When Intelligence Becomes Abundant, What Becomes Valuable?
Written by Sheelpa Patel
It is no exaggeration to point out that many global conversations about AI are dominated by hype and anxiety about what comes next.
Over the past month, as a part of SPARK 2.0, the E-Lab brought together entrepreneurs, technologists, and academics to tease out some of the trends and assumptions that inform both the excitement and the fear: Dr Umang Bhatt (Trustworthy AI, Cambridge), Professor Yonghong Peng (AI, Anglia Ruskin), Mohak Shroff (Microsoft), Ankush Sabharwal (CoRover, BharatGPT), Akshit Kandi (SkySync), and Carl Coddington and Dafydd Biffen (Pippa Labs).
Diverse though each conversation may have been, one overarching question seemed to continually crop up: as intelligence and execution become dramatically “cheaper”, what becomes more valuable? And, as complex as the untangling of this dilemma proves, the role of people, of the human, remains critical. While AI changes the economics of building a company, it doesn't yet remove the need for expertise, judgement, trust, or responsibility.
Replacement, so often a component of either hype or anxiety, became a background to a debate about what exactly entrepreneurs should do given both the possibilities and intricacies of amplification of human capability and given the growing accessibility of these tools. And a few key points really stood out:
From Artificial Intelligence to Institutional Intelligence
Professor Yonghong Peng, Professor of Artificial Intelligence and Deputy Dean for Research and Innovation at Anglia Ruskin University in Cambridge, uses the concept of “institutional intelligence” to understand the debate. His perspective is marked by an attempt to move past individual AI tools and towards the kind of organisations which are built to learn continuously.
What this means, as an argument, it that most businesses still run AI in silos: a copilot here, an agent there, a chatbot somewhere else. But as we move forward, the real shift happens when AI is embedded in how an organisation senses, decides, acts, and learns as a whole. The choice is not, therefore, ‘which model to use?’ but rather ‘how is my organisation designed to get smarter?” Turning every customer interaction, operational decision, and piece of feedback into better decisions system-wide.
What this leaves us with is a different definition of AI-native: not a business that just uses AI, but one designed to embeds a system which is then also able to learn.
Human Domain Expertise: A Bigger Moat than Software
As software gets easier to produce, what gets harder to copy? Akshit Kandi, Founder and CEO of SkySync advised our founders to “go deep, not wide”. His point was essentially that product functionality can be reproduced but that deep domain knowledge comes from understanding customers, workflows, regulation, incentives, and the edge cases of how an industry actually works.
Echoing this, Dr. Umang Bhatt, Assistant Professor in Trustworthy Artificial Intelligence at the University of Cambridge, made a similar point about “taste”. As software generation becomes abundant, the value shifts to knowing which problem is worth solving and recognising a good solution when you see one.
In these kinds of contexts, it seems that competitive advantage will depend less on building another interface, and more on understanding a domain better than anyone else.
AI May Change Headcount and Alter How We Understand Work
Building on this understanding of value, Umang described his team's philosophy at Cambridge's Trustworthy AI Lab: “Bullish on humans. Bullish on building a pro-human future.” His expectation isn't one workforce replacing another, but humans, software agents, and physical autonomous systems working alongside each other.
Startup growth and headcount have moved together for years but AI is breaking that link. Already we have seen that founders can deploy agents across research, coding, sales, marketing, support, and operations with small teams generating output that once required much larger organisations.
In recent months, these “one-person unicorns” get the headlines, but an important question to keep in mind is what happens when one person gains access to capabilities that once required an entire organisation, extending their reach dramatically? If we follow Akshit Kandi's practical framing, which describes managing AI agents like members of a workforce, then perhaps the human doesn't disappear but simply gets more leverage. Leadership becomes less about supervising every task and more about orchestrating different kinds of intelligence.
Should Humans Adapt to AI, or Should AI Adapt to Humans?
A related question came up in a sovereign AI discussion with Mohak Shroff (Microsoft) and Ankush Sabharwal (CoRover, BharatGPT): should humans adapt to AI, or should AI adapt to humans?
The usual assumption is that millions of workers must reskill to fit the technology. But what if founders built the other way round, designing AI to fit existing human behaviour, knowledge, and workflows. Whichever approach you take, if AI strips out the difficult work through which people build judgement, how will the next generation of founders learn what good looks like? Automate the tasks people once learned from, and you also have to redesign how expertise gets built.
The Agentic Web Could Change How Markets Work
Carl Coddington and Dafydd Biffen (Pippa Labs) pushed further: what happens when agents start interacting across the open web?
Individuals and businesses will each have agents acting on their behalf. Searching, comparing, negotiating with other agents, and eventually transacting. This will change the logic of the web itself.
Today's internet runs on attention: businesses pay to reach people, platforms decide what we see, advertising infers what we want. An agent doesn't need persuading by an advert. Give it an intent, and it finds the best capability to fulfil it. That shifts the model from attention-and-advertising to intent-and-capability.
For entrepreneurs it’s worth contemplating: what happens to search when an agent picks the answer? To customer acquisition when agents mediate discovery? What does brand mean when machines influence choice, and who controls the data behind that decision?
The Scarcer Resource?
Ultimately, the opportunity is to decide what should be automated, what should be augmented, and what should remain distinctly human. Because as intelligence becomes abundant, human judgement may become the scarcer, and more valuable, resource.
Sheelpa Patel is a growth advisor and founder of Mavens + Mavericks, connecting entrepreneurs, investors, industry leaders and international innovation ecosystems to help ambitious ideas achieve global impact. Her experience spans autonomous vehicle AI at Oxford-University spinout Oxa, and building and scaling INFINITI LAB, Nissan’s global startup accelerator, alongside initiatives championing women entrepreneurs across Southeast Asia. Sheelpa is a Senior Advisory Board Member of the E-Lab and spearheaded the creation of SPARK, the incubator helping Cambridge University’s most ambitious founders and researchers transform early-stage science and research into investor-ready ventures. She also serves on the Oxford–Cambridge Supercluster Board’s Science & Technology panel, advancing collaboration across one of the UK’s leading innovation ecosystems.