Every SaaS business will become a harness around a model

Every software business must now build its own AI orchestration layer, or risk commoditization.

SaaS businesses are evolving into "harnesses" that wrap around AI models, according to Shrivu Shankar. This "harness" encompasses all the necessary infrastructure, interfaces, context, and state to operationalize stateless LLMs. Initially, individuals will use these harnesses to augment their work, with functions like engineering, product, and sales pairing with agents. As this progresses, core tasks will shift to background agents, with individuals orchestrating these agents and reviewing their outputs. Eventually, proactive background agents will manage many tasks, and human roles will focus on reviewing agent outputs and providing critical judgment. This means the company itself becomes the harness, with its domain knowledge, tooling, and review loops forming the core offering. Differentiation will stem from how effectively a company constructs its harness, influencing trust, distribution, efficacy, and domain context. The article suggests this trend is already visible in in-house AI developer tools, with companies needing to own the top-level harness and integrate vendor products into it.

The core argument is that the business value will shift from the proprietary model to the "harness" that effectively integrates and operationalizes AI models. Companies that master harness construction—defining how AI learns, what it monitors, how it interacts with human oversight, and its integrations—will build defensible moats. This means an increased focus on building custom infrastructure for AI orchestration, rather than relying solely on off-the-shelf AI model APIs. As a result, expect a rise in headless software components and a re-evaluation of organizational structures to best leverage human judgment within AI-driven workflows.

This transition signals a fundamental shift in how software products are developed and delivered, moving from human-centric processes to AI-centric ones where human intervention is strategically placed. Companies that can effectively build and manage these harnesses will gain competitive advantages in trust, distribution speed, and product efficacy. The article posits that AI-native startups are well-positioned to disrupt incumbents in areas where the competitive moat can be redefined and strengthened through superior harness design and execution.