To meet the need, we:

  • Devised an AI-assisted editing process for maximum speed
  • Proposed using an existing client-approved instance of Anthropic’s Claude AI to maintain data security and address privacy concerns
  • Built a multi-stage, AI-powered article summarization engine, that was able to work within AI context-window limitations to distill the 7000-word post to 1,900 in minutes instead of hours
  • Refined and iterated on prompting style and sub-prompts to more faithfully reproduce key content and executive language
  • Employed skilled Steyer editing resources to evaluate output for quality and alignment with client voice, identifying important content gaps that slipped past the AI
  • Examined quality at each summarizer stage, including recognizing a sub-prompting issue introducing errors into the copy
  • Performed a detailed final review and copy edit confirming resolution of issues and adherence to client editorial standards

The outcome of this process was an effective summary chapter that preserved as much original language as possible, produced to exacting style and voice standards, comprehensively reviewed by skilled humans–all produced in roughly a third of the total hours that a human-only process would require. In the process, we demonstrated the importance of incrementally reviewing outputs at each step of multi-stage AI pipelines, to assure final quality and avoid wasting time addressing buggy content outputs that can multiply exponentially as they move through the pipeline.

By combining skilled project managers, editors, and writers with the latest in AI technologies, Steyer is able to rapidly create content solutions that meet business needs cost-effectively, without sacrificing final quality.

Interested in learning more about how we’re using AI and traditional methods to efficiently create quality content? Reach out today and let’s start the conversation!