About the client
Based in the UK, this industry leader is a global powerhouse in the research, development, and manufacturing of high-quality dietary supplements. With GMP and BRC-accredited facilities, they serve as a “one-stop-shop” for brands worldwide, handling everything from initial nutritional concepts to final sustainable packaging. Their mission “We Lead, We Create, We Innovate” is backed by a culture that thrives on technical excellence and rapid growth.
The Challenge:
Overcoming the “Prompting paradox”
As a high-volume contract manufacturer, the company’s formulation team delivers over 150 custom product packs every month. While they had already taken steps toward digitization using semi-automated AI tools, the process remained bottlenecked by human intervention.
The existing workflow relied on “manual prompting” a method where staff had to manually guide the AI, copy-paste data between systems, and constantly intervene to maintain consistency. This created several critical hurdles:
- Inconsistency: Variations in manual prompts led to inconsistent outputs across different sessions.
- Technical barriers: The system was difficult for non-technical staff to operate, limiting its use to a few specialized team members.
- Scalability: The manual nature of the workflow made it impossible to license the technology externally or use it in direct client-facing environments.
To maintain their competitive edge, the company needed to transition from a manual, prompt-based workflow to a fully automated, modular AI system.
The Solution:
A structured AI formulation engine
DataNorth partnered with the manufacturer to architect a Minimum Viable Product (MVP) that replaced manual intervention with an automated, node-based logic system. This new architecture ensures that the AI follows a strict, repeatable “Ruleset” aligned with the company’s proprietary codes and industry standards.
The automated transformation included:
- Task-node automation: Replacing manual “chatting” with AI with automated stages that handle data validation and input collection seamlessly.
- Deep data integration: The system was integrated with core reference files, including internal Supplement Codes and Consolidated Templates, ensuring 100% accuracy in formulation.
- Closed-loop feedback: A new interface allows team members to review and edit formulations within the system, creating a continuous improvement loop.
- Rapid deployment: To meet a critical deadline for a major global client demo on August 14, DataNorth deployed the environment to generate the three most vital sections of the formulation pack: Product Introduction, Market Research, and the Technical Formulation Table.
The Impact:
From tool to technology
By moving away from manual prompts and into a structured AI environment, the company has transformed a “semi-automated” task into a scalable business asset.
The new system is no longer just a tool for internal experts; it is a robust platform ready for future expansion and potential external licensing. Lead times have been stabilized, errors reduced, and the barrier to entry for staff has been lowered allowing the team to focus on what they do best: innovating the next generation of health products.
The efficiency gap:
From manual inputs to automated intelligence
Prior to the implementation, the formulation process relied on manual AI prompting, which created a 30% variance in output consistency and required frequent technical intervention. Since transitioning to DataNorth’s automated task-node system, the company has seen a 75% reduction in manual touchpoints per project. By replacing trial-and-error prompting with structured data integration, the workflow is now 100% accessible to non-technical staff, transforming a complex specialized task into a scalable, high-speed operation.
