Answers from your own documents
Staff ask a question in plain language and get an answer from your own regulations, files and procedures. Not a list of search results, but an answer with its source.
Your documents answer questions. Only for those allowed to see them.
Every organisation sits on years of documents that people ask questions about every day, and looking up the answer costs time over and over again. A knowledge assistant on your own documents solves that, provided it knows who is allowed to see what and where every answer comes from. besst.ai builds those assistants on your own documents, with your own permission structure, so your documents stay your documents. In your own environment: your Azure tenant or fully offline on your own hardware.
- Your own environment
- Permissions as the starting point
- Insight into sources
- Offline too
- SharePointDocuments
- OneDriveFiles
- TeamsCollaboration
- OutlookEmail
- Other sourcesTo be agreed
- Question
- How many days a week may I work from home?
- Answer
- Up to two days a week, in consultation with your manager.
- Source
Role switch
What you get
Permissions that keep holding
Anyone without access to a document should get no answer from it either. Not in the text, not in the source reference, and not as a confirmation that the document exists. That is the design principle, and it is tested in your own environment before anything goes into production.
See how permissions workIn your own environment
Everything runs in your own environment, under your own control. Your documents do not leave that environment.
From question to a working knowledge assistant
Three phases. At the end of each one it is settled what you have, and which failure point is solved.
Exploration
We look at your documents, your permission structure and the questions your people actually ask.
- permissions mapped
- sources mapped
What you have
An assessment: where does it break down for you, and is it worth doing.
- A possible application
- An inventory of your sources
- An analysis of your permissions
- The risks
- Advice on a proof of concept
We map which documents exist, how permissions are arranged today and which question should be answered first. You get a reasoned yes or no, and if it is a yes, a bounded assignment.
Proof of concept
A working setup in your own environment, on a bounded part of your documents.
- permissions
- provenance
- silence
What you have
Answers you can verify yourself, with the source attached, and with permissions left intact.
A working assistant on a bounded part of your documents, in your own environment, in the cloud or offline, with the permission separation built in. Meant to be used, not demonstrated. By the end you know whether this works for you.
Production
The setup grows to your whole corpus and becomes part of your daily work.
- staleness
- fragmentation
What you have
A service that stays current and that answers from your fragmented sources as one whole.
Extending to the rest of the documents, connecting to your existing systems, setting up management and handing over to your own people.
At the end of each phase it is settled what you have and whether the next one is worth doing. How long a phase takes depends on the size of your document set and on the pace at which you arrange access and decisions. Waiting on those is in practice the largest delaying factor, and that is also why phase 1 exists. After the exploration we name a period for the proof of concept, and not before.
“When will I know whether this works for us?”
At the end of phase 1. Not earlier, and not later either.
Where it runs
In your own Azure tenant
Your data stays in your subscription, under your control, with your permission structure.
Fully offline
No internet, no cloud, open-weight models on your own hardware. Runs today on an ordinary laptop.
- Your files
- Your mail
- Your systems
Both variants solve the same five failure points. You choose the place, not the level.
“Does our data have to go to the cloud?”
No. The same assistant also runs fully offline, without internet, on your own hardware.
Where it breaks down
Building a knowledge assistant that answers questions from your documents is not the hard part. The hard part comes after. Five things that stop almost every project:
The moment personnel files, objections and executive papers enter the same system, the assistant has to know who may see what. If that fails, either everyone sees everything or the project stops. That principle applies in both runtime environments.
See the illustration above.We work through these five, in that order, and test them in your own environment before anything goes into production. In an Azure tenant of your own, or fully offline on your own hardware. Exactly how is something we discuss in person.
TechnologyWant to know whether this works for your organisation?
An exploratory conversation of one hour usually gives a clear picture. Get in touch and we will schedule it.
We show it in your own environment. Without internet too, on a laptop at your table.
besst.ai. Your documents answer questions. Only for those allowed to see them.