AI Leadership: Perspectives from Academia and Venture on the Future of AI
Rapid advancements in AI will continue to transform the ways in which we work and live. To understand the scope of its impact is complex, requiring an understanding of multiple perspectives that transcend both practice and research. As we continue to push the frontier of AI capabilities, it is critical to shape its trajectory with foresight and responsibility. To learn more about how leaders can prepare for the future of AI, we spoke to leaders spanning industry and research who offered insights into the core technology and its business implications.
Shaun Johnson, Paul Grigas, and Alexander Fred-Ojala (Alex) each offered a unique perspective on how AI technology is evolving and how leaders can tackle the opportunities and challenges. Shaun Johnson is the founding partner at AIX Ventures, where he invests in early AI-native ventures, leveraging his background in engineering, product, and design.
On the academic side, Paul Grigas, associate professor in the Industrial Engineering and Operations Research department at UC Berkeley, has spent his career designing and analyzing algorithms for optimization, machine learning, and data-driven decision-making.
Bridging both academia and industry, Alexander Fred-Ojala studied mathematical statistics and taught applied data science courses in the College of Engineering at UC Berkeley before stepping into his current role as the head of AI at EQT Ventures, the largest early-stage venture capital firm in Europe. He has additionally served as the Research Director of SCET’s Data Lab. Despite each of their unique backgrounds and expertise, our conversations revealed several converging themes for AI leadership.
Effective AI leadership requires both a rigorous understanding of AI’s technological capabilities and methods, deep understanding of the market and high-potential opportunities, and the foresight and responsibility to set up their organizations for success. In the long term, they also need the ability to take action and make prudent decisions.
Here’s what AI leaders need to know:
Where Are We Today?
AI as a Foundational Toolbox
Paul Grigas, whose work lies at the forefront of this rapidly evolving frontier, described AI as a “foundational technology,” a toolbox applicable to every domain.
“I think AI has huge potential to accelerate scientific discovery, engineering, and the development of products and tools to better society and humanity, whether that be through improvized medication, improved logistics, etc.”
AI as an Accessible and Abundant Source of Intelligence for All
Paul additionally emphasized the increasing accessibility of these technologies. Anyone can access these state-of-the-art models and integrate them into their work, accelerating and enhancing productivity. The growing accessibility of AI democratizes its use, and transforms intelligence into an abundant commodity available to all.
“Anyone will have access to a PhD-level expert in their pocket on any topic, no matter if you are researching biotech or macroeconomics. The capabilities of AI to reason with unstructured information, process text, and communication will unlock so many opportunities for us as a species.”
Investing in AI
When evaluating AI startups, Shaun Johnson highlights investment at the intersection of AI-nativeness and market savviness—in other words, investing in teams who exhibit both strong founder-product fit and founder-market fit.
“We’re focused on thinking about teams that can bring AI experiences to the market that others cannot, and we’re thinking about teams that understand the market and have a non-consensus viewpoint as to why the market is going to evolve.”
However, Shaun noted the difficulty of finding founding teams with strong capabilities in both areas.
“Very often you might back AI-native founders that need to become more market savvy, or market savvy founders that may need to infuse more AI-nativeness.”
AI as a Driver of Velocity
When it comes to investing in AI companies, Alex adds that assessing velocity is a key factor in evaluating potential portfolio companies.
“Companies are growing at unprecedented speeds when it comes to the revenue they are generating in a short amount of time, or the millions of users they’re serving in only a couple of months.”
The combination of global information infrastructure and AI functionality allows for the instantaneous sharing of knowledge and enables leaner teams to scale efficiently at lower costs.
Traits of a Successful AI Entrepreneur
In terms of the founders themselves, Shaun identifies several promising characteristics in an AI-native founder. In particular, he looks at “horsepower,” which refers to a timely and high rate of good judgement.
“When you’re starting a company, startups must have execution speed. You want to know that the CEO has clarity of thinking at a very high rate.”
