Unsupervised Learning• Aug 26, 2025• 1:04:26Interview

Scale Expert on AI Pricing Tactics, Reactions to GPT-5 & Why Voice is Underrated

From Unsupervised Learning

Peter Dang•Investor, Felicis

Executive Summary

AI product pricing is flawed; models must evolve from per-seat or token-based systems to better align with value creation and outcomes.
Startups have a significant advantage over large model companies by focusing on niche, vertical markets with proprietary data, as incumbents target trillion-dollar opportunities.
The future of AI product development hinges on proactive agents that initiate action and the critical, underhyped role of 'evals' (evaluations) as the new form of product specification.
Defensibility for AI startups is built on three pillars: unique proprietary data, a data flywheel from product usage, and a world-class, focused team.

9 quotes

Concerns Raised

Current AI pricing models (per-seat, token-based) do not align with customer value.
The AI 'talent wars' are creating a widening compensation gap between research and other essential roles.
The term 'agent' is currently overhyped, as today's systems lack true agency and proactivity.

Opportunities Identified

Startups can build defensible businesses on proprietary data not found on the open web.
Developing proactive AI that initiates tasks represents a key frontier for product innovation.
Applying AI to optimize inefficient processes in 'boring' industries like call centers presents a massive opportunity.
Mastering 'evals' is an underhyped but critical competency for building successful AI products.

Key Themes

The Evolution of AI Pricing Models

Current AI pricing, whether per-seat or token-based, fails to capture the true value delivered. The industry needs to move towards models that are more closely aligned with the work performed and outcomes generated, even if pure outcomes-based pricing is challenging for generalist tools like ChatGPT.

This is a fundamental business model challenge for the entire AI industry, impacting profitability, customer adoption, and the ability to scale sustainably.

Startup vs. Incumbent Dynamics

While foundation model companies have advantages like early access to new models, startups can carve out defensible niches. By focusing 100% on specific verticals and leveraging proprietary data not available on the open web, startups can build valuable businesses in markets too small for incumbents to pursue.

This provides a strategic roadmap for founders and investors on how to compete and build lasting value in an ecosystem dominated by large technology players.

The Future of AI Product and Interaction

The paradigm of user interaction with software is shifting. The future lies in proactive AI that can initiate conversations and actions, moving beyond the current reactive, prompt-based model. This will likely lead to a decline in the importance of traditional graphical user interfaces for many applications.

Product leaders and designers must rethink user experience from the ground up, preparing for a world where the AI, not the user, often initiates the interaction.

The New AI Product Development Paradigm

Building AI products requires a different approach than traditional software. The 'eval' (evaluation) has become the new product specification, defining success and guiding development. This data-centric approach is a core, yet underhyped, aspect of creating effective AI systems.

This insight highlights a critical operational shift for teams building with AI, emphasizing the need for new skills and processes centered on model evaluation and performance measurement.

Get started free

Topics

AI Pricing Models Value-Based Pricing Subscription Models Startup Strategy AI Defensibility Proprietary Data Data Flywheel Foundation Models Product Management AI Product Development Model Evaluation (Evals)Proactive AI AI Agents User Interface (UI)OpenAI ChatGPT Talent Wars Organizational Culture

Processed Apr 3, 2026 yt-dlp + mlx-whisper + Gemini