When is a prediction reliable enough to use?
What information, and how much of it, do we need to use a prediction and place justified trust in it? I develop analysis and verification workflows to explore this question. Around 1,500 YouTube transcripts provide a basis for the Prediction Report, AI Bubble Tracker and Daily Checker.
From a client workflow to my own research
My work spans several research and reporting formats. For a private-equity firm, I developed a workshop and helped build an automated analysis workflow. It combines qualitative information from transcripts with quantitative inputs, evaluates incoming information and highlights relevant changes as early-warning signals. This helps decision-makers prepare for calls and follow up on promising leads. The client’s name and internal methods remain confidential. This work also informed the question behind Market Research & Reports: what information, and how much of it, is needed to use and trust a prediction, and what workflow is required to support that assessment?
My contribution
I design and develop agent-assisted analysis and verification workflows across several research and reporting formats. The focus is on bringing information together, identifying remaining uncertainty and critically assessing predictions before using them. The analysis and three report formats are in development; their reliability still needs to be tested.
Prediction Report
A structured assessment of a prediction’s evidence base, uncertainties and conditions for use.
AI Bubble Tracker
Tracking statements about the AI market and examining the expectations behind them.
Daily Checker
A daily review of statements and developments.
Want to review a report?
All three formats are in development. If you would like to read a report and provide feedback, email me for a look at the current work.