When it comes to the age of AI, it is critical to move quickly. Shaun emphasizes the sense of urgency for founders to optimize for speed and maintain a rigorous rate of learning. Founders should reflect routinely on what they have learned on a weekly basis, as well as how they can accelerate their progress in the week ahead. Additionally, Shaun notes that clear goal setting and honest self-assessment and reflection are key ingredients to success.
“Making sure you have a hypothesis that you’re testing—and being very objective with yourself as to whether or not those tests are passing or failing—and moving on quickly is incredibly important.”
What We're Getting Wrong Today and Why
Alex believes that business leaders are overestimating AI capabilities in the short term but underestimating its scope in the long term.
Keeping Up With the Pace of AI
Alex notes, “Today, is it important to not approach product-building with a shortsighted vision, merely attempting to keep up with the pace of AI advancements rather than anticipate its development.” Leaders must extrapolate and forecast where AI capabilities will be in the next six to twelve months.
“Avoid spending a lot of time, effort, and capital to optimize something that a general tool will be capable of. It’s better to build for the future. Enter specific niches that a general tool will have a hard time optimizing for.”
The pace of AI development and deployment shows no signs of waning, and whether business leaders embrace AI will distinguish the industry winners from the underperformers.
Challenging AI as a Universal Solution
Paul Grigas pointed out that artificial general intelligence (AGI) is a lot more hype than reality in conversations today.
“Having AGI as the goalpost is motivating for people to develop enhancements to AI and better AI tools and systems, but I feel that currently, AI is a tool people are using to really accelerate and enhance their productivity. It’s hard for me to see a world where we are in some kind of utopian society where AI is doing everything. I do think that's a human endeavor. ”
Business leaders must rethink how they approach problem-solving with AI. While the AI technology itself is not the remedy, it can address specific problems within domains. Championing AI as a universal solution to broad challenges in healthcare or climate change is impractical, as domain-specific knowledge and high-quality training data are critical to achieving effective outcomes.
Premature Confidence in AI Autonomy
Furthermore, business leaders tend to overlook technical nuances of AI systems, leading them to overestimate the capabilities of their models. As a result, they operate under the flawed assumption that these systems are going to be 100% reliable when in reality, human judgement and oversight remain essential for refining accuracy and checking source attributions.
Alex said, “It won’t be 100% perfect because these systems are still probabilistic. We’re going to get there in a few years, but that’s not going to happen today.”
AI Integration: Easier Said Than Done
Furthermore, business leaders less familiar with the technical details overestimate the ease with which AI can be integrated with their legacy systems, such as certain software or workflows that their company has had in place for decades.
“It’s not like the AI systems will be able to integrate into [legacy systems] and use that off the shelf. This process requires costly integration to supply the model with the right context.”
However, despite steadily decreasing compute costs, feeding AI models large amounts of data is still expensive and time-intensive.
Preparing for the Next Generation of Work
As AI changes the nature of work, it is critical that everyone—regardless of their domain—be well-versed in its applications. With its incredible capability to accelerate productivity, businesses must make a worthwhile investment in skilled talent, and allocate the proper time and resources to train employees. It is also important for leaders to establish and maintain the alignment of employees’ skill sets with broader objectives and organizational strategy.
Alex emphasized the importance of seeking people who are not only equipped with the technical background to tackle problems using AI, but also “systems-level thinkers.” People who can think about integration and structuring complex systems through a systems-thinking lens are well-poised to create value in this AI-powered workforce. Equally important are high-energy individuals who display a strong sense of agency and creativity while understanding exactly what end users want.
Ethical and Social Considerations
As AI becomes more ubiquitous, certain issues of fairness, sustainability, and regulation must be addressed to ensure the responsible implementation of AI. Responsible deployment must happen on three levels: in the engineering of AI itself, adoption by users, and conscientious funding.
