What does responsible AI look like in practice?
What eight social impact ventures can teach us about building technology for opportunity
19 October 2026, London, UK
Image: AI and Social Mobility Challenge Prize, Cohort 2026
What does it take to build AI that genuinely widens opportunity, rather than reinforcing the inequalities that already exist? Responsible AI is often discussed through principles such as fairness, transparency and accountability. But what do those principles mean for a venture developing personalised learning for children, an AI careers adviser or a platform connecting young people with employment opportunities?
The answers depend on who the technology is designed for, what it is intended to do and the circumstances in which people will use it.
In our first AI & Social Mobility Insights article, we explored five patterns emerging across the wider innovation landscape. Our second article introduced eight ventures addressing different barriers to education, employment and opportunity. Now, as the cohort approaches the final stages of the AI and Social Mobility Challenge Prize, we are looking more closely at how these ventures are approaching responsible innovation. The Challenge Prize is delivered by EY Foundation and Social Tech Trust, in partnership with the Social Mobility Commission, and supported by Microsoft, HSF Kramer, EY, Fieldfisher and LinkedIn.
Drawing on the approaches, experiences and early evidence shared by the eight ventures, we can identify four practical considerations for anyone developing AI intended to widen access to opportunity:
making access genuinely inclusive,
involving young people in design,
building trust through safeguards and human oversight, and
demonstrating meaningful impact.
1. Access needs to be genuinely inclusive
Making a digital product available is not the same as making it accessible.
Cost, connectivity, disability, digital confidence and access to devices can all influence whether someone can use a technology. For young people from lower socio-economic backgrounds, these barriers may overlap with wider disadvantages in education, employment and access to support.
Breakthrough AI works with young people who face barriers to participating in the digital economy, including those with experience of the justice system, care-experienced young people and those not in education, employment or training. The organisation has supported more than 700 associates through its programmes. Its AI-powered learning companion is being developed to provide personalised coaching and practical learning opportunities that adapt to a young person’s experience, literacy and goals. An important part of Breakthrough’s approach is helping young people become critical users of AI, able to question its outputs, recognise potential bias and understand their data rights.
FIIT Learning approaches accessibility through education. Its AI-powered platform is being developed for young people who struggle to access conventional schooling, including those experiencing persistent absence, emotionally based school avoidance or special educational needs and disabilities. The aim is to provide flexible, personalised learning that adapts to individual circumstances, while maintaining human support and oversight.
Louno Space focuses on access to STEM education at an earlier age. Through its existing workshops and programmes, it reaches thousands of children across over 100 schools. Its AI learning companion, Nova, is being developed to offer personalised STEM missions that children can explore beyond face-to-face sessions, potentially widening access to enrichment that might otherwise depend on family resources or school provision.
These ventures are tackling different problems, but they share a common question: how can technology make valuable learning and support more accessible without creating new barriers for the people who need them most? Responsible AI starts with understanding who might still be excluded, and designing with those barriers in mind.
2. Participation should influence what gets built
Young people are often invited to test technology that has already been designed for them.
Their feedback can be valuable, but participation becomes more meaningful when people have opportunities to influence decisions before they are finalised. This is particularly important when developing products for young people whose experiences may differ significantly from those of the founders, developers or institutions commissioning the technology.
HopeHub is developing Jords, an AI-supported goal coach within a wider digital youth-work platform.
The venture involves hundreds of young people and professionals in its co-design process. Its early feasibility pilot with the Mayor of London’s Violence Reduction Unit involving young people helped the team explore how digital support could contribute to motivation, goal-setting and access to opportunities.
MyPocketSkill connects young people with paid earning opportunities while integrating financial education and skills development. The venture involves young people through its ambassador programme and community co-design activities. Its approach recognises that young people’s experiences of earning, managing money and working with clients should help inform the services designed to support them.
Tangent, which aims to widen access to professional networks and career opportunities, conducted hundreds of user interviews, alongside surveys and school-based focus groups, to understand the challenges people face when entering the labour market.
These examples show different ways ventures can bring users’ experiences into product development.
But the most useful question is not simply how many people have been consulted. It is what changed because they were involved. That might mean changing the language or accessibility of a product, reconsidering how recommendations are presented, introducing additional safeguards or recognising that a proposed feature does not address the problem users are experiencing. Responsible innovation requires being willing to challenge assumptions, including those that shaped the original idea.
3. Trust requires transparency, safeguards and accountability
When AI influences someone’s education, career guidance or access to opportunities, trust matters.
People need to understand when AI is involved, what it can and cannot do, and where they can turn when something goes wrong. The risks are different across technologies. An AI learning companion for children raises different questions from a recruitment platform or a digital youth-support service.
