The AI-Accelerated Decision Quality Model
Use AI to kill bad ideas fast, deepen good ones, and fact-check your own biases before committing resources
The AI-Accelerated Decision Quality Model is Sahil Lavingia's practical framework for using AI to transform the quality and speed of decisions across business, investing, and life. The core insight is not that AI replaces thinking but that it compresses the research cycle so dramatically that the equivalent of weeks or months of investigation can happen in a single subway ride. This compression has three primary effects. First, it kills bad ideas fast: most ideas people carry on mental to-do lists for months are actually terrible, but because the research required to discover this was prohibitive, people deluded themselves into thinking the ideas were promising. Now you can evaluate an idea at depth in minutes and either discard it or commit to it with genuine conviction. Second, it deepens good ideas by enabling research at a level previously reserved for trillion-dollar companies: every material sourced, every supplier evaluated, every historical precedent examined. Third, it fact-checks your cognitive biases by forcing you to articulate mental models and test whether they generalize beyond the single example you always use. Lavingia emphasizes that AI is most useful for new explorations rather than existing work where the context in your head is too expensive to transfer into a prompt. He also stresses that AI removes the emotional component from decisions by providing an objective sounding board that does not have political motivations or emotional attachments to particular outcomes, unlike the committees of people who typically influence major business decisions.
- Most ideas on your to-do list are terrible and AI lets you discover this in minutes rather than months
- The quality of every decision improves when you can do the equivalent of a hundred hours of research instantly
- AI is most useful for new explorations not existing work where context transfer is too expensive
- Emotional bias in decision-making can be reduced by using AI as an objective sounding board
- If AI cannot generate additional examples for your mental model the model probably does not generalize
- Eliminate Your To-Do List Through Rapid Idea ValidationWhen you have a new idea, immediately invest the equivalent of deep research time through AI rather than adding it to a to-do list. Ask ChatGPT to explore the idea at depth: what are the biggest holes, what precedents exist, what does the competitive landscape look like, what are the key risks. Most ideas that seemed promising in the abstract reveal themselves as weak when examined concretely. The goal is to get to zero or to genuine conviction quickly rather than carrying a backlog of untested ideas that creates the illusion of opportunity while consuming mental bandwidth.Pro tipTell the AI to poke holes in your idea and explicitly ask for the strongest counterarguments. If it can only come up with the single example you already had in mind, the idea probably does not generalize and should be discarded.
- Deepen Research on Surviving Ideas to Expert LevelFor ideas that survive initial validation, use AI to achieve a level of research depth previously available only to large organizations with dedicated analysts. Ask specific questions about history, precedents, materials, suppliers, zoning, regulations, competitive dynamics, and technical requirements. The goal is to arrive at your first meeting with a domain expert already knowing enough to have a substantive conversation rather than starting from zero. When you show up with deep background knowledge, experts share higher-quality insights because they skip the basics and address nuanced issues they would not have raised with a novice.Pro tipFrame your research questions around your specific thesis rather than generic exploration. Instead of asking about downtown Brooklyn broadly, ask what the first roads were and whether they still carry the most traffic, because your investment thesis is based on network effects.
- Fact-Check Your Own Biases and Mental ModelsUse AI to test whether your mental models actually generalize beyond the single example you always use to explain them. Share your model and ask for five additional examples. If the AI can only generate the one example you already had, the model is weaker than you thought and should be revised or discarded. This practice combats the common pattern where people make decisions based on a rationalized narrative and then selectively gather evidence to support the conclusion they already wanted to reach. AI provides a more objective assessment of whether your reasoning actually holds.Pro tipAfter making a tentative decision, explicitly ask AI whether your stated reasons actually align with the decision. It will often reveal that your rational justification does not match your emotional preference, giving you the chance to make a more honest choice.WarningAI safety guardrails may sometimes push back on decisions that are perfectly valid. Recognize when the AI is genuinely identifying a flaw versus when it is being overly cautious due to its training.
- Empower Every Team Member as a Decision MakerUse AI to distribute expertise across your organization so that every person can make better decisions independently rather than requiring specialized knowledge holders for every question. When a designer needs to understand equity structures, when an engineer needs to grasp pricing psychology, when a marketer needs to evaluate technical feasibility, AI provides instant expert-level context that previously required scheduling meetings with internal specialists. This reduces the number of people involved in each decision, which counterintuitively improves decision quality because committees introduce emotional dynamics and political considerations that dilute clear thinking.
Lavingia used ChatGPT to research buying a building in New York City, starting with the history of downtown Brooklyn's road formations to identify network effects, narrowing from a thousand potential buildings to specific candidates, researching construction materials of specific buildings, visualizing interior designs through DALL-E, and evaluating zoning requirements. He could show up to meetings with real estate developers, construction managers, and general contractors with a depth of knowledge that normally takes years to accumulate, leading them to remark on his surprising architectural knowledge.
Lavingia estimates that Gumroad's eventual shift to 10% flat pricing, which took twelve years to implement, cost approximately seventy million dollars in unrealized profit margin. The decision itself was not complex; it was a single number change in one line of code. But the research and conviction required to make the change was prohibitive without AI-assisted analysis of pricing precedents, competitive dynamics, and customer impact scenarios.
Lavingia developed this approach through daily use of ChatGPT across every aspect of his life as both the CEO of Gumroad, one of the largest platforms for creators to sell their work online, and as an active investor evaluating real estate and startup opportunities. His most vivid demonstration involves buying a building in New York City, where he used ChatGPT to research the history of downtown Brooklyn, identify network effects in real estate going back to original road formations, evaluate zoning requirements, background-check property owners, visualize interior designs, and narrow from a thousand potential buildings to serious candidates. He estimates that a single pricing decision at Gumroad, which took twelve years to optimize, cost approximately seventy million dollars in profit margin that could have been captured earlier with better research tools. This concrete financial cost of slow decision-making became his motivation for integrating AI into every decision process.