Running and training large AI models consumes copious amounts of energy, and businesses are allocating significant resources to these processes. With this investment comes an obligation to closely monitor their carbon footprint. Because data centers are so power-hungry, they strain local infrastructure, and the presence of data centers operating in small communities have disrupted local activities. Paul noted that “there’s a lot of inequity and not enough regulation” around the environmental aspect. He advocates for extra emphasis on engineering efforts devoted to managing the energy use of a model and detect more accurately the complexity of the prompt to avoid wasting energy and resources.
Furthermore, fairness and mitigating algorithm bias is a critical, non-negotiable issue, and is paramount when integrating AI with platforms that ultimately make decisions that directly or indirectly affect the lives and welfare of people. Traces of racial bias can lead to discrimination in action, perpetuating harmful practices.
Alex noted that everyone building, funding, and regulating these systems need to make sure that AI is trustworthy, reliable, sustainable, and secure as the systems become more widespread and autonomous. He emphasized that these factors must be carefully considered while at the same time promoting innovation and competitiveness.
What's Next?
Paul characterizes the current state of AI as akin to the “early days of the internet.”
“It’s almost like the Wild West. We’re very primitive in how we are accessing and using these models.”
As a result, AI hallucinations and misinformation pose harmful threats to its reliability, issues that must be addressed as AI use continues to be more widespread and deeply embedded. Paul notes that, with the internet now, we are able to distinguish between reliable and unreliable sources; however, output from large language models lack those same trust frameworks.
"I think there needs to be more effort put into building a framework for understanding why AI models say what they say, and how reliable those outputs really are. It shouldn’t just be about telling you it’s 95% confident. It should also provide citations, sources, and a better sense of where the information is coming from, even which parts of the training data informed a specific response. That kind of framework just doesn’t exist yet."
In the future, Shaun is excited to see more tech-enabled services. As an investor, he is particularly passionate about finding a team of one cofounder with a deep understanding of a vertical, typically service-oriented, and another cofounder with AI-native building experience. Together, he believes that they can create a business “quite lean and scalable.”
Advice for Aspiring AI Founders
Alex stated, “There’s never been a better time in history to become a first-time founder and entrepreneur. The barriers have lowered, with agentic systems and external cognitive engines that can reason for us, help us build software, handle marketing tasks, or customer support. Every founder should lean in.”
Shaun advises entrepreneurs to carefully approach their AI solutions through ample reading and discussion, and understand deeply and precisely where their venture will operate. In particular, he distinguishes between two distinct directions for AI entrepreneurs: heat-seeking opportunities and truffle hunting opportunities. He encourages founders to pick their lane intentionally and conduct deliberate preliminary research to determine where their venture is best poised to succeed.
“Heat-seeking” refers to the pursuit of a space that is obvious, hot, and buzzy. To succeed in this space, one must have “a reason that they can win over the dozens of other companies that are going into that space.” On the other hand, “truffle hunting” refers to entering a quieter, under-served space that has historically received less attention. However, founders must understand why the space appears underserved, ensure that they are not entering a startup graveyard, and have a strong rationale for entering before executing.
Similarly, Alex encourages entrepreneurs to avoid building a general platform for a broad problem but rather focusing on solving a specific problem in a niche that they can scale.
The opportunities are limitless, and AI advancement has made it possible to enter domains that may not have been addressed by software before. Alex notes that, because AI is beginning to think like humans, the TAM for AI has increased dramatically to extend to the services sector as well. It has become feasible to solve tasks end-to-end that previously only humans could, and AI can now solve problems in areas of the economy that weren't possible with tech in the previous era of non-AI software.
From academia to industry, harnessing the power of AI will unlock new opportunities and solutions across all domains. Founders and business executives must be prepared to optimize for speed and good judgment, while making sure to innovate responsibly and ethically.
Interested in Learning More?
Interested in learning more insights from the UC Berkeley faculty and industry community? Our AI Leadership Accelerator is specifically-designed to help executives and senior leaders create robust AI strategies for their organizations.