Search across company documents
The answer is in a manual somewhere, but no one knows which one.
- What we do
- We build an internal search engine over your manuals, contracts and procedures, in several languages: your team asks in its own words and gets the answer together with the documents it comes from. The documents and the data remain your company's.
- How we do it
- First we look with you at which documents go into the search engine and where they are today: shared folders, the intranet, or the program where a department keeps its documents. We check that the documents and their permissions can be read from each folder or system, and we decide with you who can see what, because not everyone gets to see a contract: the search engine follows those permissions. Then we connect the search engine to those folders, so that it reads each document in its own language and keeps reading the ones that are added or changed. We test with your team's real questions: every answer has to cite the document and the section it comes from, so that whoever asked can check it, and we also test questions whose answer is not in the documents, to check that in that case the search engine says so. Once the test answers line up with the documents, we open it to the whole team.
- What you receive
- An internal assistant that cites its sources in every answer.
- What changes for you
- Your team stops searching folder by folder, or asking the one person who always knows where everything is. The answer comes with the document it is drawn from, even when that document is in another language, and someone new to the company finds the procedures without waiting for someone to explain them.
- What we need from you
- The list of the folders or systems where the documents are, with access to read them; the list of the people who will use the search engine and which kinds of document each of them may see; and one person from operations or technical support who gathers the questions your team asks most often and reviews the test answers with us.
- Who it is for
- Operations and technical support
- How you buy it
- A project, plus a recurring fee.
Example use cases
These are examples: they do not describe a real client.
A packaging machine maker in Italy with technical service teams in Spain and Mexico
The situation
The technicians who go out to customers' plants work with the manuals of many models, some in Italian and others in English, scattered across several shared folders. When a machine stops, the technician calls head office in Italy, and the answer depends on who picks up the phone.
What we do
The search engine reads the manuals, the technical notes, the bulletins and the maintenance contracts, which we gather in the connected folders. The technician asks in Spanish, for instance what an alarm means on a given model, and gets the answer in Spanish with the manual and the section it comes from, in its original language. The internal technical notes are seen only by the technicians, and the maintenance contracts only by the office.
A logistics company in Spain with offices in Italy and Colombia
The situation
The procedures of each warehouse, the contracts with customers and the instructions for each service are written in Spanish or in Italian, depending on the office that drafted them. When a customer asks what their contract says about returns, customer support searches several folders and ends up asking another office.
What we do
The search engine reads the procedures and contracts of all its offices, each in its own language. Customer support asks what that customer's contract says about returns and gets the answer with the contract and the clause it comes from, to read it before replying. The contracts are seen only by the people the company decides.
Frequently asked questions
Can an answer be wrong?
Yes, an AI tool can get things wrong. That is why every answer cites the document and the section it comes from, so that whoever asked can check it, and during the test we check with your team that the answers line up with the documents.
Who sees each document?
Whoever you decide. Before we start we decide with you who may see each kind of document, and the search engine answers each person only from the documents they may see.
Who owns the documents and the data?
Your company does. The original documents stay where they are; the search engine keeps a copy of their text to search it. The search engine, its code, its accounts and its documentation are in your name from day one.
And the documents added later?
The search engine reads them when they are added or changed in the connected folders, with their permissions, and stops using the ones that are deleted. A scanned document or a drawing is checked during the test: if it cannot be read, you know before the search engine is opened to the team.
Where is the data processed?
In the AI service we choose with you before we start: the written scope says which one it is, in which region it processes the data and which data it receives. The account for that service is in your name from day one.