Frequently asked questions

Desk turns the documents, files, and knowledge an organization already has into a knowledge base its people can ask. Files are read once; the people, organizations, projects, decisions, and numbers inside them become records with citations back to the source. You ask questions in plain language and get answers that point to the files they came from. Routines let you write a repeated question down once and run it on a schedule.

Teams whose best answers live in someone's memory: consultancies qualifying for RFPs, clinical research networks answering feasibility requests, law and advisory practices, engineering and architecture firms. Any team that wins work on precedent and spends too long finding it.

Search finds documents. Desk reads them into records and relationships, so it can filter, count, and compare across files, not just quote one. Every answer cites its sources, and Desk is instructed to report what the files say rather than infer, and to say plainly when the files are silent. See Our approach to knowledge management.

Those are assistants for the person and the moment: they help you write, and they retrieve from the files in front of you each time you ask. Desk is a shared knowledge base: files are read once into records, relationships, and cited facts that the whole team queries, counts over, and checks on a schedule. The relationship is the one between instant messaging and email: both carry information, both are worth having, and nobody asks you to choose. See the comparison articles for the detail.

Yes. Each is separate inside your organization, with its own files, records, searches, and routines. Most teams use one per body of knowledge that should not mix, such as one per practice area or per business unit. Members see only the knowledge bases they are given.

PDF, Word, Excel, PowerPoint, Markdown, plain text, CSV, and JSON, plus scanned PDFs and images through OCR. Spreadsheets are read as labeled rows so their columns become fields.

No. Upload them as they are. Desk reads what is inside and keeps the original searchable. Nothing depends on folder structure or file names.

No. Every Desk knowledge base uses the same structure: the people, organizations, places, work, agreements, events, issues, commitments, and topics your files mention, with your industry's own words kept alongside. There is nothing to design or maintain, and files are never re-read because something changed. Upload and ask. See Our approach to knowledge management.

Each new file is read as it arrives and merged with what Desk already knows: a person or study mentioned before gains a source, a new one gets a record. There is nothing to set up or redo, and earlier files are never re-read. You can ask about the new file within a minute or two.

Creating a knowledge base takes a minute. Reading the first twenty to thirty files takes a few minutes more, and you can ask while that runs. There is no structure to design or approve. Most teams are asking real questions in their first hour.

Every claim must cite a file, and Desk is instructed to state only what a file states, to keep the file's own wording for reasons and decisions, and never to connect two facts causally unless a file does. When it is asked for a judgement, it labels that sentence as its own inference. When the files are silent, it says so. See Citations or nothing.

Yes. Records and files in an answer are links. Clicking one opens the record or the file so you can see the supporting passage.

Yes. Attached files are read as context for that question only and are not added to the knowledge base. Desk cites them as attached files so you can tell them apart from your permanent records.

A repeated job written down once: a name, instructions in plain language, and a schedule. Desk runs it on demand or on schedule and saves the cited result. Routines can carry standing reference files, such as a rubric or template, and individual runs can take one-off files. See Routines.

In Desk's hosted environment, with each organization's files and records kept separate from every other customer's. Each knowledge base is its own database. Dedicated deployments are available for organizations that need their own instance. See the Security page.

No. Your files and the records derived from them are used only to answer your organization's questions and run your routines. They are not used to train any model, ours or a provider's.

Desk uses commercial language-model and embedding providers under agreements that do not permit them to retain or train on the content sent to them. The providers in use are listed on the Security page and may change as better options become available.

Access is per organization and per knowledge base. Admins manage people and knowledge bases for their organization; members see only the knowledge bases they have been given. Every request is checked against these permissions.

Yes. Deleting a file removes its text, its search index, and the records only it supported. Deleting a knowledge base removes everything in it, and only an admin can do it, by typing its name. On request we delete an organization's data entirely.

Desk is priced per organization based on the number of knowledge bases and the volume of files read. We are working with early customers directly; request a demo and we will scope it with you.

Yes, and we recommend it. One knowledge base, one team, their real files. Expand when the first team is asking questions it could not answer before.

Not yet. Every deployment currently starts with a walkthrough on your own files, so the first questions your team asks are real ones.