Snowflake‑Powered AI Helps Edmund Optics Speed Up Technical Support

Manufacturing

Enterprise

Snowflake‑Powered AI Helps Edmund Optics Speed Up Technical Support

Technology used

Snowflake

Snowflake Cortex Search

Snowflake LLM Functions

Snowflake Native Apps

Matillion ETL

Streamlit (UI Framework)

RAG Architecture

Vector Database

Multi-Agent AI Architecture

Solution

Snap Analytics designed and delivered a Snowflake‑native AI assistant that consolidates Edmund Optics’ fragmented technical knowledge into a single, intelligent interface. The solution enables engineers to rapidly retrieve, synthesise, and respond to technical queries using AI‑powered search and automation.

Results

Reduced response times by ~40%, saved ~100 hours per month, and enabled faster, more scalable technical support through AI-powered knowledge automation.

The Challenge

The Solution

The Results

The Challenge

Edmund Optics’ global technical support team faced growing pressure to respond quickly and accurately to complex engineering queries.

Highly skilled engineers were spending significant time searching across disparate sources—internal databases, technical PDFs, website content, and historic support conversations—to compile responses.

This process created several challenges:

Slow response times and repetitive effort: As engineers manually searched fragmented sources and repeatedly solved similar queries in isolation.

Scaling limitations: As demand increased but adding more engineers was not viable

Inconsistent knowledge access and long onboarding cycles: With valuable insights locked in inboxes and historical interactions, and new engineers taking up to six months to become fully productive.

Snap had the expertise across our entire stack and helped us design a solution that truly gets the best out of the technology.

Daniel Adams

Global Analytics Manager

United Kingdom

The Solution

Snap Analytics partnered closely with Edmund Optics to design, build, and evolve an enterprise AI assistant through a phased, iterative approach.

Proof Of Concept To Production-Ready Solution

The project began with a rapid proof of concept, delivered in approximately 12 weeks, to validate the feasibility of an AI-driven technical support assistant. This evolved into a full production solution and later a second-generation product featuring enhanced capabilities such as multi-agent architecture and expanded language support.

Centralised, AI-Ready Knowledge Foundation

Using Matillion and Snowflake, Snap consolidated a wide range of structured and unstructured data sources into a unified reference layer, bringing together technical application documents and PDFs, historical support conversations, product specifications, knowledge base content, and internal product datasets spanning decades. This created a searchable, AI-ready knowledge base that eliminated duplication and enabled rapid information retrieval.

AI Architecture Built For Accuracy And Control

The assistant leveraged a Snowflake-native architecture with vector search and RAG capabilities to ensure responses were grounded in trusted internal data. Snap also implemented critical AI guardrails to prevent hallucinated outputs, control inappropriate or unsafe responses, and ensure engineers retained full oversight of customer communication.

Continuous Iteration and User-First Delivery

A lightweight, agile delivery model—supported by frequent feedback sessions, user interviews, and real-world testing—enabled rapid improvement across each phase, with key practices including direct collaboration between Snap engineers and client stakeholders, pilot testing with subsets of users, and continuous iteration based on real usage feedback.

Evolved Into A Multi-Agent AI Product

The second-generation solution introduced a range of enhanced capabilities, including a multi-agent AI architecture, multi-language support for global teams, and improved reasoning capabilities such as product recommendations, alongside increased scalability and stability within Snowflake.

The Results

~40% Reduction In Response Times

Following initial rollout (Version one estimate).

~100 Hours Saved Per Month

Through reduced manual research effort.

200–300 Customer Queries Per Month

Partially handled or accelerated by the assistant.

Faster Onboarding of New Engineers

With anecdotal reduction from ~6 months to ~3–4 months.

Improved Productivity, Increased Consistency And Reuse Of Knowledge

Allowing experts to focus on complex, high-value queries and reducing duplicated effort across teams.

Centralised Access To Decades Of Technical Knowledge

Improving accuracy and confidence in responses.

Foundation Established For Future AI Use Cases

Proving the viability of internal AI-driven solutions.

Edmund Optics is a global manufacturer and supplier of optical components, imaging systems, and photonics solutions, with 1,300+ employees across 19 global locations. The business serves advanced manufacturing, life sciences, biomedical, semiconductor, and R&D markets, supporting mission-critical applications where precision, performance, and technical expertise are essential.

Next case study

Transforming Trade Spend into Measurable ROI

Transforming Cost Forecasting for an Enterprise CPG Through Intelligent Data Automation

Navigating Regulatory Changes: SJP Ongoing Advice Charges