How to Automate Relativity

Relativity is where a matter turns into documents somebody has to read

Relativity is the platform legal teams use once a case becomes data. A workspace holds everything collected for one matter, and every person on that case works inside it: the associate coding documents, the project manager watching throughput, the partner who wants to know whether privilege review will finish before the production is due.

The company traded as kCura until 2017, when it renamed itself after the product. RelativityOne is the cloud version, and the same workspace idea sits underneath both. Law firms use it, so do corporate legal departments, government agencies and the service providers who host review for them.

What separates it from ordinary business software is volume and consequence. One custodian's mailbox can run to hundreds of thousands of documents, and a privileged document that went out in a production is not a data entry mistake. It is a problem with a judge attached to it.

A custodian's laptop is chased by email, never by software

Very little of a review manager's week is spent looking at documents.

Last Tuesday went on finding out whether the custodian in the Chicago office ever handed over the laptop, which meant an email, a second email, then a call to their assistant. It went on noticing that a scheduling order had moved the substantial completion date, which the court published on its docket and told nobody about. It went on writing the status update for the client's general counsel, who has no Relativity licence and does not want one.

Then there is the exception list from processing: the files that would not extract, the password-protected archives, the containers that came back corrupt. Each is a small decision, and the pile gets worked through on a Thursday because nothing surfaced it on Monday.

Relativity

The client update written overnight Review progress turned into the sentences a general counsel actually reads, drafted before anyone arrives and waiting to be sent.

Deadlines found on the docket The court's own record checked each morning, so a moved production date reaches the schedule the day the order issues.

Exceptions gathered into one decision Processing failures swept and grouped by cause, so a dozen encrypted archives become a single call rather than a dozen small ones.

Active Learning ranks the documents once somebody has collected them

Relativity automates review thoroughly, and a team still hand-managing batches is doing work the platform stopped requiring.

Review Center builds review queues straight from saved searches, which removed the old ritual of creating, refreshing and managing batches. Its Active Learning queues come in two shapes, Prioritized Review to surface the most relevant documents first and Coverage Review to train the model quickly, and the classifier runs in the background as coding decisions land. The aiR suite goes further: aiR for Review ranks documents against the case criteria, aiR for Privilege flags what looks privileged, and aiR for Case Strategy assembles a narrative from what has been coded. Relativity Legal Hold and Collect issues the hold and pulls data from Microsoft 365, Google Workspace and Slack, and processing scales without anyone asking it to.

Every one of those acts on data already in the workspace, or on a source Relativity has been given a connector to.

A court's docket is not one of those sources. Neither is the client's own ticketing system, the opposing party's transfer site, the reviewer roster kept in someone else's HR system, or the inbox where a custodian promised to send the phone next week.

The court's calendar sits outside every workspace

Put the workspace and the systems around it on one screen each morning, and case management stops running a step behind.

The morning's review metrics turned into the update the client actually reads, in sentences rather than a screenshot of a dashboard, drafted and waiting for the case manager to send. Batch assignments proposed against who is available and what is still uncoded, so a reviewer signing on at eight has work in front of them.

The docket read for every open matter, so a production date that moved on Thursday afternoon is on the schedule Friday rather than the week it bites. Processing exceptions swept each morning and grouped by cause, so twelve encrypted archives become one decision instead of twelve. Custodians who have not handed anything over chased on a schedule, each time with the list of what is still outstanding attached.

Reviewer certifications and workspace access expiries watched, because the reviewer who loses access on Monday is always on the matter with the deadline. Quotes left sitting with a service provider followed up while the scope is still current.

Reading and drafting run unattended. Nothing gets coded, produced, filed or sent to a client except by somebody qualified who has read it first.

A privilege call is not clerical work

None of the chasing above is legal judgement. It is one person opening five systems before nine and turning what they see into a list, and it is the first thing that slides in the fortnight before a production.

WebRun is an agent that works a real Chrome browser, signed in as you. It opens the workspace alongside the court's system, the client's portal and your own inbox, reads what each one shows, and hands back the update, the chase list or the exception summary already written.

It runs on your schedule in your own private environment, sessions are not shared between tools, and a workflow can be locked to a named list of sites. You can watch a run and stop it at any point.

The workflows below are already built, and each one names what it opens.