Once the folder structure is right, the next pass cleans filenames. The AI proposes renames for the worst offenders, leaves the good ones alone, and lets you defer metadata to a separate step.
After the AI organized the library, the narrator immediately spotted naming problems — filenames like cryptic IDs and inconsistent prefixes. The next move was to ask SharePoint AI to "suggest ways to better name" the files. The AI returned a list of rename proposals with both current names and recommended names side by side.
Each suggestion is auditable. You see the rule the AI applied (consistent prefix, date format, removal of internal codes) and can approve, edit, or reject it. The bulk rename happens after you confirm — there's no surprise mass-edit.
The narrator asked the AI to "rename the worst offenders" rather than renaming everything. This is a scope filter: it tells the AI to ignore files whose names are already acceptable and focus on the ones causing the most confusion — duplicates, cryptic codes, broken dates, missing context.
After the run, the library showed consistent "014" prefixes where there had previously been a mix of formats — exactly the change the AI had recommended. Other filenames stayed untouched.
When organizing, the narrator deliberately said "I'm going to ignore metadata for now." Metadata extraction (risk levels, document types, dates, owners) is its own pipeline — it answers different questions than folder structure does, and mixing the two passes makes both worse.
Once folders and filenames are stable, the same library can be re-scanned with a metadata extraction skill — the topic of the next video in the series. Trying to do all three at once typically produces inconsistent results and harder-to-debug edge cases.
The whole demo follows a pattern: scan → review options → apply a subset → re-scan → fix the next layer. After the file moves the narrator immediately spotted naming issues and ran another pass. This is the workflow — not a single magic button, but a short loop of AI suggestions and human approvals.
By the end of the demo the library had a real folder structure, consistent naming on the previously broken files, and a clean baseline for the next pass (metadata). All from inside the SharePoint UI, with no scripts or third-party tools.