Empowering Global Journalism with GenAI
The Associated Press & Pluralit’s GenAI Collaboration
In early 2024, The Associated Press (AP) partnered with Pluralit AI Solutions to explore how Generative AI could enhance editorial workflows across languages, formats, and audiences. The goal was to accelerate production, support editorial creativity, and expand multilingual reach — without compromising on quality or journalistic integrity.
​
Together, we designed and deployed three end-to-end GenAI solutions, moving from PoC through Pilot and Pre-Production. Each solution was co-developed with the client’s editorial, product, and engineering teams using weekly sprints, user-driven feedback, and strict quality benchmarks.
We developed an AI-powered summarisation tool to support editors in producing accurate, concise summaries faster.
​
This required overcoming challenges such as hallucination control, temporal language handling, and consistent tone.
​
Key components included:
​
-
Prompt tuning and LLM benchmarking across several models.
-
Guardrails using NER-weighted chunking to detect hallucinations.
-
Tools for measuring editing distance and user effort.
-
Fine-tuning strategies to better align with editorial needs.
​
The result: faster summary creation without losing editorial oversight.
1. AI-Powered News Summaries
To expand the organisation’s global reach, we created GenAI workflows for Spanish and Portuguese translation, with editorial-level quality.
​
Highlights:
​
-
Competitive testing across multiple LLMs (GPT-4o, Claude, DeepL, etc.)
-
“Write-through” logic to manage evolving news articles without retriggering full re-translation
-
Real-time metrics on translation quality, API performance, and editor satisfaction
-
A lightweight editor UI integrated with internal systems
​​
This enabled faster, more accurate multilingual publication — at scale.
2. High-Quality Machine Translation
3. Multi-Style Headline Generation
We also introduced a headline generation tool to support fast, creative ideation across multiple tones and formats.
​
Delivered features included:
​
-
Multi-style output using LangGraph nodes
-
Character-length precision to meet editorial standards
-
Front-end tool for browsing and adapting headline ideas
-
Seamless integration into existing editorial platforms
​​
Rather than replace editors, this tool helped spark better ideas faster.
RESULTS
All three use cases moved successfully through PoC, Pilot, and Pre-Production, demonstrating the real-world readiness of GenAI in complex editorial environments.
​
Workflows were accelerated. Output expanded. And editorial teams gained intelligent, intuitive tools that respected their standards.
BUILT WITH EDITORS, POWERED BY AI.
This between AP and Pluralit AI Solutions shows what GenAI can do when designed for real people solving real problems.
With the right partnership and technical depth, GenAI becomes more than a tool — it becomes part of the newsroom.