Early case assessment

Early case assessment gives you a first look at a whole case before anyone has scanned anything: how many documents it holds, whose they are, what kinds of file they are, when they are dated and who writes to whom, with suggested review segments and next steps.

Claira reads the metadata of every document in the case (its date, title and fields such as custodian and extension), with your own Discover access, counts it, and writes an overview from those counts. It never reads the content of the documents, and it writes nothing to Discover. Running an assessment is free: it uses no tokens.

When to use it

• At intake, before you decide what to scan: see the volume, the custodians, the file types, the languages and the date spread in one place, with the months where documents spike or thin out.
• To find what to set aside: system files, empty files and very large files are counted separately.
• To plan the review: the suggested segments split the case into sets you can scan, assign or find again in Discover.
• To brief someone: download the assessment as a PDF, with the statistics it was built from.
• After new data arrives: Claira tells you when imports have loaded documents since the last assessment, so you know when a new one is worth running.

Running an assessment

1. Open the Claira menu (the ☰ icon), hover over More at the bottom, and choose Early case assessment. It sits beside Case Context.
2. To count a pick list of your own on top of the standard fields, open Settings above the button (see the next section). Most of the time there is nothing to change here.
3. Click Run assessment. Once the case has an assessment, the button reads Run a new assessment.

Choosing the fields to count

Every assessment counts these standard pick lists when the case has them:

• Custodian
• Native File Extension
• Document Type
• Document Kind
• Languages
• People - From
• People - To
• People - CC

A case that does not have one of them simply skips it. You do not have to choose them.

To count another pick list, such as Issues, open Settings above the button and choose it from Choose a pick list. You can add up to five. Remove one with the × on its chip. The pick lists you add apply to the run you are about to start.

While it runs

Claira shows what it is doing under the button: counting documents, finding fields, looking for a file size field, reading every document's metadata page by page, reading the import history, then writing the assessment. Reading takes about a minute for every 200,000 documents; on a large case, Claira shows how many minutes are left.

While Claira is reading from Discover, you can click Stop. Nothing is saved.

Once Claira is writing the assessment, you can leave the mode: the assessment appears here when it is ready.

Only one assessment of a case can be written at a time. If one is already being written, Claira says so, and it appears here when it is ready.

What the assessment contains

The top line says when the assessment was run and by whom. Under it, four headline figures: Documents, Custodians, Date range and Imports and ingestions. A figure Claira could not read shows None.

Under the figures, The assessment holds what Claira wrote, in a box of its own so it is never confused with the counts below it. Its seven sections appear one by one while Claira writes, each in its place, with the sections still to come shown as grey lines:

The overview is written from the statistics alone. It treats what the case is about as a hypothesis for the review to test, and marks its inferences as such. AI can make mistakes: check the points that matter in Discover.

Data analysis

Below the assessment, Data analysis shows the statistics it was written from. It is collapsed; click it to open it. The statistics were counted over every document when the assessment was run, with the access of the person who ran it:

• File types: a bar splitting the case into User-generated, System-generated and Unclassified documents, the count for each file family, and the most common extensions (see How file types are grouped).
• Timeline: documents per month over the case's core period, the months that hold 99% of the plausibly dated documents. Spikes are highlighted and gaps shaded, the axis under the chart is marked in quarters, years or several years depending on how long the case runs, and three boxes under it give the Spikes, the Gaps and the Busiest days. A spike is a month with at least three times the documents of the median month; a gap is two or more months with under a tenth of it. Claira computes them from the counts, so the overview states them rather than guessing.
• Each counted field: the number of documents for each value, with its share of the case and a bar, and how many documents have No value. The first eight values are shown, and a link such as Show all 12 opens the rest. A line at the end counts any other values.
• Custodians by year, Custodians by file family and File families by year: tables with a row for each of the 15 largest custodians (or each family), the rest summed in one row, and documents without a custodian in No value. Darker cells hold more documents. On a narrow screen, scroll a table sideways.
• Custodian coverage: a bar per custodian with one block per year, darker where that year holds more of their documents and empty where they have none, followed by when their dated documents start and stop and any stretch of three or more empty months in between.
• Who writes to whom: the most frequent pairs from People - From to People - To. On a case with a very large number of distinct pairs, Claira drops the rarest while counting and says the counts are lower bounds.
• File sizes: how many documents fall into each size band, from empty (0 bytes) to 1 GB and over, and the Largest files with their Document IDs, so you can find them in Discover. Claira reads sizes only from a file size field; when the case has none, it says so.
• Document dates: the Range, the Dated documents, how many are Dated before 1990 (often placeholder dates) or Dated after the assessment, and the counts By year.
• Imports and ingestions: how many jobs of each kind the case has had, then each job with its kind (an Ingestion, an Import from a load file, SQL or a database, or Individual documents), the documents it loaded, the duplicates and NIST files it suppressed, and its exceptions.
• Sample of titles: up to 150 titles, taken at even spacing across the documents that have one.

They are there so you can check any figure in the overview against its source, and so an older assessment still makes sense after the case has changed.

