Adam Goldfarb: Harnessing AI for Economic Opportunity
Adam Goldfarb, Founding Executive Director of Opportunity AI, a collaborative of more than 100 foundations focused on AI's impact on jobs and economic mobility, discusses how philanthropy and the social sector can work together to ensure AI expands rather than narrows economic opportunity. Drawing on his experience leading Schmidt Futures' workforce technology initiative and co-founding Opportunity AI, Adam maps the emerging landscape of economic opportunity funders into four distinct camps: those focused on worker voice and governance, those preparing transition systems for potential disruption, those building AI-powered tools for the social sector, and those watching the growing entry of wealth from the AI industry itself. The conversation explores the most promising categories of AI use cases in workforce development, from career navigation tools and AI-powered coaching to benefits access and operational efficiency for nonprofits, and why philanthropy needs real-time labor market observability to detect how AI is reshaping jobs and skills before it's too late to respond. Adam also makes the case that agility, for workers, training organizations, and funders alike, may be the defining touchstone of this era, and offers a vision for how the sector can move faster and more experimentally without losing sight of the learners it exists to serve. Transcript Julian Alssid: Welcome to the Work Forces podcast. I'm Julian Alssid. Kaitlin LeMoine: And I'm Kaitlin LeMoine, and we speak with innovators who are shaping the future of work and learning. Julian Alssid: Together, we unpack the complex elements of workforce and career preparation, and offer practical solutions that can be scaled and sustained. Kaitlin LeMoine: This podcast is an outgrowth of our Work Forces consulting practice. Through weekly discussions, we seek to share the trends and themes we see in our work and amplify impactful efforts happening in higher education, industry, and workforce development all across the country. We are grateful to Lumina Foundation for its past support during the initial development and launch of this podcast, and invite future sponsors of this effort. Please check out our Work Forces podcast website to learn more. And so, with that, let's dive in. Julian Alssid: As we continue Season 7, there is one topic that has come to dominate virtually every conversation we have in our consulting work, on this podcast, and frankly, everywhere we turn: artificial intelligence. The central question for our field isn't whether AI will change the nature of work and learning; it will. The question is how can we harness these tools in ways that genuinely expand economic opportunity for the people who need it most. Kaitlin LeMoine: And that question puts a real responsibility on the organizations and individuals who are in a position to shape how this moment unfolds. Funders, policymakers, and practitioners all have a role to play, and getting it right requires deliberate strategy and coordination, a willingness to move faster than maybe ever before, and a clear-eyed view of both the challenges and opportunities that come with AI tools as they emerge. Julian Alssid: Which brings us to our guest today. Adam Goldfarb is the founding executive director of Opportunity AI, a collaborative of more than 100 foundations focused on AI's impact on jobs and economic mobility. Before starting Opportunity AI, Adam launched and led Eric and Wendy Schmidt's grant-making program devoted to innovation and career training. Earlier in his career, Adam served as chief of staff of a state education agency, co-founded a charter high school, coordinated policy advocacy for the city of Newark, New Jersey, and led strategy for national nonprofits. Adam, welcome to Work Forces. We are so glad to have you with us. Adam Goldfarb: Thank you. It's amazing to be here. You know, all the longtime listener, first time caller, notes apply here. Thanks. So thank you. Kaitlin LeMoine: We're so glad you can join us today, Adam. We really appreciate it and look forward to diving in with you. As we get started, please tell us a bit more about your background and your current role at Opportunity AI. Adam Goldfarb: So, Julian, as you mentioned, I spent most of my early career working for mayors and governors. I did economic development policy for the city of Newark, New Jersey. I was chief of staff of the Connecticut Education Department, and the origin story here is: while I was in Newark, as you mentioned, I co-founded a high school, served as the board chair, and this was a school you're familiar with. You know, we did everything in our power to prepare a group of low income high school students to get into and to succeed in college. And long story short, we had great college acceptance rates, rates that beat the odds, and which were a wonderful achievement for the students, but relatively poor college completion rates. And I had personally, I think, underestimated the barriers that the traditional postsecondary system throws in front of low-income first-gen students. And so I think, like a lot of us, that inspired me to begin working on shorter-term interventions that can help low-income students get their first upwardly mobile job, achieve economic freedom. And so I took on roles in that space, and that's what led me eventually to Schmidt Futures, which was Eric and Wendy Schmidt's philanthropy. That's when I met you, Julian. I was helping them rethink their shared prosperity work. We had a program using technology to help workforce training entities improve their growth and scalability with your help. So yeah, I got to work with a lot of the most amazing organizations in our space. Organizations that I think your listeners are really familiar with: Social Finance, Empower, Propel America, Craft Education System, Upwardly Global. But what happened was, this takes us into 2023, and AI starts to grab the headlines, right, with ChatGPT coming out. And