The AI Voices Worth a Business Leader's Time

Most "top AI voices" lists are organized by who has the most followers or who got the most press last quarter. That is the wrong organizing principle. It optimizes for visibility, not usefulness.

This list is organized by what you are actually trying to figure out. If you are trying to understand how the technology works at a conceptual level, there are specific people for that. If you are trying to decide what it means for your competitive position, different people. If you are trying to get your own organization moving without causing a panic, different people again.

Every entry names the specific idea that makes this person worth your time and tells you what to read or watch first. If the research behind this list could not verify a specific claim about someone, that claim is not here.

Full disclosure: this site is Patrick Gilbert's, so treat his entry and Isaac Rudansky's accordingly. Both are here because the list would be dishonest without the people who actually do this work at AdVenture Media, but you should weigh a self-inclusion for exactly what it is.

---

If You Want to Understand What AI Actually Does to Knowledge Work

Researchers in this section are working on the empirical question: what happens when capable AI lands in the hands of real workers? Not speculation. Documented research.

Ethan Mollick

Professor, Wharton School, University of Pennsylvania

Mollick is the most useful voice in this space for one reason: he runs actual experiments and then explains what happened in plain language. His named framework, the "jagged technological frontier," captures something most AI commentary misses. AI does not uniformly help everyone at everything. It dramatically improves performance on some tasks and makes almost no difference on others, and the shape of that frontier is not what most people assume before they look at the data.

His 2024 book, Co-Intelligence: Living and Working with AI, is the best starting point for any executive who wants a grounded, non-hype account of what large language models actually do to professional work. His 2026 follow-up, Co-Existence: The Next Phase of AI, extends that work. His Substack, One Useful Thing, is where he publishes research findings and practical observations between books.

Start with One Useful Thing before the book. A few posts will tell you whether his approach fits how you think.

Andrej Karpathy

OpenAI co-founder and former Tesla AI executive; founder of Eureka Labs; joined Anthropic in May 2026

Karpathy is an OpenAI co-founder and former Tesla AI executive. He thinks in systems, not products. His public writing and talks explain how large language models actually work at a mechanical level, which matters if you want to stop treating AI as a black box and start making better decisions about where to apply it.

His prior educational venture, Eureka Labs, reflected a consistent interest in making deep technical knowledge accessible. His move to Anthropic in May 2026 puts him at the frontier of where the technology is going. He is worth following not for strategy advice but for grounding. Leaders who understand the mechanism make fewer bad bets.

Search for his public talks on neural networks and LLMs. They are long and technical by design, but an hour with one will permanently change how you read AI product announcements.

Andrew Ng

Founder, DeepLearning.AI

Ng built DeepLearning.AI to close the gap between frontier AI research and practical application. His writing is consistently focused on what practitioners and organizations actually need to know, not what makes for impressive conference slides.

His 2026 "AI Engineering Skills Map," co-presented with DeepLearning.AI, is a useful diagnostic for any leader trying to figure out what capabilities their team actually needs to build. It cuts through the noise about "AI strategy" and gets to the specific technical competencies that determine whether AI implementations work or fail.

Start at his writing page. It is organized by topic and skips the promotional material.

---

If You Want to Think About Competitive Strategy

Voices in this section are working on the harder question: what does AI mean for how businesses compete, not just how they operate?

Karim Lakhani

Dorothy & Michael Hintze Professor of Business Administration, Harvard Business School; Founding Chair, HBS AI Institute and Laboratory for Innovation Science at Harvard

Lakhani and his co-author Marco Iansiti wrote the strategy book that most serious executives were reading in the early AI era: Competing in the Age of AI (2019). Their central argument is that AI-native firms operate on fundamentally different logic than traditional firms, not just faster horses but a different kind of vehicle.

As Founding Chair of the HBS AI Institute, Lakhani's current work sits at the intersection of AI adoption and organizational design. If you are a leader in a legacy business trying to figure out what "AI transformation" actually requires structurally, his body of work is the most rigorous available.