Ultra Education is developing Abby, an AI-powered entrepreneurship mentor designed to provide personalised, age-appropriate guidance to young people. Ultra’s established education and mentoring programmes work with around thousands of young people annually through more than 200 schools and 15 local authorities. Abby is at an earlier stage of development. The venture is designing it to support, rather than replace, teachers, parents, counsellors and safeguarding professionals. Human oversight remains important, particularly where young people need support beyond the scope of an AI system.
HopeHub has also described clear boundaries around its technology. Its approach includes escalating safeguarding concerns to trusted professionals rather than allowing AI to make those decisions independently. The team reports postponing peer-to-peer features to reduce potential risks during early development.
For Tangent, the challenge is different. Professional networks and recruitment processes can reproduce existing inequalities, including biases affecting who is considered for an opportunity. The venture describes an anonymised candidate-review option intended to reduce the influence of identifying characteristics. It also positions AI as a tool to support candidate discovery, with employers retaining responsibility for hiring decisions.
These approaches reflect an important principle. Responsible AI is not about removing people from important decisions. It is about understanding where technology can provide useful support, where human judgement remains essential, and who is accountable when something goes wrong. Transparency and explainability are part of this. People should be able to understand how AI is being used and question important outputs or recommendations that affect their opportunities. Trust cannot simply be assumed because a product is technically effective. It has to be earned through the way technology is designed, governed and used.
4. Impact needs evidence, not just ambition
For ventures working on social mobility, success cannot be measured only by how many people use a product.
A learning platform may attract users without necessarily improving educational outcomes. A careers tool might provide useful information without helping someone take the next step. A professional network might create connections without those connections leading to opportunities. If the goal is social mobility, we need evidence that something meaningful is changing.
AVA Academy is developing an AI-supported approach to university admissions and career pathways, with a focus on making personalised preparation more accessible to students who may not have access to expensive tutoring or professional networks. AVA conducted a randomised controlled trial across several medical schools, comparing its AI-supported approach with a video-based course. The AI-supported group demonstrated more than five times the engagement of the comparison group.
This provides an example of how ventures can begin testing whether technology changes how people engage with learning, while continuing to examine what that engagement means for longer-term outcomes. Evidence from an existing programme can inform the development of a new technology, but it does not automatically prove that the technology will achieve the same outcomes. For ventures working on social mobility, the real measure of success is not simply how many people can access a product. It is whether that access changes what someone can learn, do or achieve. This requires clarity about what has already been demonstrated, what remains uncertain and what evidence still needs to be collected.
The Youth Voice Forum asks…
EY Foundation’s Youth Voice Forum has challenged us throughout the programme to consider AI from the perspective of the young people whose opportunities it is intended to influence. Its manifesto, Don’t Build AI Without Us, calls for AI that is accessible, understandable and fair, and for young people to have a meaningful role in shaping the technology that affects their futures.
Two questions are particularly relevant to the approaches explored in this article:
How have young people helped shape your product, and what changed because of their involvement?
How will you know your technology has genuinely improved opportunities, rather than simply increased efficiency or profit?
These questions bring us back to the relationship between participation and accountability. Young people should not only be protected from potential harm. They should also have opportunities to influence the decisions, technologies and systems that affect them.
Building technology that earns trust
Across the eight ventures, responsible AI takes different forms. It can mean making learning more accessible to young people whose needs are not met by conventional education. It can mean creating opportunities for young people to influence product design, introducing safeguards for children, maintaining human oversight or addressing potential bias in recruitment.
It also means being prepared to test assumptions and demonstrate impact. The cohort includes ventures at different stages of development, with different levels of evidence and different approaches to measuring success. That diversity is valuable. It reminds us that responsible AI is not a fixed set of features that can simply be added to a product. It is an ongoing process of understanding risks, listening to people, making considered decisions and learning from evidence.
AI alone cannot overcome the structural barriers that shape social mobility. Education systems, employers, public services and communities all influence the opportunities available to young people.
But technology may help change how people access learning, guidance, networks and support.
The challenge is ensuring that those possibilities translate into meaningful improvements in people’s lives.
As the eight ventures prepare for the final judging session and the Challenge Prize Parliamentary showcase on 29 October, we will be highlighting their work and the different approaches they are taking to widen access to opportunity on Social Tech Trust's LinkedIn page. The winner and runner-up will be announced at the Parliamentary event, marking an important milestone in the ventures’ journeys.
Follow the final stages of the AI and Social Mobility Challenge Prize as we introduce the eight ventures and explore how responsible innovation can help widen access to opportunity.