for me, and for I think a bunch of my colleagues who found themselves in grant maker roles, this was like an "oh crap" moment. I could imagine my boss calling me and saying, "What's your AI strategy? How does this change what we're doing with your program?" And I wasn't really sure. I don't come from a technical background. I wasn't really sure what I would say, and it was leading to a lot of crises of confidence of like, how can we deign to support this if we don't know the answer to that question? And so what happened was, I and the head of grant making at GitLab Foundation, whose name is Matt Ziegler—I think you had Ellie Bertani on a previous episode, right—so Matt and I started trading notes on what the two of us were reading about AI, and we pulled in some of our friends and we said, "Let's go on a regular Zoom." And those Zooms just sort of eventually became the community that became what we call now Opportunity AI, which we spun out with the help of a bunch of foundations in the space, like Gates and Comcast and Segal and Workday Foundation and Annie E. Casey, some others. And so today, those first Zooms have spawned a funder collaborative. We have over 200 members, as you mentioned, drawn from over 100 foundations. And Opportunity AI is trying to solve a problem that is endemic in philanthropy and long term in philanthropy, but that is becoming more acute. We think that the social sector, nonprofits, impact organizations that may be organized as for-profits, have a huge role to play in helping workers and learners adapt and succeed in the AI era. And for that to happen, funders have to step up to the plate. But unfortunately, economic opportunity funders are, on average, like me, not technical. We're not like former software engineers. We know a thing or two, hopefully, about economic opportunity and modalities there. So, how are we going to keep up with AI's developments given those non-technical backgrounds? And then two, philanthropy is just famously fragmented and super slow. You have too many duplicate efforts. We make would-be grantee partners deal with too much rigmarole. We descend into strategy refresh processes that take months and years, and so Opportunity AI is trying to be the collaboration infrastructure to solve those problems. We're helping funders learn and take action together in the economic opportunity world. We do that by learning, right? So our funders come together, they meet with the most cutting-edge practitioners, they hear from the frontier labs themselves, they learn about what's going well and what's not going well in use cases around AI in economic opportunity, and we bring funds together to align their funding and to craft new multi-donor projects together. Julian Alssid: You may not be techie, but you certainly have immersed yourself a bit now Adam Goldfarb: Yes Julian Alssid: in this brave new world. So we'd like to start broad and dig deeper, beginning with hearing from you about just your observations of how the AI funder and practitioner landscape has evolved over the past few years, and what shifts you are seeing. Adam Goldfarb: I think there are a few phenomena. I'm curious what you guys think. So one is just the simplest one: growth, growth of interest. Our community has grown. I think that you may not have thought of yourself as a practitioner or a funder as an AI funder or an AI practitioner, but suddenly you are finding yourself trying to place workers into an AI-inflected job market, and so you have found yourself—you may not have known this at first—but you are that. And so just growth of interest in the space relative to those origins I mentioned in '23. And the second thing, and I think this is a big difference from 2023 and 2024, it can seem from the outside, I feel like you go to a funder's website, a practitioner's website, and it just sounds like everyone's using the same words in a slightly different order about something happening with AI and something happening with economic opportunity. I do think that it's possible to parse the funders into a few discrete camps. They're all friendly, they're all overlapping, but there are groups of funders trying to answer different questions. Okay, and so one camp that we see is what I call the voice camp, and they're basically asking: who gets a say in how AI is built and governed? Okay, and with these folks you'll see programs related to worker power, community agency, and what should the guardrails be on big tech. There are amazing funders in our community that I really admire in this space: Kaphan, Irvine in California, Omidyar; and some of those funders have come together to launch a coalition that your listeners have probably heard of called Humanity AI. And there's a preexisting organization called Families and Workers Fund that is prominent in this realm. So there's that voice camp. And then we're seeing another camp, especially this year, that we think of as the quote-unquote bridge camp, and they're asking: are transition systems ready for AI's potential disruptions? Right. So these are funders who want to make sure that the unemployment insurance system is up to snuff for what could be a different shape of job disruption in this economy. You're seeing funders partner with state governments, with the NGA, to help governors shift their strategies in light of the regional economies that they're in. In particular in this space, former Rhode Island Governor Gina Raimondo has founded something called the Raise Us Coalition, which is a very impressive, sophisticated effort in this space. Gary Community Ventures in Colorado is another philanthropy very adept at partnering with government and helping them make this transition. And then the third camp that I'd say, and this is where a lot of Opportunity AI members in particular resonate, these are the builders, okay? And these folks are asking: what tools can the social sector build? How can the social sector adapt itself, the nonprofit community especially, to the AI era? These funders