AI Applications and Vertical Integration

2024-09-30 · Source: Tanay’s Newsletter · Field: Business & Management — Corporate Strategy & Leadership, Entrepreneurship & Start-ups, Artificial Intelligence & Machine Learning · Depth: Intermediate, short

Summary

AI application companies are increasingly adopting a "full-stack" vertical integration strategy, moving beyond traditional middle-layer application development. This trend manifests in two primary directions. Some companies integrate "down" into the model layer, developing proprietary intelligence. Examples include Cursor's Composer 2, which uses Kimi K2.5 as a base and applies continued pretraining and reinforcement learning for coding tasks, and Intercom's Fin Apex, now powering its English-language customer conversations. This integration is driven by the value of proprietary usage traces for model improvement, reduced costs, faster speeds, and product differentiation. Conversely, other companies integrate "up" into the human or service layer, selling complete outcomes rather than just software. This "services-as-software" model, exemplified by Crosby AI's "Neofirm" approach, WithCoverage, Harper, and Mechanical Orchard, becomes more viable as AI improves agentic task capabilities, allowing companies to own the end-to-end process and fill AI's remaining gaps.

Key takeaway

For AI Product Managers evaluating long-term strategy, recognize that AI application companies are converging on vertical integration. Your team should assess whether to integrate "down" by developing proprietary models using unique usage data for cost, speed, and differentiation, or "up" by offering full outcomes as a service. This strategic choice impacts competitive positioning and value capture, demanding a clear roadmap for expanding your stack.

Key insights

AI application companies are vertically integrating into either the model or service layer for full-stack operations.

Principles

• Proprietary usage data drives AI model performance.
• Vertical integration boosts differentiation and efficiency.
• AI enables selling outcomes, not just software.

In practice

• Fine-tune models using proprietary usage traces.
• Combine AI with human services for end-to-end outcomes.

Topics

• AI Vertical Integration
• AI Application Strategy
• Proprietary AI Models
• AI-native Services
• Model Fine-tuning
• Full-stack Startups

Best for: Entrepreneur, CTO, VP of Engineering/Data, Director of AI/ML, AI Product Manager, Investor

Related on AIssential

• Microsoft’s AI chief on the greatest game of catchup ever played — Microsoft is aggressively building proprietary AI models, chips, and integrated software to achieve cost leadership and deep product integration, aiming to surpass competitors.
• Choosing between in-house development and AI integration consulting for your next project — Choosing between in-house AI development and external AI integration consulting impacts cost, speed, risk, and long-term strategy.
• Software Finally Gets to Work: The Opportunity in Vertical AI — Vertical AI transforms software from assisting to executing, targeting labor costs and building compounding value through deep domain integration.
• The Updated Map of AI — AI industry leaders are vertically integrating across infrastructure, models, and distribution, consolidating power.
• AI's Great Divergence — AI's rapid advancement is creating significant divergences in public perception, economic benefits, and workforce impact.
• The Era of Vertical AI Models — Domain-specific AI models, enhanced by post-training on last-mile usage data, can now outperform general-purpose frontier models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Tanay’s Newsletter.