--- title: Before you turn AI loose on your shared drive description: AI search answers from whatever files it can reach. What to clean up first, from expired records and overshared folders to old scans AI can't read. --- [Skip to content](https://www.dtiibs.com/insights/before-you-turn-ai-loose-on-your-shared-drive#main) [**DTI**Integrated Business Solutions](https://www.dtiibs.com/) Menu [Documents and records](https://www.dtiibs.com/document-management) [Document management](https://www.dtiibs.com/document-management) [Document scanning](https://www.dtiibs.com/document-scanning) [Large-format scanning](https://www.dtiibs.com/large-format-scanning) [Scanning Scope Worksheet](https://www.dtiibs.com/scanning-scope-worksheet) [Accounts payable](https://www.dtiibs.com/accounts-payable-automation) [AP automation](https://www.dtiibs.com/accounts-payable-automation) [Invoice processing services](https://www.dtiibs.com/invoice-processing) [Vendor payments](https://www.dtiibs.com/vendor-payments) [For Penta customers](https://www.dtiibs.com/penta) [AP Assessment](https://www.dtiibs.com/ap-assessment) [Insights](https://www.dtiibs.com/insights) [About](https://www.dtiibs.com/about) [336.346.1580](tel:+13363461580) [Talk with DTI](https://www.dtiibs.com/contact) [Insights](https://www.dtiibs.com/insights)/Document Management # Before you turn AI loose on your shared drive R **Ryan White** November 2, 2026 · 7 min read AI search tools answer from whatever files they can reach, so outdated copies, expired records and folders shared too widely all end up in the answers. Before you turn one on, settle who can see what, clear out what you no longer need to keep, and make sure your scanned records have text a machine can actually read. ## AI can see whatever your people can see Microsoft's [guidance on preparing SharePoint for Copilot](https://learn.microsoft.com/en-us/sharepoint/sharepoint-copilot-best-practices), updated in August 2026, explains that Copilot respects the permissions and sharing settings you already have. It doesn't give anyone access they didn't already have. The flip side is that it will happily surface anything your current settings allow, and Microsoft notes that SharePoint's default sharing settings are the most permissive ones. That's why Microsoft's own checklist starts with tightening sharing, reviewing who has access to what, and archiving inactive sites. If your organization is planning to turn on Copilot, that review belongs with whoever manages your Microsoft 365 environment, and it should happen first. ## Most organizations don't know what's in their files Permissions are only part of it. The bigger problem is that most organizations have lost track of what they've stored. In a [BARC study](https://barc.com/news/unstructured-data-barc-study/) published in June 2026, only 29% of roughly 220 data and AI leaders said they fully know where their AI-relevant documents and files live. Seventy percent said less than half of that information is discoverable and usable for AI. A [Komprise survey](https://www.blocksandfiles.com/data-management/2026/10/19/komprise-ai-needs-ai-ready-data-and-there-isnt-enough-of-it/5303841) reported this month found that only 7% have classified and cleaned the majority of their file data. Those surveys focus on larger organizations with dedicated data teams. If your company doesn't have one, it's worth asking how your own files would score. ## The history nobody deleted When a customer asks us to help move out of an old file share or archive and into a newer, more secure platform, there's usually a lot of history in it. Not all of it needs to be kept. We regularly find records that passed their retention date years ago and should have been cleaned out. Deciding what to remove is hard when there's this much of it, especially for a company that has never deleted anything. Without a clear way to decide what stays and what goes, the cleanup gets overwhelming quickly and tends to stall. And every expired record you keep is one more thing an AI tool can find and put in front of someone. ## Old scans that AI can't read We also see a lot of documents that are images only. Companies that started with document management early often scanned their records as TIFF images, and those files don't have the full-text layer that most modern searchable PDFs include. A person can read the page, and an AI assistant can often read it too if you hand it that one file. Search tools, and AI assistants searching across a library, generally need a text layer to find it in the first place. Some older document management systems did run full-text OCR, but kept that text in a separate database instead of inside the files. When you migrate out of one of those systems, the images can come along while the searchable text stays behind, unless the migration plans for it. For those records to be searchable by AI, the text has to be pulled out first, either by running them through optical character recognition (OCR) or through newer AI capture tools that read the page and create that text layer. If your archive dates back to the early days of scanning, assume a good share of it is in this condition until you check. The same goes for any paper still sitting in boxes: it's invisible to AI until it's [scanned with searchable text](https://www.dtiibs.com/document-scanning). ## Asking an AI agent to find a record One customer had us scan their records into searchable PDFs. Their plan was to store everything in SharePoint behind Teams and use an AI agent to search for and retrieve records. Once it was running, they found it didn't give them everything they wanted. Asking Copilot to "go find me" a specific record or file turned out to be a slow way to retrieve records they needed regularly. They wanted to pull up a particular record quickly and know it was the right one. AI tools are useful for answering questions across a lot of material. When your team needs a specific record by name, number or date, we think a document management system with proper index fields still does that job better. Document management still plays a big role in keeping information findable and organized, and we're not at the point where AI agents can run everything. ## When documents belong in a document management system A document belongs in a [document management system](https://www.dtiibs.com/document-management), instead of a SharePoint site or a network folder, when there's a real business case behind it and people regularly search for and retrieve it. Contracts, employee files, invoices and customer records usually qualify. A folder of meeting notes usually doesn't. How you manage a record over its life should be decided in this order: 1. How useful