was the smallest number of tables that we can use that are going to feed all the data products that we could envision. Um, and so that
[11:39] was a really important part of just that and tackling of getting from not quite prepared to prepared. And then what did we start to tackle and in what order?
[11:52] And it's funny because the first two things are in the wrong order, but you know, we started with departmental dashboards. So just getting one department, we picked our fundraising team, building some dashboards, getting them excited, roll out to another one,
[12:05] another one, and then you start just sort of seeing this excitement build of people knowing that, okay, wow, this is different. That's not a spreadsheet. This is this is um, you know, looks professional. And then we turned on
[12:17] Genie and people were very excited by that. And I'm going to tell you a little secret. We didn't spend a lot of time training the Genie space. We clicked the button on the dashboard that said enable genie. Um, it was good enough, right?
[12:31] You know, it could be better. I'm sure there's a reason they tell us to train the genie spaces, but for our team and our size, that was super helpful. Um, and that kind of got us, you know, from
[12:43] A to B on natural language query and helping people be able to do that. Then we turn on GD1, give people access through that sort of business user front
[12:55] door, give them the consumer access u privilege and not intimidate them with that left navbar of data bricks. Yeah, go into G. It's all you see access to
[13:08] and yeah, very straightforward. Everybody gets the same link. That was for internal users. And then for external users uh embedded dashboards into the system they were
[13:20] already in which for us is our care portal or strong family um platform and service tools. So once we had that, we had a pretty good foundation and then we could start
[13:34] with the more interesting things uh such as donor personalization, being able to build a one-pager so that that fundraiser could go to that potential donor with really targeted information
[13:46] about, you know, his or her community. do notebook analytics. We could use the AI extract and you know other SQL uh AI tools and functions to go through a long
[13:59] list of free form comments that we've had for years and be able to sort of pull to the top. Here's three or four that look really interesting. Why don't you turn those into a blog post or social media video? You know what are
[14:12] interesting stories and use the AI tools right there in SQL to be able to do that. So these are again not dramatic things but it's all right there in data bricks all very helpful all in one
[14:24] place. So one quick vignette and then I'll be done. Um one of the things I think is really been powerful is this moving away from the data team owning
[14:39] you analysis to democratizing it. And so this sort of you spot a trend to be able to ask like I wonder kind of questions to then predict. And so this is a simple
[14:52] chart. Yeah. And in terms it's you know uh consumer you know customer acquisition and churn. You know for us it's about volunteer organizations which you know for us happen to be churches in most cases but it's a volunteer
[15:04] organization. And the blue excuse me the green are overtime volunteer organizations that are still engaged. The red is ones that have turned dormant. That's a problem, right? Just
[15:16] like it is in the private sector. And so you can then ask that I wonder question. And historically that I wonder question would have gone to our data team, figure it out, do things in Excel and figure it out. But now the president of care
[15:28] portal hops on Genie right there in that dashboard asks that question gets some thoughts gets some ideas and takes that also in Genie maybe in a notebook maybe
[15:41] not and says okay that's interesting now that we know some characteristics of why churches or volunteer organizations go dormant well we just signed up a bunch of organizations in three years if we
[15:53] stick on that same pattern what's it going to look like and he just sets some goals for our chief operating officer to beat and didn't talk to our data analyst once and that's a good thing and so that's been really helpful. So
[16:08] last slide, you know, for those who are thinking about uh is data bricks right for you? Is the data bricks for good program? Some things I would recommend you focus on or or do as you're working through this is one, we spend a lot of
[16:22] time sort of nailing our gold schema, getting that foundation of tables ready for that all those data products. Super helpful. to lean into the data bricks team not as doers necessarily but as
[16:34] coaches you needed to learn um you know we're a team of three u myself a data analyst a data engineer needed to learn how to do these things so lean into them as coaches lastly let the genie be the
[16:48] doer and a 247 coach uh and that really accelerated our ability to learn and to move quickly so thank you and then bring up >>
[17:19] All right. My name is Rupish. I work for Crisis Text Line. I'm gonna explain what that is in a few minutes, but I think it was actually really smart of Joy and Michael to choose uh Corey and uh and the Go
[17:32] project with Crisis Text Line. There's some there's some fees and I think there's some also really really interesting differences, too. Um uh so at Crisis Text Line, I have a fun dual role. I'm the chief operating officer
[17:44] and I help help to run and manage our crisis service and I'm also the chief product officer. Um, and I I I mentioned that just because it's like really interesting how you have to think about the product decisions that we're making
[17:56] and kind of connect it to the implications on the service. Uh, which you you'll see kind of throughout this presentation. But, um, first a little bit about Crisis Text Line. How many of you are familiar with Crisis Text Line?