are longtime funders of direct service organizations, and they're asking how this adaptation can happen. You've seen Next Ladder Ventures, a prominent multi-donor effort launched by Gates, Valhalla, a few others, GitLab Foundation, where Ellie and my co-founder Matt come from. So there's this camp focused on: let's help the social sector itself adapt, let's help the social sector use AI tools to go farther with their programming. And then the last camp that sort of has to be mentioned, but it's too early to speak too much about it, is just the present and coming entry of wealth from the AI industry itself. OpenAI Foundation has recently announced itself; it has a corpus of funds by virtue of the nonprofit/corporate setup of OpenAI that is bigger than many of the biggest other philanthropies in the United States combined, like 130 billion dollars in assets that they're launching with. And so, as they and as other folks from the AI industry enter the space, it's going to shift what's possible quite a bit. So this is just an emerging trend worth watching. It's really interesting to see these camps emerging and coalescing. And what we're hoping is that Opportunity AI can be a place where these funders can come together and make sure that they are acting harmoniously, not duplicating efforts, and trading notes. Kaitlin LeMoine: It's really interesting to hear you break down the different funder archetypes in those groups across those themes. I feel like it must be very interesting from your vantage point to see over a few years these different affinity groups evolve, recognizing that of course there's some overlap, but really interesting to hear you break them out in that way. I'm curious, Adam, what are the current leading categories of AI use cases you're seeing across training, education, and adjacent areas too, like access to benefits? Adam Goldfarb: I think we are seeing a lot of tool building happening, especially from that final category that I mentioned, the builders… Kaitlin LeMoine: The builders, yeah. Adam Goldfarb: Funders who are funding that kind of work, and so I do think there are phenomena emerging here too, right? So one of them is navigation; career navigation is emerging as this meta category, a category of categories of initiatives. As you know, so much of the magic of an amazing career intervention, whether it's at a community college or a workforce training organization, is embodied in the career coach, the person who helps you think about your job search strategy. They're tailoring your resume. Maybe they're also helping you deal with the balance of life and your job search and your training. They help you understand what an informational interview is, all these weird rituals that make up ways to enter the labor market. And so there's a lot of optimism, temperate optimism, that software can help capture some of the magic of a career coach in your pocket. You can imagine a digital coach that knows your regional labor market data really well, knows your pre-existing credentials really well, knows the training options in your area, knows which ones have high ROI and which ones are a little more dubious. And especially exciting is imagining the agentic application tier, where that piece of software in your pocket can take action on your behalf, right, can say, "I know you're a little shy about this career transition you're trying to make, but there's a networking event that has no bus transfers in your area that I can sign you up for right now," and sort of start to get you exposure into this field. So some of these navigation tools are being built to be learner-facing, and some are being built to be career coach-facing to enable a career coach to serve more people and more effectively. And it makes sense why this is an exciting use case, right, because AI is really good at the matching problem, taking a big mess of data and making sense of it. In New York State, there are 900 apprenticeship programs, and there's a matching problem between apprenticeship seeker and apprenticeship provider, and matching problems are something AI is just really well suited for. Another use case within career navigation is AI is really good at giving you fleshed-out examples of something, right? We all have fun with AI saying, "read this to me in this voice" or "translate The Odyssey into modern vernacular." And so, as a supplemental tutor, AI has a lot of value in throwing low-pressure at-bats at a learner with practice problems. Climb Together is a nonprofit whose chatbot helps low-income students at community colleges have the chance to practice informational interviews, which is, as I mentioned before, a very odd ritual: how do you talk about yourself and your skills in this very ritualized way where you don't sound so into yourself, and you're making a connection with the person across the desk? AI can throw at-bats at you and provide you that simulation. So all of those fall in that bucket of career navigation, where there's a lot of promise, though there's a lot of peril—we have to make sure that the data being fed into the tools is good enough, etc. So career navigation is sort of a meta category. And then you mentioned access to benefits, right? Government does not make it easy for us to navigate its systems. There's a thicket of eligibility rules, there are weird forms, there are benefits cliffs where if you sign up for these programs, it makes you ineligible for those. So AI can help people navigate the UI system and other stability systems to provide the job seeker with stability during their job search. Here in New Jersey, just translating the forms into Spanish increased applications by 172 percent—people who were always eligible, but the system just didn't speak their language, and it was labor-intensive for us to translate all of the various interfaces that a person would experience. Another really interesting use case: different items in the grocery store are eligible for food assistance programs or not based on whether