How file types are grouped

Claira decides each document's file family from its extension alone, using a fixed list:

System files are what a computer writes rather than a person: programs and libraries, configuration, logs, temporary, cache and backup files, shortcuts and fonts. They count as System-generated, and are often candidates to set aside before review. Every other family counts as User-generated. An extension that is not on the list counts as Other, and a document with no extension as No extension; both count as Unclassified.

Large cases and your access

Every document, every time. Claira reads the metadata of every document, however large the case: nothing is sampled, and every figure covers the whole case. Reading takes about a minute for every 200,000 documents. Only the sample of titles is a sample, spread across the case.

Your Discover access. Every read runs as you. Every figure comes from a Discover search, so it covers the documents you can see: two people with different access to the case can get different figures.

If your access does not include the import history, Claira leaves it out and still runs the assessment. The Imports and ingestions figure then shows None, and the same section under Data analysis says the access used did not include the import history.

Everyone who uses Claira in the case sees the same assessments, so an assessment shows what the person who ran it could see.

Past assessments, the PDF and deleting

Past assessments

Each run is a new, dated assessment; an assessment is never revised. The mode opens on the newest one. When the case has more than one, Past assessments lists them with their date, who ran them and their document count: click one to show it. An assessment still being written is marked Writing, and one that did not finish is marked Failed.

Downloading the PDF

Click Download PDF on a finished assessment. The PDF starts with the case, who ran the assessment, when, and the document count, followed by the seven sections. Data analysis follows on a page of its own, as tables of one width: the file types, the timeline with its spikes and gaps, the counts and shares for each field, the tables by custodian, year and file family, custodian coverage, who writes to whom, file sizes and the largest files with their Document IDs, the document dates, the imports and ingestions with their kinds, and the sample of titles.

The filename follows the pattern Claira-Early-case-assessment-<case>-<YYYYMMDD>.pdf.

Deleting an assessment

Click the bin icon (Delete assessment) beside Download PDF. Claira asks first: the assessment is removed for everyone on the case, with the statistics it was built from. Nothing in Discover changes.

Who can run and delete assessments

When your organization enforces roles in the Admin Portal, running an assessment needs the Run an early case assessment permission, which the User role has by default. Deleting one needs Delete an early case assessment, which it does not.

When the case has changed

When you open the mode and the case has had imports since the latest assessment, Claira says so above the button, for example: One import job has loaded 1,240 documents since the assessment of Sep 21, 2026. Run a new assessment to include them.

If your access does not include the import history, Claira compares the document count instead: when it has changed, Claira tells you how many documents the case holds now against how many it held at the assessment.

What Claira keeps

Claira keeps each assessment, including the statistics it was built from: the counts for each field together with their values (custodian and sender names, for example), the file types, the timeline, the tables by custodian, the sender and recipient pairs, the file sizes with the titles and Document IDs of the largest files, the import history, and the sample of titles. It keeps them until someone deletes the assessment or the case's data is deleted. This is how an Insight keeps the field values it was generated from.

An assessment never holds the text of your documents: it is built from metadata, not from reading documents, and it keeps counts, not a copy of each document's metadata. See Privacy and Security and Deleting a case's data.

Asking the Agent

In Agent Mode, the Agent knows whether this case has an assessment, when it was run and who ran it, and it reads the latest one when you ask about the case: for example, how many documents are in this case?, who are the custodians? or where should I start? It reads the figures (including the file types, the spikes and gaps, custodian coverage, the top sender and recipient pairs and the file sizes) and the overview, not the sample of titles. It tells you the date the assessment was run, because the case may have changed since.

Ask about an earlier one and it lists the assessments this case holds with their dates, so you can have it compare two runs and say what changed. Ask to see one and it offers a button that opens it here, on the run you asked about.

In a case where no assessment has been run, one of the Agent's opening suggestions is Give me a first look at what's in this case, and its answer offers a button that brings you to this screen with Run assessment ready for you. While an assessment is being written it says so, instead of describing a report it cannot see yet.

When the Agent proposes a bulk scan of all documents in the case, the proposal shows the document count from the latest assessment and the date it was counted. See Agent Mode.

The Agent cannot run an assessment itself: its button only brings you here, and the reading happens in your browser with this screen open. To get a new one, press Run assessment.

Claira Private cases

On a Claira Private case, early case assessment never uses one of Claira's models. Where Claira cannot run the assessment on your case's own model, running one is refused with a message saying so, and nothing is saved.

While you have a local model turned on, running an assessment is not available: if you run one, Claira says so and nothing is saved. Past assessments stay available to read, download and delete.

The Early case assessment Insight template

Early case assessment replaces the Insight template of the same name, which wrote a memo from the per-document summaries of a bulk scan. That template is no longer offered in Insights. Insights you already generated from it are unchanged and stay in Past insights.

To write a memo from document summaries yourself, run the Summarize Document prompt as a bulk scan, then generate an Insight over its results with your own prompt.

Where to go next

• Case Context overview: write down the parties, issues and instructions before you scan.
• Bulk Scan: scan a suggested segment.
• Prompt templates: the Quickstart templates for relevance, privilege, objective coding and more.
• Insights and Insight templates: turn scan results into one cited output.