From search migration to enterprise knowledge platform
Our client, a global software and services provider, needed a search foundation that could keep up with a growing number of internal and customer-facing applications. What started as a migration of two applications from IBM Watson to Lucidworks Fusion has grown over about four years into a shared platform that serves many of the company’s search-heavy applications.
Project Overview
New Elevation extended Fusion with a custom retrieval-augmented generation (RAG) framework, a built-in evaluation pipeline, and API access for AI agents. Employees use the platform through tools such as an internal search portal and a document management system. Customers and vendors use it through incident management applications that answer questions from company content and link to the sources.
The Challenge
The original scope was focused: move two applications from Watson to Fusion to gain more control over relevance and a clearer path for growth. Once those applications went live, other teams asked for the same capability, and each came with different needs:
Fast, precise keyword search across structured and semi-structured content
Natural-language answers drawn from tickets, manuals, and wiki pages
Strict access control for sensitive content
Support for users across regions and languages
A single fixed configuration couldn’t serve all of these needs, and a separate stack for each application would have been costly and difficult to govern. The client needed one shared platform that could be configured for each application while keeping quality, security, and onboarding speed consistent.
Our Approach
New Elevation built the platform in layers on top of Fusion, so each application could get the behavior it needed from a common foundation.
1. An app-based foundation in Fusion Lucidworks Fusion is a collection of apps hosted in the same environment.
Each app has its own configuration, pipelines, search behavior, data, and resources, and apps can run independently or share information when a use case calls for it. This structure let the team configure and tune each application’s search experience on its own while every application benefited from the same underlying platform and shared practices.
2. A layered RAG framework Custom layers sit on top of that foundation.
Each layer handles one part of the path from question to answer and can be tuned per app, so a keyword-focused employee portal and a RAG-driven incident assistant can run on the same platform and behave very differently. Applications that need traditional search can bypass the RAG layer entirely.
The layers cover three areas:
Search and retrieval: Queries are analyzed and rewritten to capture intent, with recognition of synonyms and business entities, so a search for one term can find content that uses another. Multilingual search works across languages, and each application can choose lexical, semantic, or hybrid retrieval and set its own ranking and personalization rules.
Data and knowledge: Connectors ingest content from databases (JDBC), websites, REST APIs, SharePoint, and Amazon S3. Built-in and custom parsing handles documents, transcriptions, tables, wiki pages, and tickets. Content is split into meaningful passages and enriched with metadata, so the system can retrieve the exact passage that answers a question. Results, including generated answers, honor each user’s access rights.
Platform and operations: Observability and traceability let teams follow a request from query to answer and see what was retrieved and why. Analytics and ranking explanations support continuous improvement. High availability and disaster recovery keep search dependable. Compliance, audit, retention, and data residency controls support regulated and global use cases.
3. Built-in quality evaluation
An evaluation pipeline continuously checks whether the right content was retrieved and whether each generated answer is accurate, relevant, and supported by that content. Teams use it to test changes before release, apply the same quality standard to every application, and validate new applications during onboarding. It is a key reason the platform could keep adding applications without losing quality.
4. Ready for AI agents
Any upstream AI agent can connect to the platform through its API, which gives the agent access to enterprise content that respects user permissions and meets the same retrieval and evaluation standards as the human-facing applications. A Model Context Protocol (MCP) integration is in progress. MCP is an open standard that lets AI agents discover and use tools consistently, so the integration will make it easier for other agents to connect.
The Results
Today the platform serves employees, customers, and vendors across many of the client’s search-heavy applications, all from a single environment. Each application gets the search or RAG behavior it needs without a separate stack to build and govern. Reusable layers and the evaluation pipeline let new applications go live quickly at a consistent quality bar, and AI agents can already query the platform through its API.
Technologies Used
More Case Studies
Related Work
Taking Client Setup Out of the CRM
Our customer’s business relies on a distributed sales team to identify corporate clients that align with their sales goals. But setting up a new client required sales professionals to navigate a back-office CRM, understand its data structures and relationships, and enter information correctly. Because the system was used infrequently, records were often late, inaccurate, or duplicated. New Elevation simplified the process by moving it out of the CRM entirely. A search-led experience in the firm’s company-wide portal helps users find existing records before creating new ones, making it easier for sales teams to work efficiently while improving data quality.
View Case Study →
AI-Enabled Website Revamp
A three-person team led a full rebuild of the New Elevation website using AI and MCP tooling. The effort reimagined how internal digital platforms can be delivered, using AI to significantly reduce team size, timelines, and reliance on traditional role structures.
View Case Study →
A Complete League Management System Without Enterprise Overhead
A small business that runs Mahjong leagues and an online storefront needed a custom application to manage its operations without adding significant cost. New Elevation helped turn its vision into a production-ready MVP in about three weeks, enabling it to streamline league management and create a foundation for a new product offering.
View Case Study →
Optimizing Adobe Experience Manager for IT client
A fortune 500 company needed clarity on whether their existing AEM implementation was optimized and whether a cloud migration made sense. New Elevation conducted a comprehensive audit — spanning templates, workflows, and stakeholder interviews — and delivered actionable recommendations that gave the client the confidence to make an informed migration decision.
View Case Study →