Leadership wants data prepared for AI, but there isn't a clear use case or scope. The content is spread across old file shares and document systems, with inconsistent access and duplicate versions. How do you turn that request into something manageable?
How do you scope a vague request to get data ready for AI?
Asked by
+briefly!in_transit
Answers
Start with what a specific group of users wants to do, not with a cleanup of everything. If they want to search a few document sites through a chat interface, those sites become the initial scope.
Agree on which content is included, who can access it, who owns it, and what a useful result looks like. That gives you a way to assess the work before promising an organisation-wide rollout.
Reply to this answer
Do a scoped audit before anyone estimates the AI work. Look at access permissions, duplicates, conflicting versions, and whether anyone can say which copy is authoritative. Messy content and broad access can become much more visible when a search tool makes it easier to find.
The output can be a list of decisions and cleanup tasks, not a promise to fix every repository. AI might help people find things, but it won't decide which document is the approved one or who should be allowed to see it.
Reply to this answer
Ask who is sponsoring the cleanup and where the time comes from. Doing it internally still has a cost if the same people are expected to keep up with their existing work. An outside provider needs a budget too.
Put the assessment and remediation on the roadmap as actual work, with an owner and an estimate after the scope is known. Otherwise it stays invisible until the deadline slips.
Reply to this answer
Ownership is the sticking point. A policy saying departments must manage their content doesn't help much if nobody will back it up. Get a named business owner for each in-scope collection and agree who can make decisions about obsolete or conflicting material.
If those decisions can't be made, flag that as a dependency rather than quietly leaving IT to guess.
+briefly!in_transit:
That distinction helps. The technical team can identify conflicts, but the business still needs to decide which content is valid.
Reply to this answer
Keep the access question separate from the content cleanup. An AI tool shouldn't get broad administrative access just to make a pilot easier. Check what the intended users and the tool can access before connecting any repository.
Reply to this answer
Add your answer
Answers are reviewed before publication.
!steady~the+envelope:
That also gives the business a chance to say which sources actually matter. An old file share doesn't have to be in the first phase just because it exists.