Co-Intelligence and the Future of AI in Business: A Conversation with Dr. Ethan Mollick

The rapid advancements in artificial intelligence, especially large language models (LLMs), are poised to transform nearly every aspect of business. To understand the implications and potential of this technology, Gene Kim sat down with Dr. Ethan Mollick, associate professor at the Wharton School, University of Pennsylvania and author of the New York Times bestseller Co-Intelligence, at the 2024 Enterprise Technology Leadership Summit Virtual Europe.

Dr. Mollick, who has studied innovation and entrepreneurship for many years, sees three key trends in AI that will shape the future: specialized devices, large context windows, and agents. He notes that while he is not a technical expert himself, his focus has always been on the business applications and societal implications of AI, dating back to his work with pioneering computer scientist Marvin Minsky at the MIT Media Lab.

A key insight from Dr. Mollick’s book is that employees across organizations are already quietly experimenting with using AI tools to dramatically improve their productivity and work quality—but they are often reluctant to share these innovations with their employers. He attributes this reticence to several factors:

1) Unclear rules and fear of punishment for “incorrect” usage of AI.
2) Desire to be seen as a “wizard” delivering amazing results without revealing the AI assistance.
3) Concern about job security if the extent of productivity gains from AI becomes known.
4) Lack of incentives or recognition for initiative in leveraging AI.

To overcome this and foster a culture of open innovation around AI, Dr. Mollick advises leaders to radically rethink incentives. This could range from substantial spot bonuses for the most impactful AI prompts each week to job security guarantees to assuage fears around AI-driven layoffs. Ultimately, organizations need to adapt their culture to one of trust and sharing for AI breakthroughs to come into the open.

Looking ahead, Dr. Mollick cautions against overreliance on retrieval-augmented generation (RAG), a technique of using AI to analyze an organization’s own documents to generate highly relevant responses. While RAG has valuable applications, such as in well-bounded domains like customer service, leaders must recognize that LLMs can still hallucinate convincing nonsense even when fed curated data. Additionally, the current cost constraints that make RAG appealing compared to ingesting everything into an LLM’s context window may diminish rapidly as the technology continues its exponential improvement.

Dr. Mollick’s core advice for technology leaders is simple yet challenging: “Use AI for everything.” Too many organizations are either not leveraging the technology at all outside of software development or are artificially constraining its usage. To build true competence and discover transformative opportunities, leaders and their teams must commit significant hands-on time with the most advanced LLMs available, pushing them to their limits.

This exploratory usage should expressly include probing for weaknesses and failure modes, not just aspirational use cases. Dr. Mollick emphasizes that even the most powerful LLMs today have immense room for improvement in areas like prompt engineering and response quality metrics—and end users outside the organization likely have access to equally capable models without the same usage constraints.

To navigate this new reality, Dr. Mollick advocates adopting the mindset of “building for the future” even if model capabilities temporarily plateau. Concretely, this means designing systems that can readily swap in more advanced LLMs as they emerge rather than over-optimizing around today’s cost and performance limitations. For organizations facing short-term delivery pressures, he suggests a two-track approach: assembling a minimum viable product with current tools while concurrently preparing to leapfrog forward as the technology matures.

Zooming out, Dr. Mollick calls for the AI community to establish greater “agency” in shaping the future of the field through more public communication and new benchmarks. He notes that by virtue of being an early, prominent voice on the business implications of AI, his ideas have gained surprising influence—and technologists have a similar opportunity and responsibility to highlight positive examples of AI augmenting human potential rather than replacing jobs wholesale.

Finally, Dr. Mollick contends that the most widely used benchmarks for language models, which skew heavily toward formal math and coding problems, fail to capture the fuller range of linguistic and reasoning capabilities that matter for business use cases. He sees a vital need for the practitioner community itself to create new open benchmarks assessing performance on realistic conversational and analytical workflows, as well as explicitly testing for alignment with human values.

While the road ahead for enterprise AI will undoubtedly have many twists and turns, Dr. Mollick remains optimistic about the potential for this technology to unlock human creativity and prosperity on an unprecedented scale. But realizing that potential will require active, thoughtful stewardship and open collaboration between technology leaders across industries. The next decade will be a defining one—not just for artificial intelligence but for deciding what kind of future we want to build with our increasingly intelligent machines.

Watch the full interview in our video library here.

Sign up for the next Enterprise Technology Leadership Summit here.

- About The Authors

Summary by IT Revolution

Articles created by summarizing a piece of original content from the author (with the help of AI).