Wilkinson argues that building with AI is a purer form of creation. It removes the friction, misinterpretation, and compromise inherent in managing a team, allowing a creator's vision to be translated directly into a final product without dilution.
The barrier to creating software is collapsing due to AI. This will lead to intense competition and pricing pressure, similar to the restaurant industry, which is a difficult business not because of a lack of demand, but because of an oversupply of competitors.
Wilkinson built an automation that scans his newsletters and Readwise account, selects stories based on his interests, and generates a daily audio brief using a synthetic voice. This creates a hyper-personalized media diet, filtering out irrelevant or negative news.
Andrew Wilkinson uses a vector database trained on all his company data to query complex operational questions. This allows him, as the head of a conglomerate, to instantly spot trends, issues, and anomalies across multiple businesses—a task impossible for a human to do alone.
Andrew Wilkinson reveals the hidden cost of using AI agents for automation. He spends the majority of his time debugging and improving them, with only a small fraction dedicated to actual productive output. This highlights the immaturity of current agent technology despite its power.
Wilkinson’s AI agents triage his emails, identify high-priority items, and draft several potential responses in his voice. He simply replies with "1B" or "2C" to execute complex actions, transforming his management workflow into a simple series of multiple-choice questions.
Wilkinson built "Deep Personality," a SaaS app, and automates its operations using AI agents. These agents handle customer support tickets (even fixing bugs and deploying code), manage ad campaigns on Meta and Reddit, and assist with development, showcasing a new model for lean startups.
Instead of trying to craft a perfect, detailed prompt, Andrew Wilkinson tells the AI his high-level goal and instructs it to "ask me a shitload of questions to determine your prompt." The AI then conducts a 5-10 minute interview, gathering all necessary context to produce a superior result.
Wilkinson’s CFO, with no prior coding experience, used AI tools to build a sophisticated, customized portfolio management dashboard. This replaced Adapar, a service costing up to $100k annually, demonstrating how AI empowers non-engineers to build complex internal tools and disrupt expensive enterprise software.