Back to Foresight

Publications on Towards Finance

Every Foresight test is written up and published on Towards Finance.

Read the full articles there.

Test 1 Published Aug 2026

An AI Agent Outperforms Senior Financial Experts

Foresight, given the same closed data room as a 4-person Stern EMBA team, picked the same financial models on its own and scored 34% higher in blind evaluation — in about 1 hour instead of 30.

Test 2 Published Aug 2026

External Data Fetching Amplifies AI Agent's Outputs

Letting Foresight fetch a single checkable fact when the data room fell short won every blind comparison, scoring 57% higher on average — closing a real factual gap in minutes instead of hours.

Test 3 Published Aug 2026

Unrestricted Open Data Destroys AI Agent's Outputs

Removing the data room entirely and letting Foresight collect everything from the open internet dropped output quality by roughly 44% — though it still beat senior human experts by 7.5%.

Test 4 Published Oct 2026

Can Default AI Agents Run Financial Analysis Alone?

Given two sentences and no instructions, a default AI agent handled the same distressed-investment analysis and outscored Foresight by about 33% in blind evaluation — in 20 minutes instead of 90.