ProVentures Research Brief

AI is making execution and discovery easier. This edition looks at what still takes work: measuring founder progress, earning buyer confidence, and understanding who bears the risk behind AI products.

Measuring the Impact of Impact Measurement

Emanuele Colonnelli, Timothy McQuade, Natalia Rigol and Benjamin N. Roth · Becker Friedman Institute working paper · November 17, 2026

A randomized trial with 163 Indian social enterprises compared a common training course with the same course plus hands-on implementation, stakeholder research and a tailored report. Firms receiving that extra support recorded roughly twice the revenue growth and raised roughly twice as much external capital. A separate investor experiment found that quantified customer-impact metrics increased interest; generic impact statements did not. The trial tested a bundle, so it cannot isolate the effect of any one part.

For ProVentures: Make advisor engagements produce something founders can use: a baseline, an agreed metric, stakeholder feedback and a documented decision.

Newsletter angle: The Advice Wasn’t the Product. The Evidence System Was.

Read the working paper

B2B Content Engagement & Buyer Behavior Report

Informa TechTarget · November 16, 2026

In a survey of 499 technology buyers, 44% reported using AI during active evaluation, versus 33% during discovery and 20% during decision-making. Newsletters were the most cited discovery channel (46%). And while 58% often clicked through when AI cited a vendor, only 42% said a citation built trust. The respondents were drawn from TechTarget’s own opt-in audience.

For ProVentures: Use the newsletter to introduce named experts, then give buying groups the evidence they can take into an internal meeting: costs, implementation requirements and measurable outcomes.

Newsletter angle: AI Gets You Clicked. Experts Get You Chosen.

Read the report · Read the public findings

Clouds of Credit: Mapping the Financial Web Behind AI Compute

Stijn Van Nieuwerburgh and Zijin Su · NBER working paper · November 2026; first draft November 1

Debt finances nearly 60% of the compute expansion in the authors’ model. In its low-productivity scenario, losses reach $16.8 trillion across shareholders and creditors, with a 40% loss given default on debt. Removing contractual spillovers between key firms reduces modeled private losses by roughly 30%. These are stress-test results, not forecasts.

For ProVentures: Ask AI founders which suppliers their margins depend on, what happens if API prices rise, and whether a substitute could be deployed if a supplier became unavailable.

Newsletter angle: Every AI Startup Has a Hidden Balance Sheet.

Read the working paper

AI as Shared Business Infrastructure: Firm Growth and the Distribution of Gains

Panle Jia Barwick, Lu Fang, Jinhan Liu, Huilan Xu and Zhe Yuan · NBER working paper · November 2026

After a Chinese e-commerce platform embedded AI into merchant software, the authors estimated 51.6% higher monthly sales and 47% more orders among adopting firms with annual sales below RMB 5 million. The smallest firms gained most in percentage terms. Category-level evidence pointed toward more entry and new products, not just sales shifting between merchants. Results come from one platform and should not be treated as a universal benchmark.

For ProVentures: Give smaller marketplace participants access to shared research, listing and measurement tools. Let their own expertise and judgment distinguish their offers.

Newsletter angle: The Best Marketplace AI May Be Infrastructure, Not a Competitor.

Read the working paper

Limited Capital? Explaining Transatlantic Divergences in Artificial Intelligence Startup Financing

Christian Hermann Hennings and Dirk Schiereck · PLOS One · November 5, 2026

Across roughly 8,000 PitchBook-tracked rounds, European AI deal counts broadly kept pace with the U.S., but financing volumes diverged. In the authors’ later observation period, AI represented about 60% of U.S. venture funding by value versus about 31% in Europe. Deal counts are an imperfect proxy for founder formation; the result also depends on PitchBook’s AI classification.

For ProVentures: Match founders to investors by geography, stage, check size and follow-on capacity. A long contact list is no substitute for knowing who can finance the next stage.

Newsletter angle: Europe May Not Have an AI-Founder Problem. It Has a Scale-Capital Problem.

Read the peer-reviewed paper

Sources for this edition

  1. Founder and advisor programs: Becker Friedman Institute publication record and full working paper and methods.
  2. Demand generation: Informa TechTarget report and findings and survey methodology.
  3. AI infrastructure financing: NBER Working Paper 36032.
  4. Marketplace design: NBER Working Paper 36003.
  5. Transatlantic AI financing: PLOS One peer-reviewed article and methods.

The “For ProVentures” notes and newsletter angles are editorial interpretations, not findings from the sources.