[18:08] Raise your hand. Oh, okay. Pretty good. Pretty good. All right. We um started in 2013. Uh we're a texton free 247 anonymous crisis service. We're staffed
[18:23] um the first responder is a trained volunteer. So anyone in this room after going through a 15 to 20 hour training on how to build good rapport, how to listen, um how to show empathy, uh can
[18:37] become a volunteer crisis counselor. Se over 700 amazing human beings show up on the platform between midnight and midnight. So every minute of the day to
[18:49] volunteer and help people help people in people in their in their moment of of crisis which which could be they've have uh you know intense thoughts of suicide
[19:01] or it could be that they had a really bad day. And what motivates me and I think what motivated Joy and Michael and data bricks for good team is um we get about 10 of those quotes every day.
[19:17] Like just like I'm going to pause for 10 seconds because y'all have to read that.
[19:32] A person wrote that to us. Think about this. A person wrote that to us after having on average a 40inut 20 message text conversation with a complete stranger.
[19:47] So Joy, this is what you guys are enabling. Um we believe strongly that we have a solution or at least a flavor of a solution to a global problem. So we operate in the United States and
[20:00] we supported last year 1.5 million conversations in the United States. We have we have some small affiliates in Europe. But a lot of what I'm going to talk about and a lot of what uh data
[20:13] bricks for good team is helping us is to set up for global expansion. Um before I get before before we focus on this slide let me let me sort of say
[20:25] one more thing. Um when we have been thinking about global expansion, crisis text line um probably has the most sophisticated, most scaled crisis
[20:37] conversation platform in the world. Again, it's been built and refined for 13 years. It's handled 13 million conversations. There's been 300 million messages that have passed back and forth. And so we we kind of have some
[20:50] experience doing that. When we thought about going global, we thought that was what we were going to do. We were going to find organizations around the world doing crisis support and be able to make
[21:02] them better with our crisis conversation platform. That hypothesis didn't necessarily turn out to be true in terms of our ability to scale and help others in the best way. What we do, what our
[21:15] new hypothesis is, what we believe is the is the is the way that crisis text line can affect mental health globally is through our data and understanding
[21:27] the mental health of a population. And so we had to get our data platform in absolutely perfect shape uh in in all the ways I'm going to describe uh in a
[21:40] moment. So that was the key hypothesis which then led three years ago to starting a relationship with data bricks and taking taking the work that the team had done which was good but not nearly at the level it needed to be in order to
[21:52] to make sure that the data all that conversation data all the data with people's most vulnerable moments all their suicidal thoughts their immigration status any number of things that people share in these conversations
[22:04] had to be protected deidentified stored secured uh in in all the ways that that we need to do when we when we do this around the world. So um we started the around the world journey uh in in a big
[22:18] way last year by uh launching a story. Akiastoy means I'm here uh in Spanish. Um it's a what's it's a chat line um
[22:34] that's based on WhatsApp, right? Which of course almost everyone in Latin America has WhatsApp. So, they're just a simple text away. There's 625 million people that speak Spanish. Uh, Kia Story serves people in 24 different countries
[22:47] from Spain all the way through the Caribbean, Central America, all of Latin America. We're working on Brazil, Brazil, Portuguese, and we're not quite there yet. Um, but in order to do this and to do this at the scale that we want
[22:59] to, like we're doing in the United States, we had a lot of a lot of work to do on the data. And so we had three things that we were trying to solve and they were kind of at odds with each other. Um, you know, we
[23:11] had to be, you know, really take into account the privacy of the data. I mentioned how sensitive the data is. Um, you know, I think those of you that have been playing with AI and AI models, you probably have had moments where the cost
[23:23] has gone up and over the budget that you were anticipating. At the scale of data we have and with the amount of of AI engineers and models we're developing, that was definitely a risk. And so we had to make sure that we had really
[23:36] clear governance and cost controls. Um and then of course we wanted to scale which meant we had different compliance requirements in different parts of the world. So um and we had to do all of this. I mean uh Joy and Michael set this
[23:48] up really well by saying that they totally understand the nonprofit, right? We have big ambitions, big miss big missions, but we have nonprofit resources. Um we we have we have a team. We our our company is 200 people. Um we
[24:01] we manage a data engineering team of five or six people. Um and so uh we you know we're we're in a little bit better shape than than than uh than Corey, but we but but given this amount of work, we still needed a lot of help from the data
[24:13] bricks for good team. And you know, I think um you know, Corey used the word sort of blocking and tackling. I mean, I think this is maybe this is kind of a version of blocking and tackling. Um, you know, we I I was I
