Project
Coltri Knowledge System
An ongoing project to transform Coltri documentation into a single, searchable and evolving knowledge system.
Client
Coltri Compressors
Period
In progress · 2026
Category
Integrated systems · UX/UI design
From documentation to knowledge
Before building the portal, we designed the system that must feed it.
CKS began with a functional analysis workshop conducted with Coltri to map processes, pain points, users and responsibilities across the existing document ecosystem. The goal was not to draw screens immediately, but to understand how to turn knowledge distributed across SharePoint, archives and manual procedures into one coherent system.
The analysis defined a first functional architecture, data model, publication flows, search modes and an evolutionary roadmap. The project is now underway: how much we tell represents the perimeter designed and the direction we are working on, not a product already finished.
State
Project in progress
Method
Workshop + functional analysis
System
SharePoint, Next.js, Knowledge Layer
Development
Advanced search + AI assistant
From fragmentation to a shared map.
Catalogs, technical cards, manuals, spare parts, certificates and videos constitute a vast, multilingual and subject to revisions. Before designing CKS we have reconstructed where these contents live, who updates them, how they are published and where duplicates or disalignments arise.
Functional analysis was not an annex to the project: it is the project in its first form. It has transformed needs and criticality into entities, metadata, roles, permits, operational flows and criteria with which to evaluate the successive technological choices.
One point of truth. More services that use it.
SharePoint has been identified as the official source of documentation. A sync layer will have to read new content and revisions, check mandatory metadata, normalize information and update a database optimized for consultation.
Knowledge Layer will organize relationships, languages, categories, models, visibility and revisions. Over this base they can live a Next.js portal, a more effective search and AI assistant, without duplicating documents or forcing departments to upgrade multiple platforms.
Search designed around how technicians, dealers and customers actually work.
Not everyone knows the internal structure of an archive. There are those who start from the compressor model, who from the document type and who from a language or from a specific revision. This is why the experience has been designed with more consistent inputs and filters that can be combined.
The document remains a single entity: languages and revisions are connected, not scattered copies. The user can thus reach the correct version and immediately understand what related content is available.
You don't have to know everything. You have to find the correct answer.
The assistant designed for CKS is not a generic chatbot. It will have to operate exclusively on Coltri’s official content, retrieve relevant information and build the answer through a RAG architecture, indicating the sources used.
When the knowledge base does not contain a reliable answer, the system will have to declare it and direct the user to technical, commercial or related documentation. The goal is to reduce research time without replacing skills and responsibilities with uncheckable answers.
The first release is the core. Not the limit.
The first phase includes centralization, synchronization with SharePoint, metadata, document portal, advanced research, multilingual management and AI assistant. It is the core necessary to make the current heritage more accessible and reduce manual activities.
The architecture is designed to later accommodate restricted dealer areas, service content, maintenance, mobile applications and further business systems. Each extension will have to use the same central knowledge, avoiding creating a new silo.
State of the project
A defined ecosystem. Now under construction.
Functional analysis, architecture and prototypes have defined the perimeter of the first phase. Implementation and validation will proceed for progressive releases.
SharePoint
Unique Source
A central process to update documents and metadata without duplicate publications.
Knowledge Layer
Information related
Models, languages, revisions, visibility and relationships become structured knowledge.
RAG + sources
A controlled
The answers will have to use the official documentation and show where they come from.
Let’s build something useful.