they're classified as food or junk food. And it's annoying to have to carry those items to the register and find out whether this sparkling water is classified as soda or sparkling water. And so, another organization in our space, Propel, lets you scan the barcode, and it tells you the eligibility of that product. So just exciting things that AI can make possible. And then the last thing I'll say is social sector organizations who are building are using AI to help themselves reach more students at scale at a reasonable cost. AI can absorb the repeatable, annoying parts of human services and free up the teacher at a job training program from so much paper grading to just being the deliverer of content. If we help every social sector organization realize an operational efficiency enhancement, the dollars that we put towards the social sector can go farther. So broadly, enabling scale is an exciting, non-glamorous way that AI is going to enhance our space. Julian Alssid: Yeah, it's really incredible seeing the emerging applications in our field, and at the same time, there is a lot of worry, a lot of concern, particularly that AI is going to reduce job opportunities or certainly change work as we know it as more tasks become automated. How is philanthropy working to address this issue? Adam Goldfarb: I think that philanthropic activity in this space is going to be very significant. This is an area that's just not in the headlines yet that people don't know about, because this is a big problem. There are these doomsday predictions that make the headlines sometimes from people who know the technology best, that a white-collar jobs apocalypse is coming. There's disagreement between sort of the people at the frontier labs making those predictions and labor economists, and we want the people in our sector—the head of curriculum at a community college, a workforce board director—to be able to know how demand is shifting in their regions, how AI may be augmenting or automating various skills, and how that's changing the demand for various occupations, and how a given job description may look different than it did six months ago in light of AI's effect on the technology relevant in that space. Government data products are not equal to this task. There's too much of a lag, and they're not adept at spotting disparate effects across companies or across regional labor markets. A company in one city may automate a certain function, and the identical company across town may not automate it. It's uneven, and we can't see those phenomena in the national statistics. And so I am hearing from funders—we had 80 funders together in Washington a couple of weeks ago—that this conversation is really moving forward. Philanthropy and the social sector would like to partner together to create an observability function, where the practitioners that we care about the most have greater observability into AI's effects on jobs and skills as they unfold in a given labor market: by data infrastructure work to make data more available to tools, by tool building itself to make sense of that data, and by funding research. This will enable us to see the unpredictable ways in which AI is affecting jobs and skills, and enable us to adapt accordingly. We had Molly Kinder speak to us in DC, who is recently of the Brookings Institution, and she pointed out that there were these predictions of labor market disruption particular to a certain type of coding job, but meanwhile, it's clerical and administrative work that is seeing the biggest exposure to AI automation, and a lot of those roles are traditionally female-coded in the labor market. We're just not ready for it; we don't have the detection tools to assess whether those things are happening and how government policy and nonprofit response has to react accordingly. And so there's going to be a lot of energy entering that problem space. Kaitlin LeMoine: Well, and even I would imagine the lag of the information, as you said—thinking back over the last three years, if your data collection efforts are from last year, we're even more behind than ever before with respect to how seemingly within weeks things can shift around how jobs or roles can be impacted. I would imagine that's a really challenging piece of this as well. Adam Goldfarb: Yeah, and like a lot of your listeners probably subscribe to Lightcast, there are tools available, but we just need more data brought to bear here, and we need more tools to make sense of the data. The AI labs themselves, for example, have proprietary data on how people are actually using AI tools in given occupations, and so we need the right sort of data sharing agreements to extract some of that to help us use the predictive power of that. Julian Alssid: Well, as with earlier technological innovation, there'll be jobs that we don't know exist, and so while clearly a lot of the simpler tasks are automated, and that clearly impacts less skilled and younger workers, for example, we really don't entirely know where this goes. Adam Goldfarb: Exactly. One probable and exciting shift here amid all this concern about the negative effects is that it does seem very likely that AI could lead to a surge in solopreneurship and folks hanging out their own shingle trying something, that the barriers to entering the labor market and entering entrepreneurship are lower. Workday Foundation has a big partnership in this space around encouraging AI-powered solopreneurship. I feel like more workers will find themselves in those kinds of roles than before. Kaitlin LeMoine: Building off of we don't know where we're headed exactly, but despite that, Adam, what advice would you offer to education and workforce leaders who want to move forward thoughtfully in this unpredictable and ever-shifting environment? Adam Goldfarb: So I do think that there is a potential shift here that may help actors in our sector get through this. Over the last decade, we've seen the emergence of these touchstone ideas in our sector. There was