[24:27] wanted to say that a data bricks for good didn't come in with like a big flashy AI thing. It was like look here here are the four fundamental found foundational things that you guys need. Uh and and basically they helped us
[24:39] deliver all this. So uh their their team uh coached our team, their team helped us with pricing. Their team helped with implementation and uh and maybe unlike Corey's thing, we learned but they also
[24:52] did work for us. So we were the beneficiaries of of Joy finding resources to work for us and to do this works. In the last year, all four of these things have been delivered. We now we now have uh all the data all the data
[25:05] and the costs are tagged so we know where the money is going and who's spending on what. Um we've done a lot of the work on the catalog. Um we are much stronger in terms of our security and and governance posture. um and um we
[25:18] have automated compliance checks which helps us with HIPPA, GDPR and all the other compliance uh frameworks that we have to follow. So this was this gave us that strong foundation that um that will
[25:30] allow that allows us now to start to scale a story and um you know uh Akiastoy is Latin America and Spanish also can also support conversations in Spain and so um because of GDP we have
[25:46] uh put put the next data center in the EU and so we're going to Frankfurt and so that will help crisis text line you know really prepare for that next step of of offering our service through partners in other parts of the world by
[25:59] being one of the few nonprofits that has compliant scalable mental health platform and this is sort of just the beginning for us. We imagine um a lot doing a lot more together. Um if you think about um mental health is one of
[26:13] the most underfunded, underresourced but yet acute problems in the world. Um and uh that you know when you have a a service like ours the conversations that
[26:25] you have and the pe things that people share give you know governments deep insight into at a population level not individual level population level what's happening and where do they need to put resources change regulation adjust their
[26:39] health care policies to support the people and we feel like we have a we have a role there and of course you know as we do that we need lots of help Um, you know, uh, we have a bunch of fun stuff coming up next, right? Um, uh,
[26:52] utilizing, um, I've been talking a lot about data and data data protection at the higher level. Um, we also want to use, um, uh, we want to we want to use AI and ML to use the data and the
[27:05] conversations to make our volunteers and our clinical staff more effective.
[27:18] An example of that is a conversation may be going on for you know uh 45 minutes or an hour. A volunteer may have to leave. A volunteers's shift may end partially through a conversation. So our
[27:30] system brings another volunteer ready to take on that conversation. Um a five minute if it takes five minutes to provide a summary of the first part of the conversation that's not really helpful. Right? No one's going to wait
[27:44] that long. So if we can deliver it in under a minute with with better support from, you know, from from rag models and things like that, then we can deliver even better service to the texter and make the volunteers experience better by by making them giving them uh um giving
[27:59] them tools to make them much more efficient. So those are some of the other areas that we're that we're working on and exploring. And I mean this is sort of my concluding slide. you know, we we we we are set up
[28:13] now with the partnership from data bricks for good to have to offer this service to 625 million more people. Um our goal in five years actually is uh 25% of the world's population has access to a service like ours. Um Akias Stoy
[28:27] gives us a big step forward. Um but you know because of the work we've done with them we are uh we are HIPPA compliant uh at over 90% level. We're going to be getting conversation summaries as one
[28:39] small example of platform efficiency in less than a minute. Um and you know we're going to support at least 40 countries with one platform. So um I guess my concluding thought is um you know for for those for for the
[28:51] nonprofits in the room I think you know just based upon what Corey and I shared it sort of doesn't matter where you are on the journey. I think, you know, bring your big vision and bring your current
[29:03] state to the team and they'll work at the at the right level, at the right cadence to get you to where you want to be. Thank you so much.
[29:26] Well, >> Corey Rupesh, thank you so much for sharing your stories. I can't believe that Corey got into the organization in 10 months with him and a data engineer and a data scientist. This says possible
[29:41] like that is pretty pretty awesome. uh it gives
[29:54] rightly shared Rupesh's team is larger Corey's team is small they're in similar areas but data bricks can do this for your organizations we focus very heavily in our early days uh in the enterprise
[30:08] segment but we're super motivated to work with the nonprofit sector the fact that you're here is tells us that you believe in data and AI and the potential of it. So, we're looking forward to working with you. Um there is a lot to
[30:22] be done and please don't hesitate to reach out to us at forgood databicks.com and uh yeah enjoy the rest of data and AI summit and thank you again for being
[30:34] here. We'll be around for questions and answers. Uh we love to talk. Cheers.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.