Saved 200+ weekly hours at Cooke Chile

The Challenge

Cooke Aquaculture manages large volumes of operational data across multiple Excel files, covering production metrics, compliance reports, anomalies, and performance indicators across sites and zones.

While this data was available, extracting insights required manual analysis, cross-checking spreadsheets, and relying on technical teams to answer operational questions. This slowed decision-making and created friction between data, operations, and leadership.

Cooke needed a way for teams to ask questions in natural language and get accurate, structured answers directly from their Excel data, without having to open or understand the spreadsheets themselves.

The Solution

EarlyShift designed and implemented a custom AI chatbot platform capable of understanding, searching, and reasoning over tabular data sourced from Excel files.

Instead of treating spreadsheets as static documents, we built a system that:

• Ingests Excel files
• Converts them into a structured, tabular JSON format
• Preserves rows, columns, headers, and relationships
• Makes the data queryable via Natural Language Queries (NLQ)

This allowed the AI to behave less like a generic chatbot and more like an interactive data analyst.

Each chatbot was custom-configured for Cooke’s internal use cases, ensuring relevance, accuracy, and secure access to operational information.

How It Works (High-Level Architecture)

1. Excel Processing PipelineUploaded Excel files are parsed and transformed into normalized, structured JSON tables that maintain the integrity of the original data.
2. Tabular Data Intelligence LayerThe AI is trained to understand column semantics, numeric values, categories, and time-based trends, enabling precise reasoning over structured data.
3. Natural Language Query InterfaceUsers ask questions like:
4. Structured AI ResponsesThe chatbot returns concise, structured answers, often including summaries, comparisons, and trends—without exposing raw spreadsheets.

The Result

• Instant access to insights from complex Excel datasets
• No technical expertise required to query operational data
• Faster decision-making across operations and management
• Reduced dependency on manual reporting and data extraction
• A scalable foundation for future AI-driven analytics

Most importantly, Cooke moved from reading spreadsheets to having conversations with their data.

Why It Matters

This project demonstrates how AI becomes truly valuable when it understands structure, not just text.

By transforming Excel files into a format the AI can reason over, EarlyShift enabled:

• Real operational intelligence
• Secure, internal AI tooling
• A reusable, multi-bot platform architecture

This was not a generic chatbot, it was a custom AI data interface built around Cooke’s real operational workflows.