Exploring the Triple Frontier: Building Consumer, Enterprise, and Developer AI Products

The New Era of Building AI Products: Balancing ToC, ToB, and ToD

In a compelling conversation helmed by Sarah Guo, founder of the venture capital firm Conviction, industry veterans Kevin Weil and Mike Krieger provide insights into their journey from managing Instagram to shaping products at AI frontrunners like OpenAI and Anthropic.

Kevin Weil's Perspective

Kevin Weil, now deeply involved in OpenAI as Chief Product Officer, reveals his excitement about this transformative role. "It's the most interesting and impactful role I've taken," Weil emphasizes, acknowledging the unpredictable landscape of AI product challenges, where every few months a computer achieves something unprecedented.

Mike Krieger's Journey

Mike, similarly exhilarated by the shift from a consumer-centric Instagram environment to the more complex enterprise market, relishes the new rhythms of feedback cycles. "Enterprise feels more like an investment," he notes, pointing to the long cycles in enterprise deployment compared to consumer applications.

"Enterprise feedback is about the waiting game," Krieger reflects. "Once deployed, direct customer insights become a treasure trove for continuous improvement."

Exploring the Nuances of AI Product Development

The Interplay of Consumers, Enterprises, and Developers

Managing products across varied user bases concurrently requires a delicate balance and a shift in conventional product intuition. Weil notes this transition with humor: "I've meetings where they love the product but want 60 days' notice on any changes."

Intuitive Product Development
The spontaneity AI brings to the product design process shifts the reliance on intuition from certainty to probability of success. New model capabilities emerge mid-project—an entirely new kind of development cycle where the unknown looms large.

"It's like seeing a shape in the fog. You’re waiting to see what it can become," says Weil, highlighting the iterative, curious interplay between product and research teams.

Efficiencies and Effectiveness When Model Accuracy is Imperfect

Utilizing Suboptimal Models

Guo probes: "How do you plan when future functionality is uncertain?" The challenge lies in using models with varying degrees of accuracy. Weil and Krieger suggest focusing on creating robust design anticipation, with Weil remarking on GitHub Copilot as an illustration of investing in models that complement human endeavor despite being imperfect.

The Power of a 60% Solution

Models with 60% accuracy are not seen as failures but as opportunities to enhance cooperation between human input and AI capabilities. The idea is that AI, even flawed, can save critical time, particularly in complex, time-consuming tasks.

Krieger muses about the "magic 60%" as an arbitrary threshold illuminating the unpredictability of AI performance, much akin to the Mendoza Line in baseball-as a measure of performance tolerance.

AI Product Managers: Equipping Themselves for an AI Future

Imbuing Precision through Evaluation Standards

Assessing AI involves refined evaluation standards, a core skill for future product managers. As Krieger shares: "Our product managers excel by rewriting poor prompts into eloquent, functional standards," demonstrating the instrumental role of evaluation in AI development.

"It’s a core skill - writing evaluation standards is crucial for product managers," states Weil, emphasizing its importance in the AI ecosystem.

The Dynamics of the Future AI Product

The Vision of Proactivity and Asynchronicity

In envisioning future landscapes of AI, Krieger focuses on two key capabilities: proactivity and asynchronicity. AI that initiates action or provides recommendations based on user habits, alongside performing tasks asynchronously, marks a progressive direction in AI product development.

"Imagine AI proactively managing your schedule, or asynchronously researching," envisions Krieger, highlighting potential quality-of-life improvements.

Krieger and Weil discuss recognizing and leveraging 'power users’—those pioneering AI adoption within enterprises—as critical components in facilitating widespread AI acceptance.

Integrating Human-like AI Interaction

Weil continues, touching on recent developments in voice-based AI, "like a universal translator, providing real-time conversation in different languages—which will soon be common yet magical."

Closing Thoughts: A New Horizon for AI in Consumer, Enterprise, and Developer Spheres

With every step in the practical exploration of AI, from consumer to enterprise scales, the intersection of intuitions, predictions, and purely human elements shape the discourse. As Guo notes, engaging AI like TikTok users do—conversationally, interactively—offers a glimpse at the depth of change these models are facilitating in real-time.

"AI is becoming a part of our culture," reflects Guo. AI isn't just another tool; it’s a societal catalyst transforming how users, businesses, and creators interact.

Model Behavior and User Interactions

Understanding model personality—a product challenge for both AI developers and users—introduces a nuanced layer where users may even choose AI based on its perceived 'persona'. Experimental feedback loops, involving both anthropomorphic models and consumer reactions, highlight these emerging dynamics, emphasizing that AI interfaces increasingly mirror human social relationships.

As we look towards a future where AI is embedded in every facet of interaction, the dialog between industry leaders Weil, Krieger, and Guo reveals a horizon laden with opportunity, where AI helps redefine boundaries at the intersection of technology and human experience.

Midjourney prompt for the cover image: A diverse group of tech industry leaders discuss AI product development, surrounded by holographic models and futuristic screens displaying complex data. The setting is a high-tech conference room, viewed from above to capture the dynamic interaction and state-of-the-art technology. Unique details include digital waves flowing across surfaces, emphasizing the fusion of human insight and AI potential. Sketch Cartoon Style with a forward-thinking and collaborative mood.