Start with Competing in the Age of AI. It is several years old but holds up precisely because it is built on strategic logic rather than product announcements.

Marco Iansiti

David Sarnoff Professor of Business Administration, Harvard Business School; Head, Technology and Operations Management Unit and Digital Initiative at HBS

Iansiti's contribution to Competing in the Age of AI is the operations and platform side of the argument. His research on platform economics and digital transformation explains why AI-native businesses achieve cost and scale advantages that traditional operational models cannot match.

His ongoing work through the HBS Digital Initiative tracks how incumbents are (and are not) making the transition. He is less visible than some commentators on this list but more rigorous. When he publishes something, it is worth reading.

Ajay Agrawal

Geoffrey Taber Chair in Entrepreneurship and Innovation, Rotman School of Management, University of Toronto; Research Associate, National Bureau of Economic Research

Agrawal's contribution to the field is the economic framing. His core argument, developed across multiple books and now extended in The Economics of Transformative AI (University of Chicago Press, September 2026), is that AI is fundamentally a prediction technology and that understanding it through that lens explains which jobs, industries, and decisions it will change most.

For leaders who want to think clearly about AI marketing strategy and where AI creates genuine economic value rather than just operational convenience, Agrawal's framework is the most transferable to business decisions. He is a research associate at NBER, which means his work is grounded in economic methodology, not vendor enthusiasm.

Start with whatever is most recent from University of Chicago Press. The prediction-machine framing will change how you evaluate AI product claims.

Azeem Azhar

Founder, Exponential View; Non-Executive Director, Ada Lovelace Institute; Executive Fellow, Harvard Business School

Azhar's named framework, the "exponential gap," is useful for leaders who need to explain to boards and leadership teams why AI disruption feels both overhyped and underestimated at the same time. His argument: technology improves exponentially while institutions, regulations, and human habits adapt slowly and linearly. That gap is where most of the disruption happens.

His 2021 book, Exponential (published in the US as The Exponential Age), made this case before the current AI wave. Reading it now, the framework holds. His Exponential View newsletter applies the same lens to current events.

Azhar is also one of the cleaner writers in this space. He does not talk down to business readers or up to technologists. Start with a few issues of Exponential View before committing to the book.

Reid Hoffman

Co-founder, LinkedIn; Co-founder, Inflection AI; Co-founder and Executive Board Chair, Manas AI; Affiliated with the Berggruen Institute

Hoffman's contribution is not research. It is pattern recognition from operating at the frontier of platform businesses and AI ventures across multiple cycles. His 2025 book, Superagency, addresses the strategic and societal implications of AI more directly than his earlier work on blitzscaling.

His value is in the questions he asks, not the answers he provides. Masters of Scale consistently surfaces conversations about AI and competitive dynamics that are harder to find in academic literature. Hoffman is also one of the few people on this list who has actually funded, built, and operated AI companies, which gives his commentary a different texture than pure analysis.

Start with the Masters of Scale episodes focused on AI strategy. The format is conversational and works in transit.

Mustafa Suleyman

CEO and Executive Vice President, Microsoft AI; Co-founder, DeepMind; Co-founder, Inflection AI

Suleyman's 2023 book, The Coming Wave: Technology, Power, and the 21st Century's Greatest Dilemma (co-authored with Michael Bhaskar), is the most serious treatment available of the governance and containment problem in AI. His argument is that the technologies being built now are powerful enough to be genuinely destabilizing, and that society does not currently have adequate mechanisms to manage that.

That might sound abstract for a business leader. It is not. Leaders who understand why governments and regulators are moving the way they are on AI will make better strategic decisions than those who are surprised by each new policy development. Suleyman's background, co-founding DeepMind and leading Microsoft's consumer AI organization, means he is not a detached observer. He is describing a problem he is living inside.

Read The Coming Wave before the next election cycle makes it feel urgent rather than optional.

---

If You Want to Track What AI Is Actually Doing in Practice

Entries in this section are doing the work of observation rather than prediction. Useful for staying calibrated without drinking the Kool-Aid.

