Project Knowledge
We have built this before. But where are the documents?
An AI search across your project folders, quotations and reports finds earlier solutions when someone asks in their own words. With the source and, where the documents say so, the name of the person who worked on it back then. For technical B2B companies whose knowledge is spread across many years of project work.
Connect project folders, requirement specifications, bills of materials and reports.
Finds the belt conveyor even when the question asks about a material handling system.
Every statement links to the original document, visible only to those with access.
In short
How do you make project knowledge searchable?
You connect your existing storage, meaning project folders, quotations, requirement specifications, acceptance reports and complaints, to an AI search that looks for meaning instead of file names. Two conditions apply from the start: every answer names the document it comes from, and the search only shows what the person asking is allowed to open in the original storage. You start small: one area, real questions from your team and a test of whether answers, sources and access rights are correct.
What the search cannot do: it does not find knowledge that was never written down. That takes conversations with experienced colleagues before they leave.
Day to day
The knowledge exists. It is just buried in a thousand folders.
In sales
The old quotation is somewhere.
A customer wants “something like 2019, only bigger”. The quotation from back then is called “Quote_final_v3” and sits somewhere under “Projects_old”. It turns up through an email to everyone, if someone remembers.
In engineering
The solution is worked out a second time.
A commissioning problem was solved years ago. The solution is in an acceptance report that nobody knows about today.
Generational change
A colleague retires in two years.
He can explain every custom solution. Part of what he knows is in his folders, part of it only in his head.
The usual file search often does not help much, because it looks for words, not meaning. Anyone typing “material handling system” will not find the “belt conveyor”. On top of that, every department has organised its folders differently.
What it can look like
One question. Three earlier projects. The sources.
Inside sales asks in their own words. The search finds the matching projects and shows where each statement comes from.
“Have we ever quoted a stainless steel belt conveyor for the food industry?”
2019: quotation to a dairy, no order. Stainless steel 1.4301 according to the quotation text.Project folder 19-044, quotation p. 3
2021: quotation and order for a bakery products manufacturer, with acceptance report. Stainless steel according to the bill of materials.Project folder 21-112, bill of materials rev. C
2023: enquiry without quotation.Project folder 23-017
The 2021 acceptance report notes a cleaning problem at the deflection pulleys. The change is in bill of materials rev. C.Project lead 2021 according to the project sheet: H. Berger
Your team opens the original documents before reusing the old solution.
Example with sample data – not a customer reference and no guarantee of results.
How it works
From the first conversation to a tested pilot.
One area, for example the projects of one product line from the last ten years, and around 30 real test questions. Then you decide: expand, improve or stop.
01
Collect questions
What do sales, engineering and service ask about earlier projects? We collect real examples, not invented ones.
02
Review storage
Where are project folders, quotations and reports? What is readable text, what is an old scan? Who may see what?
03
Build and Test the Pilot
Your experts define in advance which projects are the right answer and when the pilot has passed. We check hits, sources and access rights, and whether the search honestly says “nothing found”.
04
Decide
You decide based on the test results, not on a demo.
To be honest
What the search can and cannot do.
What it can do
Find, combine, show contradictions
- Find documents by meaning, even when they use different terms than the question.
- Combine information from several documents, each with its source.
- Find old projects as well as new ones, as long as the documents are readable as text.
- Show contradictions, such as two versions of a bill of materials, instead of quietly picking one.
What it cannot do
Invent knowledge or replace experts
- Find knowledge that was never written down.
- Understand the content of CAD models. It finds drawings through the project number, bill of materials and accompanying documents.
- Decide whether an old solution still applies today. Your specialist checks that.
- Guarantee that it has found everything. A source shows where a statement comes from, not that there are no others.
Our recommendation
Capture the knowledge in people’s heads
Colleagues who retire in the next few years explain in short conversations beforehand why custom solutions were built the way they were. Filed as minutes, this becomes searchable project knowledge. We support you with this as part of the digitalisation project: we prepare the conversations, record them as minutes and file them so that the search finds them.
Is this right for you?
Making project knowledge searchable pays off when …
Variants
… you build many similar but never identical projects
Such as special machines or variants of a product line.
Quotations
… new quotations often build on earlier ones
And searching for the right old quotation takes time.
Generational change
… experienced colleagues will leave in the next few years
Then it matters that their knowledge can be found in the documents while they can still answer questions.
And if not: if your projects are properly tagged in the ERP and the search there is enough, you do not need AI for this. We will tell you that in the introductory meeting too.
Who is behind it
Advice and delivery from one source.
You talk to the person who also builds your project search.
Client project · in use
AI knowledge assistant for APIORIUM
A multilingual website chatbot with its own knowledge base answers questions about services, locations and products.
Personal
Martin Trixner
Has been building software and integrations since 2002 and is a certified AI-Automations-Manager.
For your projects
Internal, with sources and access rights
For your internal project search, sources, access rights and the connection to your storage are added.
Frequently Asked Questions
Frequently asked questions about project knowledge and AI.
Isn’t Microsoft 365 Copilot enough for this?
Sometimes, yes. Copilot searches what is stored in Microsoft 365 and what you are allowed to read. If project folders are on a file server, in an older archive or in the ERP, Copilot will not find them without an additional connection. For some storage locations there are ready-made Copilot connectors, for example for Windows file shares. We first check whether that is enough for your case and tell you if it is.
Do we have to tidy up our storage first?
No, not completely. For the pilot we choose one area. We mark outdated versions and duplicates, and what should not go in stays out.
Who sees which projects?
The same people as in your storage today. The search takes over the access rights of your storage, even when they change or a document is deleted. This is a fixed acceptance condition: before go-live we test it with questions from different roles and with changed access rights.
Will our project documents go to external AI services?
Not without an explicit agreement. Where the AI runs and which data it sees are agreed jointly and in writing before the pilot.
What is the difference from the AI knowledge base in service?
The knowledge base answers questions from manuals and service documents. This is about completed projects: what was quoted, built and complained about. Technically it is the same building block, the sources and questions are different.
What does it cost?
It depends on the amount and type of documents, the number of users and the connections. After reviewing your storage you receive a written quotation, with implementation and operation listed separately. If you first want to find out where AI pays off in your company at all, start with the Compact AI opportunity assessment for €1,900 excl. VAT.
Next Step
Which earlier project did your team last search for?
Bring two or three questions your team asks about earlier projects, and a rough description of your storage. In the free introductory meeting with Martin Trixner we clarify whether a pilot is worthwhile and what preparation it needs.