By:Jeff Johnson, Chief Innovation Officer, Purpose Legal
For more than twenty years, and for many of us, Early Case Assessment has meant one thing: search terms.
We interview custodians.
We draft keywords.
We run a search term report.
We negotiate whether 75,000 or 150,000 or 1,000,000 documents is reasonable.
We dive into document review.
I am not dismissing search terms. They are familiar. They can be defensible and, in many cases, remain helpful or necessary. Even if this describes your matters, keep reading!
Even so, relying on search terms as the primary engine of our ECA strategy is no longer wise or sufficient. The data has changed. The risk has changed. The possibilities and expectations have changed. Our workflows need to change with them.
The Assumption Problem
Search terms are built on assumptions.
Before we review a single document, we try to predict what relevant material will say. We anticipate terminology. We guess at language. We draft terms based on complaints, interviews, and experience.
Sometimes that works well.
But human language is complex. People abbreviate. They misspell. They use shorthand and code words. They refer to “Cathy” instead of “Katherine Smith.” They say “talk to her about the plan” instead of referencing the formal project name.
Search terms retrieve what we specifically include. They do not infer context. They do not recognize nuance. They do not connect fragmented conversations across platforms.
Search terms are excellent at helping you find what you already know. They are far less effective at helping you discover what you do not.
That distinction matters.
Hit Counts Are Not Intelligence
Too often, ECA negotiations revolve around numbers.
How many hits did this term return?
How many unique hits?
How many documents including families?
Those numbers tell us very little about meaning.
A term may return 50,000 documents. That does not mean those documents are relevant. A narrower term may return 5,000 documents and miss critical communications entirely.
When early case assessment is merely a hit count debate, we delay real strategy. Insight arrives later, during review – often too late.
Modern Data Broke the Old Model
When ECA first emerged, email dominated. Data volumes were smaller. Communication channels were limited.
Today, a single matter may involve:
- Text messages
- Microsoft Teams, Slack, and other collaboration platforms
- Shared cloud documents
- Modern attachments and collaboration threads
People communicate differently across platforms. Tone shifts. Vocabulary shifts. Message length shifts. Context fragments.
All of this creates significant challenges for a keyword strategy. It increases effort, time, and cost to craft an effective keyword strategy, without delivering additional insight or strategic value.
Modern litigation demands an earlier understanding, and modern solutions provide it.
From Filtering to Understanding
This is where the latest Early Case Intelligence workflows change the conversation.
ECI is not just a new, faster approach to data culling (though it can be, with proper triaging and validation of any not relevant classifications).
It is about moving from filtering to understanding.
Modern AI-driven ECI workflows allow us to:
- Analyze the collected data in context of a matter/investigation overview
- Surface key events and key people, grounded in actual documents
- Generate summaries and strategic suggestions tied directly to source materials
- Triage documents by the likelihood they are key or relevant
Instead of asking “How many documents hit this term?”
We can ask, “What story emerges from this data?”
That is a materially different starting point.
Oversight Is Not Optional
Let me be clear about something.
These tools are powerful. They are not perfect.
We use AI to generate analysis input, but we review and improve it. We test. We evaluate results. We adjust. We validate through statistically meaningful sampling.
Human expertise remains central. In fact, the more we rely on AI at scale, the more important human oversight becomes. The quality of the oversight and validation methodology determines whether the output is reliable.
This is not automation in a vacuum. It is AI + experienced legal judgment working together.
Search Terms Still Have a Role
This is not an argument to abandon search terms.
Search terms remain useful. In many matters, they remain required or helpful. In these cases, ECI can inform, accelerate, and improve search term development. It can test proposed terms. It can reveal conceptual gaps.
The difference is that we no longer begin and end our early assessment with keywords alone.
Traditional ECA asked, “What can we filter out?”
Modern ECI asks, “What can we learn right now?”
Given the volume, velocity, and variability of today’s data, that shift is not optional.
It is necessary.
If you would like more information or a demo of our PurposeXi solutions, reach out to our AI experts here.
