In February 2025, over 300 participants gathered in Madrid not to pitch startups or chase investment, but to reimagine what AI could become when rooted in something far less common in innovation narratives: vulnerability. 

The OdiseIA4Good Hackathon, supported by Google.org and the Pablo VI Foundation within the cAIre research project, differed from traditional accelerators. Supported by CEU San Pablo University, ONCE Foundation, Innova-tsn, and GDG Madrid, the event emphasized care, dignity, and tailored solutions over disruption. Over three days, 42 teams of technologists, ethicists, students, and activists co-created AI tools to address social issues like migration, disability, healthcare, and digital exclusion. 

But the actual test of innovation comes after the spotlight fades. Months later, what emerges isn’t just market-ready apps but more meaningful, living processes shaped by collaboration, precarity, and real-world relevance.

Not all of the 42 original projects continued. Some remained at the prototype stage, while others, as often happens in open calls, did not respond to follow-up. But seven living and diverse trajectories offer something more valuable than polished products: persistent, evolving processes. These teams, each in their way, are still exploring how to turn early insights into long-term solutions. 

Among the standout initiatives, Migration and Asylum Assistant offers guidance to migrants and asylum seekers navigating administrative procedures. DiloClaro is an AI tool that enables vulnerable groups to interact with public administrations using plain language and easy-to-read formats. ElderCare Connect combines an AI-powered app with a robotic interface to support elders in combating loneliness. Opills is an AI-driven application designed to help manage medication in care-dependent settings. Tramits-Easy simplifies access to complex administrative processes through automation. WeRepresent explores the use of AI agents to generate meaningful contacts for emerging artists and athletes. Finally, Woowii promotes employment based on affinity rather than conventional, bias-prone recruitment models.

These teams are not just building AI for vulnerable communities; they are doing so from within conditions of vulnerability. Many lack a legal entity, funding, or even consistent team members. Their shared challenges, lack of validation environments, misaligned mentoring, and absence from standard funding schemes highlight a blind spot in current innovation policy: the gap between where promising ideas emerge and where institutional support is available.
 

These seven projects, designed for and stemming from vulnerability, align with our proposed framework in "AI for Good: Questioning the idea of human vulnerability". We advocate shifting from viewing vulnerability as an external trait of “the other' to understanding it as a structural, relational, and shared condition that affects both designers and users. The focus should be on designing technologies that reduce or avoid increasing vulnerability, rather than on who is vulnerable.

From this perspective, what unites these projects is their challenge to dominant assumptions, rather than their development stage. For them, innovation becomes less about technological sophistication and more about responsiveness: Does the tool reduce cognitive burdens? Can it lower barriers for those outside formal institutions, be it independent artists, migrants, or elderly users? These are not auxiliary questions; they are core design principles in what we call “AI4Good”.

This shift in mindset has implications far beyond hackathons. A recent report, "Funding Ideas, Not Companies", published by the Institute for European Policymaking and the ifo Institute, echoes many of the structural insights we’ve drawn from the follow-up to these projects in the Hackathon’s aftermath. Its critique is sharp: the existing system places disproportionate emphasis on funding companies that conform to conventional business models. As a result, it tends to exclude or actively discourage innovation emerging from informal contexts, social collectives, cultural networks, or grassroots initiatives. 

Instead, the report proposes a shift toward funding ideas, even before they are formally established, and measuring impact beyond market returns. Five core recommendations from the report mirror our field experience:

1. Fund Ideas Without Legal Prerequisites: Many early-stage, socially valuable projects stall due to the requirement of having a legal structure. Direct-access funds for pre-formalized initiatives would enable experimentation where it matters most.
2. Reduce Bureaucratic Barriers: Administrative complexity favors well-resourced actors. For underfunded collectives or student teams, red tape is not a safeguard; it’s an exit point.
3. Rethink Mentoring Models: Many support schemes are tailored to startup logic, offering pitch coaching or investment-readiness advice. However, these seven teams need contextual and flexible support, including backend development support, legal scaffolding, and pathways to public piloting.
4. Broaden Metrics of Success: ROI is not the only meaningful outcome. Projects that enable care, accessibility, or representation contribute real public value and deserve recognition accordingly.
5. Support Innovation from the Margins: Some of OdiseIA4Good's most promising ideas came from migrant associations, artistic cooperatives, or individuals with unstable incomes. These margins are not anomalies; they are reservoirs of lived insight and creativity.

Changing who we fund means rethinking how we define innovation. Foundations, universities, public agencies, and firms have a dual role, not only as funders but as gatekeepers who decide whose ideas qualify and what forms they must take. 

The proposal is not to increase public spending but to rethink its architecture. This could mean: 

• Launching calls open to non-registered groups or early-phase ideas;
• Offering institutional testing grounds in schools, hospitals, or cultural venues;
• Providing technical, legal, and ethical support networks that go beyond traditional investment models;
• Introducing and implementing impact indicators that value collaboration, sustainability, and contextual relevance, not just economic projections. 

The potential exists. Some Hackathon teams seek investment, while others remain small and purpose-driven. What unites them isn't their market path but their proximity to the problem and dedication to solving it with humility and intent. To foster meaningful change, policymakers and funders must shift from rigid, ROI-driven models to flexible frameworks that value lived experience, collaboration, and long-term impact.