How to Use AI More Ambitionally: Height, Depth, Breadth & Verification

One: Height — From Task Completion to Goal Clarity

Felix Rieseberg argues that most users still treat AI as a faster executor of existing workflows. He illustrates with a quarterly business review: instead of asking AI to format a PowerPoint, ask it to analyze the underlying business data, identify problems, and propose decisions. Going further, if AI can continuously monitor metrics, the quarterly review cadence itself may become obsolete. The key shift is to ask: "What do I really want to happen next? Is there a deeper goal behind the current target?"

Two: Depth — From Role-Playing to Professional Methods

Adding "you are a 20-year expert" to prompts only changes vocabulary, not methodology. Rieseberg demonstrates with a user-churn scenario: rather than accepting the team's hypothesis that poor onboarding causes churn, feed raw behavioral data (registration timestamps, feature usage, login frequency, churn dates) to AI and require it to use statistical tools and databases to test the hypothesis. If churn clusters in the first days, onboarding is suspect; if it appears months later, the cause lies elsewhere. This forces AI to validate assumptions instead of rubber-stamping them.

Three: Breadth — From Single Lens to Cross-Disciplinary Synthesis

Complex problems benefit from multiple professional perspectives. For the same churn problem, Rieseberg sequences four methods: (1) data analyst — segment retention by cohort and time to pinpoint drop-off points; (2) investigative journalist — design interview guides from data patterns, analyze support tickets, compare feedback to uncover why users leave; (3) product designer — map the failing user journey, identify confusing steps, propose testable UI changes; (4) short-seller researcher — stress-test conclusions: are we ignoring high-value user loss? Do metric definitions hide deterioration? He emphasizes: do not ask AI to role-play these experts; ask it to execute their actual analytical methods.

Four: Verification — From Trusting Answers to Testing Results

Ambitious use demands rigorous validation. If AI builds a churn-prediction model, evaluate it on hold-out data not used in training. If AI writes code, run the tests. If AI produces a business report, audit data sources and calculations. Rieseberg warns against letting models self-assess; the user must define success criteria upfront and verify independently.

Source: Felix Rieseberg, "How to be more ambitious: A tutorial for humans with models" — https://felixrieseberg.com/how-to-be-more-ambitious-a-tutorial-for-humans-with-models/