Simon Willison

Independent developer and creator of Datasette

Willison has no institutional affiliation and no corporate agenda. He built Datasette, an open-source tool for exploring and publishing data, and he writes daily about what large language models can and cannot actually do when you put them in front of real technical problems.

His TIL (Today I Learned) archive at til.simonwillison.net is a running log of practical experiments with LLMs, coding agents, model context protocol, sandboxing, and Python tooling. It is not written for executives. Read it anyway. The gap between what AI vendors claim and what Willison documents is instructive, and his observations tend to surface real capability shifts before they appear in the business press.

Start with his main blog. Filter for the LLMs tag and read a week's worth of posts. You will recalibrate your sense of where the technology actually is.

Benedict Evans

Independent technology analyst

Evans publishes twice-yearly major presentations on technology strategy. His Spring 2026 presentation, "AI eats the world," does what his presentations always do: find the structural shift underneath the noise and explain what it means for how industries work.

He operates with no employer and no institutional affiliation, which means his analysis does not carry the distorting gravity of a vendor relationship or an investment thesis he needs to defend. His writing is not warm or narrative. It is precise and occasionally blunt. That is the point.

All presentations are free on his site. Start with the most recent one.

Allie K. Miller

Founder and CEO, Open Machine

Miller founded Open Machine, an advisory firm focused on enterprise AI applications. Her public commentary focuses on the practical side of enterprise AI adoption: what organizations actually encounter when they try to implement this technology at scale, not what the pitch deck says they will encounter.

For leaders dealing with the organizational and implementation side of AI, her perspective is grounded in the messy reality of real deployments. Worth following if your current question is less "should we do this" and more "why isn't this working yet."

---

If You Want the Operator's Perspective

Both entries below are self-inclusions. The disclosure at the top of this article applies here.

Patrick Gilbert

Author, Never Always, Never Never; Founder, AdVenture Media

In Never Always, Never Never, Patrick Gilbert writes about the difference between adopting AI tools and actually building an AI-first culture. That distinction, which the book develops across chapters on the AI Double Helix and the AI Maturity Ladder, is the frame for most of the practical AI content on this site.

His AI-First Leader workshop is built from the same frameworks. The book is the starting point. If you want to understand how the ideas in that book connect to competitive strategy and marketing effectiveness, this site has you covered.

Isaac Rudansky

Co-founder, AdVenture Media

Rudansky is the practitioner's practitioner in this operation. As Patrick Gilbert describes in Never Always, Never Never, Rudansky's background as a working artist gives him a different relationship to AI creative tools than most digital marketers. He was producing serious creative work with Midjourney within minutes of first using it, because taste and visual understanding are inputs the tool cannot supply.

His value on AI training is in the practical specifics. He is not explaining AI in the abstract. He is showing what it takes to get useful output from real tools in real workflows. If your organization is trying to move people up the AI maturity ladder from Dabbler to Practitioner, his approach to training is worth examining.

---

How to Use This List

Do not follow all of them. Pick one or two based on your current question, not based on who has the largest platform.

For primarily strategic questions, start with Lakhani and Iansiti's Competing in the Age of AI, then add Agrawal's economics framing. For operational questions, Mollick's One Useful Thing and Willison's blog will do more for your thinking than any strategy book. For governance and policy risk, Suleyman is the only person on this list who has written a serious book on that specific problem.

A pattern in Never Always, Never Never holds here: AI gives expertise reach it never had before, but the expertise has to already exist. Everyone on this list is a source of expertise, not a substitute for building your own. Read them as inputs to your judgment, not as authorities to defer to.

For a longer treatment of what building an AI-first culture actually requires organizationally, the book covers that in detail. For the practical question of how to use AI in marketing without getting distracted by tools that add noise rather than value, start with the AI Double Helix framework before picking a vendor. And if you want to understand why AI rollouts fail in most organizations before they produce results, that post is worth twenty minutes of your time.

This list will tell you who is thinking clearly. What you do with that clarity is still on you.