The work in classifying unstructured data happens after the first scan. The first scan tells you what you hold, and then the rules start moving. Your regulator asks a new question, your legal team rewrites a retention clause, one of your business units brings a new kind of document into scope, and every one of those changes has to get into production and reach the files you've already scanned. At enterprise scale, with AI reading the content of millions of files, every pass across that estate carries real time and processing.
We built 8.7 for that part of the job, so a change to your classification rules moves the way any other change to a production system does. You test it first, release it deliberately, and apply it across your data on your schedule, without reprocessing everything you hold. When the audit asks whether the change was tested before it reached production, and when your regulator asks whether it applied to last year's files too, the answer is yes.
Carry Tested Classifiers from Pre-Production to Production
Your team can export the classifiers of one DataXray environment as a single bundle and import that bundle into another environment or organization. The bundle holds the labels, detection rules, extraction instructions, and settings that were built and approved in pre-production, so production receives the same classification logic that was tested rather than a rebuilt copy of it.
On import, DataXray matches classifiers by name. Anything that already exists in the destination is updated, anything new is created, and nothing already there is removed. The import is all or nothing, and both environments must run the same DataXray version. The bundle carries the classification logic and not the connections to your data or the AI model each extraction uses, so your team connects production data sources and chooses models in the production environment itself. That keeps control of production data where it belongs.
When a classifier changes later, the team exports an updated bundle and imports it again, and every environment receives the same change.
Change a Classifier Once and It Reaches Every File You've Already Scanned
Say a regulation changes what you need to know about a contract, and you improve the instruction that extracts that detail. The next time DataXray scans those files, it applies your improved instruction to every contract it finds there, including the ones it classified a year ago, so your whole estate answers the new question. Re-reading every file takes time and AI processing, which is why DataXray does it only when your change affects what gets extracted. When you adjust a file filter or make an edit that doesn't touch extraction, your existing values stay where they are and DataXray extracts only from files that are new or have changed.
Two settings leave the finer choices with you. Keep extracted values holds on to answers you no longer extract, so narrowing what you extract doesn't throw away results you've already spent scan time on. Extract new files lets a data source keep indexing and classifying new files without extracting from them. When you upgrade, your existing setups keep the behavior they had, so Keep extracted values is off on them until you switch it on, and anything you create in 8.7 has it on from the start.
Classification Stays Within the Scope You Set
Governance obligations usually name kinds of documents rather than file shares, and file filters in 8.7 let you work the same way. Rather than listing every file type to skip, you name the ones to classify and DataXray leaves the rest alone. For a contracts program, that might be just PDFs and Word documents. Those other files are still discovered and still searchable by their metadata, so your inventory stays complete. What narrows is the classification effort, and that becomes a decision you make rather than a list you have to maintain.
Backups Finish Sooner and Restores Are More Dependable
Backups now compress on every processor core in the host and no longer make a full copy of the data first, so they finish sooner and need less temporary disk along the way. Restoring from a backup is more dependable too.
You can upgrade to 8.7 from 8.6 only. Your account manager or partner manager can plan the redeployment with you. Once you're on 8.7, the three things I'd do first are export a classifier bundle from pre-production, check whether naming the file types to classify can replace a long exclusion list, and decide whether Keep extracted values should be on for your existing setups. The full release notes are in the DataXray documentation.