a skills conversation—every learner needs hard-edged skills, reactive to labor market demand, industry-recognized certifications. That was a moment that very much dominated the conversation. Then there was a broadening of that conversation to say the hard skills conversation needs to be rounded out by durable skills, or skills that aren't technical in nature, and that we need to imbue our learners with social capital, like our ability to find a job. So our sector has responded to frailties in the way it works by embracing these sort of touchstones. And I think—this is an idea that I've been working on with a colleague at an organization called CareerVillage, Jared Chung—it's very possible that an emerging touchstone for this coming era is captured in the word agility. I think up and down the spectrum, from job seeker and worker to workforce training organization that serves those workers to the funders fueling those job training organizations, we're going to embrace the term agility in new ways. So, what does that mean? For workers and job seekers, AI might cause all of our roles to shift more often. We may keep the same job title, but what that job title means changes as AI augments certain subtasks and automates others. People might change careers more often. We're all living longer, so we may be changing careers more often as well. And for many of us, when that happens, that does violence to our sense of our own self-identity, and it's stressful to switch careers and to have your role change. So we may need fresh thinking about how we imbue workers and job seekers with an agile mindset where we're picking up skills on the side, we're happy to experiment, we roll with the punches when some of these transitions happen—not to minimize how actually life-altering and tragic some of these transitions can be, we know that from past waves of innovation, but that this may be a helpful mindset to imbue. And then for the workforce training organizations, they need to be more agile as well. They need to stand up new programs in response to probably faster-shifting labor market demand, and they need to jettison programs with no sentimentality that are no longer meeting the moment. They need to use AI to make their internal operations more efficient so that they can act with more agility in light of greater uncertainty. We want all of our cherished community colleges and workforce training organizations that produce these odds-beating results for their learners to thrive through this era, and that might be part of what it takes. And I think for my sector, funders need to become more agile as well. We can't get lost in these years-long strategy refreshes. We can't be paralyzed by the uncertainty that it's not clear which skills are going to be relevant when. I think funders can step into this space with more shorter-term, more experimental grant making, where each shorter-term grant making sort of produces evidence that informs the next round. We can shift capital toward models with the best evidence in that way. We need to fund differently as well. As more organizations become builders, that changes the way the cadence of funding has to work. There's this old-fashioned way of "we'll give you a million dollars in four equal tranches, and you only get the second tranche after you meet a bunch of criteria for how the first tranche went." But an organization building technology has to front-load their acquisition of technical talent. So the cadence of funding needs to be responsive to the kinds of tasks that social sector organizations are taking on right now. Fast Forward is an organization in the Bay Area that has amazing playbooks about this, both for grantees and grant makers. So there are a bunch of shifts that I feel like our sector could take to better meet the moment, and it's possible that agility could be a term that unifies the shifts that are needed. Julian Alssid: Well, Adam, as we wind down our conversation, it's clear that you have your finger on the pulse here, so we're going to be counting on you to be both agile and keep us all abreast of all the shifts and iterations taking place in the world. No pressure. Adam Goldfarb: I'll do my best. Julian Alssid: How can our listeners learn more about your work and follow what Opportunity AI is doing? Adam Goldfarb: We don't have a massive or sophisticated online presence. We give folks a sense of what we're up to over on LinkedIn. Always happy to compare notes with folks. If you are an economic opportunity funder interested in joining other funders on their AI learning journey, I'm extremely excited to connect—please reach out if that describes you. And I think in the coming few months and year, we're going to be supporting a few bigger collaborative projects, things that individual funders can't achieve on their own, but that we can achieve by collective action, so folks can stay tuned for updates on those projects. Kaitlin LeMoine: Thank you so much for taking the time to speak with us and to share what you're seeing emerge across this complex and exciting moment in time. And yeah, we do look forward to continuing to follow your efforts in this next chapter. Adam Goldfarb: Well, thank you for having me. I'm excited to keep the conversation going. Julian Alssid: Yeah, really appreciate it. Thank you so much. Kaitlin LeMoine: We hope you enjoyed today's conversation, and appreciate you tuning in to Work Forces. Thank you to our listeners and guests for their ongoing support, and a special thanks to our producer Dustin Ramsdell. If you're interested in sponsoring the podcast or want to check out more episodes, please visit workforces.info/podcast. You can also find Work Forces wherever you regularly listen to your favorite podcasts. If you enjoyed this episode, please subscribe, like, and share it with your colleagues and friends. And if you're interested in learning more about Work Forces consulting, please visit workforces.info/consulting for more details about our multi-service practice.