it still is to the business 2. What it costs to keep it where it is 3. What your retention program requires When a record starts in a solid document management system, that system can usually carry it through its whole life, from active use to final disposal, without much added cost. That often removes the need to move older records to cheaper storage tiers, which is usually where companies start losing track of what they have and who can see it. ## What to do this week You don't need a big project to get started. A few practical steps: - Ask whoever manages your Microsoft 365 environment whether sharing has been reviewed before any AI tool is turned on. - Pick one shared drive or archive and find the records that are past their retention date. Confirm your retention requirements with your records officer or legal counsel before destroying anything. - Open a few of your oldest scanned files and try to search or select the text. If you can't, those files need OCR before AI can use them. - List the records your team retrieves most often. Those are the first candidates for a document management system with proper index fields. - Count the boxes of paper that are still in storage, and decide which ones are worth scanning. ## Frequently asked questions **Will Copilot show employees files they shouldn't see?** Copilot only shows people what they already have permission to access. The problem is that many folders are shared more widely than anyone realizes, so Copilot can surface files that were technically accessible but never meant to be found. Reviewing sharing settings before you turn it on is Microsoft's own first recommendation. **Can AI read our old scanned documents?** Sometimes, but not the way most people expect. Many AI assistants can read a scanned page if you open or upload that one file. Searching across thousands of image-only files is different: search tools, and AI assistants searching a whole library, rely on a text layer, and older scans saved as TIFF images or image-only PDFs usually don't have one. Those files need OCR or an AI capture tool first. Some platforms can do this automatically, but usually only for new uploads, not the archive you already have. **Can we put our scanned records in SharePoint and search them with Copilot?** You can, and it works for some purposes. For records your team retrieves regularly by name, number or date, a document management system with index fields is usually faster and more reliable than asking an AI assistant to find them. **Should we clean up our files before turning on AI search?** Yes. Clearing out expired records, retiring duplicate versions and fixing overshared folders all improve what AI tools return, and they reduce the risk of the wrong file reaching the wrong person. **How do we decide what to keep?** Look at how useful each type of record still is to the business, then what it costs to keep where it is, then what your retention program requires. Confirm retention periods with your records officer or legal counsel before destroying anything. ## Let's get your records in shape If you're sorting out an old archive, a file share or boxes of paper before AI tools arrive, we can help you decide what to scan, what to index and where it should live. Start with our [Scanning Scope Worksheet](https://www.dtiibs.com/scanning-scope-worksheet), or [talk with us](https://www.dtiibs.com/contact) about your records. About the author **Ryan White** Talk with DTI ### Have a question about your own process or records? Tell us what you're working on and we'll set up a call with the right person. [Start a conversation](https://www.dtiibs.com/contact) ## Keep reading ### [Accounts Payable Workflow Automation What goes wrong in AP automation projects Most AP automation projects stall on exception rules, approvals and testing. Here is what we see go wrong and what working projects do before go-live. November 2, 2026](https://www.dtiibs.com/insights/what-goes-wrong-in-ap-automation-projects) ### [Document Scanning What to do with a room full of old records Clearing a file room for a move or remodel? Why reviewing every folder stalls the project, when to scan first, and when you can shred the originals. November 2, 2026](https://www.dtiibs.com/insights/what-to-do-with-a-room-full-of-old-records) **DTI**Integrated Business Solutions DTI Integrated Business Solutions 3825 W. Market St., Suite 102 Greensboro, NC 27407 [336.346.1580](tel:+13363461580) **Customer support** [866.384.7378](tel:+18663847378) [support@dtiibs.com](mailto:support@dtiibs.com) [Support portal login](https://connectwise.dtiibs.com/v4_6_release/services/system_io/customerportal/portal.html?company=dtiibs&locale=en) #### Documents and records [Document management](https://www.dtiibs.com/document-management)[Document scanning](https://www.dtiibs.com/document-scanning)[Large-format scanning](https://www.dtiibs.com/large-format-scanning)[Scanning Scope Worksheet](https://www.dtiibs.com/scanning-scope-worksheet) #### Accounts payable [AP automation](https://www.dtiibs.com/accounts-payable-automation)[Invoice processing services](https://www.dtiibs.com/invoice-processing)[Vendor payments](https://www.dtiibs.com/vendor-payments)[For Penta customers](https://www.dtiibs.com/penta)[AP Assessment](https://www.dtiibs.com/ap-assessment) #### Company [About DTI](https://www.dtiibs.com/about)[Insights](https://www.dtiibs.com/insights)[Partners and referrals](https://www.dtiibs.com/partners)[Contact](https://www.dtiibs.com/contact) © 2026 DTI Integrated Business Solutions. [Privacy policy](https://www.dtiibs.com/privacy-policy)Internal draft, not for publication. ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Ryan White", "url" : "https://www.dtiibs.com/insights/author/ryan-white" }, "dateModified" : "2026-11-02T21:49:19.919Z", "datePublished" : "2026-11-02T19:32:26.000Z", "headline" : "Before you turn AI loose on your shared drive", "mainEntityOfPage" : { "@id" : "https://www.dtiibs.com/insights/before-you-turn-ai-loose-on-your-shared-drive", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://www.dtiibs.com/hubfs/logo_main.png" }, "name" : "DTI Integrated Business Solution" } } ``` ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Ryan White" }, "dateModified" : "2026-11-02", "datePublished" : "2026-11-02", "description" : "AI search answers from whatever files it can reach. What to clean up first, from expired records and overshared folders to old scans AI can't read.", "headline" : "Before you turn AI loose on your shared drive", "mainEntityOfPage" : "https://www.dtiibs.com/insights/before-you-turn-ai-loose-on-your-shared-drive", "publisher" : { "@type" : "Organization", "name" : "DTI Integrated Business Solutions" } } ```