Speaking with TechGraph, Arjun Balaji, Co-Founder and Programme Director of ImpactAI Foundry, discussed how the organization is helping make artificial intelligence adoption more accessible for nonprofit organisations by combining practical guidance with solutions tailored to the day-to-day realities of the social sector.
Balaji also spoke about how the organisation embeds responsible decision-making into the adoption process from the outset, helping nonprofit teams build confidence in using AI while ensuring the technology continues to support their long-term mission and impact.
Read the interview in detail:
TechGraph: AI adoption today is largely being driven by enterprises with strong technical and financial resources, while many nonprofits are still at the stage of understanding how these tools fit into their work. What made ImpactAI Foundry focus specifically on building AI readiness within the social impact ecosystem?
Arjun Balaji: The gap was too visible to ignore. Enterprises have dedicated technology teams, budgets, and vendors helping them navigate AI adoption. Nonprofits have none of that, and yet they are expected to keep pace with a rapidly shifting landscape while running lean on every front. The social sector was not being left behind by accident. It was simply not being included in the conversation.
We built ImpactAI Foundry because we believed that with the right support, nonprofit teams could build and own practical AI tools themselves. Not as a future aspiration, but right now, with hands-on guidance tailored to their actual work.
TechGraph: Many nonprofit teams operate with limited bandwidth and immediate execution pressures, which makes structured learning difficult. In such an environment, what are the biggest obstacles organisations face when trying to build practical AI capability internally?
Arjun Balaji: The biggest obstacle is not technical; it is bandwidth. Most nonprofit teams are so deep in execution that carving out time to experiment with anything new feels impossible, even when they can see the potential value. There is also a confidence gap. People assume AI requires a level of technical knowledge they do not have, which makes the starting point feel much further away than it actually is.
The third obstacle is relevance. Generic AI training rarely connects to the specific workflows a team deals with every day, so enthusiasm fades quickly. What actually works is starting with a real problem the team is already frustrated by and building from there.
TechGraph: Unlike businesses that often adopt AI to improve efficiency or profitability, nonprofits tend to evaluate technology through the lens of accessibility and community impact. How does this difference shape the way social sector organisations approach AI adoption?
Arjun Balaji: It makes them more careful, which is not a bad thing. Nonprofit teams tend to ask harder questions upfront: who does this serve, could this cause harm, will this work for the communities we are accountable to? That instinct is valuable, even if it sometimes slows things down.
It also means the bar for usefulness is different. A tool that improves a business metric is considered a success in most enterprise contexts. In the social sector, the question is whether it actually changes something for the people the organisation exists to serve. That is a more demanding standard, and it pushes toward more grounded, thoughtful adoption rather than technology for its own sake.
TechGraph: There is increasing pressure on nonprofits to strengthen documentation, reporting, and impact measurement as funding environments become more outcome-focused. Where do you see AI creating the most practical value for organisations managing these responsibilities with limited resources?
Arjun Balaji: These are exactly the areas where AI can make an immediate difference. Reporting and documentation consume a disproportionate amount of time in most nonprofit teams, time that could otherwise go toward direct programme work. AI tools can help draft reports, synthesise field notes, structure data, and generate first versions of documents that teams then review and refine. The time savings are real, and they add up quickly.
Impact measurement is slightly more complex because the underlying data is often fragmented or inconsistently collected. But even here, AI can help organisations make better sense of what they already have, identify patterns, and communicate outcomes more clearly to funders and stakeholders.
TechGraph: A large part of the current AI ecosystem is designed around mainstream business workflows, while nonprofits frequently work with regional languages, fragmented information, and underserved communities. Do you think today’s AI tools are realistically equipped for these on-ground conditions?
Arjun Balaji: Honestly, not fully. Most AI tools are built around assumptions that do not hold in many nonprofit contexts: reliable internet connectivity, structured data, standard workflows, and users who are comfortable with technology. When those conditions are not met, the tools underperform or require significant adaptation.
That said, the gap is closing. And the more important point is that even imperfect tools, used thoughtfully, can create real value. The work is not to wait for perfect tools but to figure out where existing ones are genuinely useful and build from there. That is the approach we take with every organisation we work with.
TechGraph: As AI becomes more accessible, concerns around misinformation, bias, and responsible usage are also becoming more important, particularly for organisations working with sensitive community data. How does ImpactAI Foundry approach AI education in a way that keeps ethical awareness central to the learning process?
Arjun Balaji: We weave it into the build process rather than treating it as a separate module. When an organisation is deciding what tool to build, questions about data privacy, consent, and potential bias come up naturally as part of that conversation. Who will this tool affect? What data does it use? What happens if it gets something wrong? These are not abstract ethics questions; they are practical design questions, and nonprofit teams are often well-equipped to think through them because they are already accountable to communities in ways most businesses are not.
The goal is not to make teams anxious about AI but to make them thoughtful users of it. That is a more useful and more durable outcome than any technical skill we could teach.
TechGraph: Technology training often creates short-term enthusiasm, but long-term adoption usually depends on whether teams can connect those tools to their daily work. What have you learned about how nonprofit professionals move from experimentation to consistent AI usage in practice?
Arjun Balaji: The shift happens when a tool solves something specific that was genuinely painful. Not a demonstration, not a prototype built during a session, but something a team member actually uses on a Tuesday afternoon because it makes their work easier. That moment of real utility is what changes the relationship with the technology.
What we have also learned is that peer credibility matters enormously. When someone on the team becomes a confident user and shares that with colleagues, adoption spreads much faster than any formal training could achieve. So part of our focus is identifying and supporting those internal champions, because they carry the culture change forward long after our programme ends.
TechGraph: Looking ahead, if AI skills become as essential as digital literacy over the next few years, what role do you see organisations like ImpactAI Foundry playing in preparing the nonprofit sector for that shift?
Arjun Balaji: The role is access and translation. As AI becomes more embedded in how organisations operate, the risk is that the gap between well-resourced and under-resourced organisations widens further. Organisations like ImpactAI Foundry exist to make sure that does not happen by default.
The translation piece matters just as much as access. AI capability is not just about knowing which tools exist. It is about understanding how to apply them to your specific context, constraints, and community. That requires human support, not just online courses. And it requires people who understand both the technology and the sector. That is the space we are trying to occupy, and we think it will only become more important over time.
