People ask how Desk compares with Microsoft Copilot, and the question keeps coming up for a good reason: both read your files and answer in sentences. The answer is that they are not competing. They are built for different jobs.
A familiar version of the same thing is email and instant messaging. Microsoft ships both Outlook and Teams and has never suggested you pick one. Both move words between people. A chat message is for the next ten minutes: quick, informal, and scrolled out of sight by Thursday. Email is the record: addressed, kept, and forwardable to someone who was not in the room. Two tools in the same category, not interchangeable, and nobody argues about it.
Copilot and Desk are that kind of pair. Copilot is built for the work in front of you. Desk is built for what your organization knows, and still knows a year after the person who learned it has moved on.
What a general assistant is for
Copilot, and assistants like it, live inside the tools you already use. They draft the email, summarize the meeting, rewrite the paragraph, and pull relevant items from your mailbox, calendar, and files when you ask. They are optimized for the person and the moment: what you are working on right now, with what you have access to right now.
That design is a strength for individual productivity and a limit for organizational memory. Each conversation is a fresh retrieval over your own permissions. Nothing is read in advance, nothing is turned into a shared, structured record, and two colleagues asking the same question get whatever their own searches happen to surface. When the conversation closes, the work of finding the answer is thrown away, and the next person who needs it starts over.
What a Desk Knowledge Base is for
Desk is built for the question "what do we, as an organization, know?" It reads every uploaded file once and turns it into records, relationships, and cited facts shared by the people you have given access to. Access is granted per knowledge base, so each team sees its own files and nobody else's. Ask it which of your people have done a particular kind of work, and it filters the work records and follows their links to people. Ask it what a client committed to last year, and it returns the commitment with the quote and the file. Ask on Monday and on Friday and you get the same answer, because it comes from the same records.
Three practical differences follow from that:
- Counting and comparing. Desk can answer "how many", "which of our", and "compare" across hundreds of files, because the numbers were read out of the files once and can be filtered. A per-conversation retrieval cannot enumerate.
- One record per thing. A person, company, or project mentioned in a dozen files is one record with a dozen sources, so "what have we done with them" is a lookup rather than a search you hope is complete.
- Standing work. Routines run a question on a schedule and file the cited result, so the Monday check happens whether or not anyone remembers to ask.
What Copilot does that Desk does not
Desk does not draft your emails, sit inside Word, or read your calendar. It does not try to. If the job is "help me write this," use an assistant. If the job is "tell me what we know, and show me where it says so," that is Desk.
Using both
Most teams we talk to keep their assistant for personal work and put their shared institutional knowledge in Desk: proposals, reports, contracts, close-out notes, the files that answer "have we done this before?" The two do not compete for the same question, and neither replaces the other. Nobody uninstalls Outlook because Teams is faster. The quick tool and the one that keeps a record are not doing the same job, and an organization that has only the quick one accumulates nothing.
How Desk helps
Desk is the shared memory; keep your assistant for the work in front of you.
- Files are read once for the whole team, into records shared by the people you grant access to, per knowledge base, so two people asking the same question get the same cited answer.
- Counting, comparing, and following connections across hundreds of files, which a per-conversation retrieval cannot do.
- Routines run the standing questions on a schedule and file the cited results.
Thinking about how this would work for your team?
Let’s chatKeep reading
"Have we done this before?" is the most expensive question in your company
It gets asked a dozen times a week, it is almost always answerable from files you already have, and the answer usually arrives too late to matter.
Read more →Tribal knowledge: what it costs and how to capture it without a reorg
The knowledge that walks out the door is rarely written nowhere. It is written somewhere unfindable. The fix is not asking people to document more.
Read more →