When AI stands for augmented intelligence

IMAGINE you are given a million dollars to invest in a pool of 15 startups. You have two hours to review their pitch decks and make a swift but sound assessment. Do you spread your eggs across several baskets or go big on one outstanding startup?

You mull over your choice, measuring them against your standards of innovativeness and business potential. Drawing on your industry experience, you identify the stars and eliminate the weaker teams. Rinse and repeat.

This is essentially the challenge confronting universities, investors and partners involved in open innovation – how to scale this process of selection and judgment. As universities take on a more active role in nurturing entrepreneurship, they must evaluate startups fairly and allocate scarce resources including funding, mentorship and incubator space to the ventures with the greatest potential.

These evaluation criteria and resulting datasets become core institutional assets, fundamental to the university’s credibility in identifying and supporting innovators.

At Singapore Management University, this imperative was at the forefront when we hosted the 12th Lee Kuan Yew Global Business Plan Competition. With more than 1,500 submissions from over 1,200 universities across 91 countries, we faced a complex challenge: ensuring that judging remains fair, transparent and free from the cognitive biases that shape human judgment.

Indeed, research shows that roughly one-third of venture capital transactions involve founders and investors who hail from the same university because venture capitalists are significantly more inclined to back startups led by fellow alumni.

Other research shows that globally, even though female-founded startups often yield higher returns, they receive less than 5 per cent of venture capital. These human biases are real, and need to be addressed rather than amplified.

Is artificial intelligence (AI) the solution? Our exploration of generative AI in this context began in 2023, when we collaborated with Valuer.AI, a platform known for AI-driven startup scouting.

Valuer.AI identified a startup, Medea Biopharma, from the Technical University of Munich, that our 200 human judges had missed. Medea ultimately won the pre-revenue category, and has since established a cutting-edge S2 laboratory.

This demonstrated the value of human-AI complementarity: While expert panels bring contextual and industry-specific insights, AI uncovers overlooked opportunities and mitigates bias.

These lessons paved the way for the DueAI Challenge, where AI is tasked with due diligence on all startups. In this competition, 14 startups developed AI agents to evaluate pitch decks using their proprietary tools and rubrics. They worked alongside more than 230 international judges assessing innovation, scalability, market potential and financial viability.

The AI agents were evaluated on three criteria: expert alignment; explainability and transparency; and ability to identify outliers.

Expert alignment refers to how closely the AI’s selections mirror those of the international judging panel, measured by the proportion of overlapping top picks.

Explainability and transparency assess the AI agent’s ability to provide evidence-based insights and clear, succinct justifications for each shortlisted startup. Identifying outliers focuses on how AI agents surface promising ventures overlooked in earlier human reviews, thereby widening the pool of potential high-impact startups.

The results were telling. While AI could not yet pick winners, it excelled at identifying false negatives – promising startups that had been eliminated. Three of the eight startups that advanced to the grand finals emerged from this outlier pool, having been overlooked by human judges but flagged for reconsideration by multiple AI agents.

This shows how AI can expand the opportunity landscape while strengthening due diligence in large-scale competitions that span sectors and continents. Rather than replacing human judgment, the DueAI approach revealed blind spots that can stem from unconscious biases, shaped by past preferences and familiarity with certain sectors or technologies. By augmenting human discernment with AI, we can enable crucial investment choices to be made with greater objectivity.

But the DueAI outcomes also raise broader questions. With AI increasingly shaping decisions in Singapore’s public services, how do we ensure that these systems are both intelligent and accountable?

For instance, the Land Transport Authority plans to use AI to optimise road safety and bus routes, and the Singapore Police Force uses AI-enabled systems to process crime report data for surfacing previously hidden connections and patterns across cases.

While such tools will undoubtedly improve efficiency, the societal impact of their decisions demands heightened scrutiny around explainability, bias mitigation and ethical safeguards.

As Singapore advances towards a future where intelligent systems will support more tasks, from infrastructure planning to environmental monitoring, the challenge is how to deploy AI responsibly and sensibly.

As a global AI hub, Singapore has looked to proactively address these risks with its Model AI Governance Framework, emphasising human oversight, transparency, and explainability.

However, having frameworks alone is not enough. We also need to better expose students to these ethical principles and how to put them into practice.

Equipping our students with judgement to strategically and ethically use AI is a key education mission today. In Singapore’s universities, ongoing pedagogical shifts enable students in medicine, law and computing to use AI to design assistive technologies, practise cross-examinations, and deepen their engagement with learning.

These experiences build students’ confidence and prime them for a future in which human-AI collaboration will be the norm. Yet as AI systems grow more powerful and pervasive, students must also be trained to anticipate ethical risks, question opaque outputs and recognise when algorithmic decisions may compromise fairness or accountability.

Meaningful exposure to AI ethics ensures that they can engage with intelligent tools responsibly and safeguard public trust as future professionals.

The DueAI experience demonstrates that innovation and integrity can go hand in hand. This is the promise of intelligent accountability: a model where AI can support human judgment rather than eclipses it, strengthens trust rather than weakens it, and ultimately enhances – rather than undermines – human potential.

The writer is vice-president, partnerships and engagement at Singapore Management University and Lee Kong Chian professor of communication and technology. Her latest book is Humanising Technology: Reflections on Design, Ethics and Inclusion.