Written by: Cassie Lesh
For decades, keyword searches have been one of the most trusted tools in eDiscovery. They are familiar, efficient, and often highly effective when you are looking for something specific. For example: project codes, contract numbers, product references, or other concrete terms.
But the way people communicate has changed. And our searching strategies must change with the times.
Today’s evidence increasingly lives in Microsoft Teams, Slack, text messages, and other collaboration platforms where conversations are shorter, less formal, and heavily dependent on context. A recent federal court decision involving Microsoft Teams recognized exactly why that matters:
In Kim v. Cushman & Wakefield (Wakefield), the court explained:
“Keyword searches alone, without more advanced and thoughtful search techniques, will be inadequate for Teams data-a medium where conversations are shorter, more informal, and less likely to include full names than email.”
That’s a noteworthy statement—but much like a chat, without context it’s also easy to misinterpret.
The court’s statement refers to a traditional keyword strategy—one that required the plaintiff’s full name as an anchor—and ultimately decides such a formal approach is insufficient for modern chat data.
In other words, the problem was not keyword searching. The problem was treating modern chat data like traditional email data.
Same Goal. Very Different Data.
Email usually gives reviewers more structure to work with. Subject lines identify topics. Messages are longer. And participants often repeat key terms, projects, or issues throughout a thread.
Modern chat data is different.
A conversation might unfold across dozens of short messages:
- Ask C
- Different spreadsheet, same 🗑️
- 📲 I’ll ping her
- 👋 Post-J era loading… 🔥
- To the people chatting in real time, they may be perfectly clear. To a review team months later, it’s an ediscovery puzzle.
That is the challenge with modern collaboration data. Important context may be spread across an entire conversation, buried in shorthand, or implied by an emoji. People use first names, nicknames, abbreviations, GIFs, and inside references. They do not always restate the full project name, legal issue, or person involved.
The Court Did Not Prescribe a Tool. It Prescribed a Standard.
One of the most important parts of the Wakefield order is that the court did not tell the defendant exactly how to conduct the supplemental Teams search.
Instead, it said:
“Defendant is free to consult with its eDiscovery consultant on the most efficient and defensible methods for searching, reviewing, and producing Teams data.”
The court then made clear that the defendant could use “custodian-based collection, refined keyword queries, or technology-assisted review,” so long as the search was reasonable and the production complete
That distinction matters because the goal is not to use AI because it is new, or keywords because they are familiar. The goal is to build a search strategy that is reasonable, proportional, and appropriate for the data type.
When Keywords Need Backup.
To the court’s point, the challenge with modern chat data is not always finding the right keyword. Sometimes, it is understanding the tone and context surrounding the conversation.
- That is where Relativity’s broader toolkit becomes especially useful.
- Using the termination issue in Wakefield as an example, aiR Assist could help reviewers ask natural language questions like:
- What conversations discuss transition planning or reassignment of responsibilities?
- Who communicated about the timing of the termination decisions?
- Are there chats discussing workload, coverage, maternity leave, return-to-work issues, or indirect/code-word references?
- Similarly, Relativity Sentiment Analysis can help surface or prioritize conversations that may contain negativity, anger, or other emotionally charged language. For chat-heavy matters, that can be valuable because the most important messages are not always the most formal ones. A conversation may never say “termination decision” or “pregnancy discrimination,” but it may still show frustration, urgency, resentment, relief, or a meaningful shift in tone.
Better Search Has Never Been About One Tool.
The best eDiscovery workflows have never depended on a single method.
The Sedona Conference has long recognized that keyword searching remains valuable but has documented deficiencies, and that alternative search tools may properly supplement keyword and Boolean search techniques.
That is the real lesson of Wakefield.
The court did not say keyword searches are inherently inadequate. It said that a keyword strategy designed for email may be inadequate for Teams data. And that principle should feel familiar to anyone working with Slack, text messages, Google Chat, WhatsApp, or other modern communication sources. The ruling was about Teams, but the broader warning is about data-aware search strategy.
A Consulting Problem, Not Just a Technology Problem.
Perhaps the most interesting part of the Wakefield order is that the court recognized the value experienced eDiscovery teams bring in crafting a more comprehensive search strategy.
That is where Page One comes in. We’re not just any eDiscovery vendor. We pride ourselves in helping clients build workflows that combine the right tools for the right problem: keyword searches, analytics, aiR, validation, QC, and the human judgment needed to make the whole thing defensible.