Why modern collaboration platforms require a different discovery methodology—and what Kim v. Cushman & Wakefield tells us about where courts are heading.
There’s a version of eDiscovery that still runs on assumptions built in the early 2000s: that the relevant data lives in email, that people communicate in full sentences, and that the words they use in serious conversations are the same words you’d think to type into a search query. Courts have been patient with that assumption for a long time. They’re becoming less so.
A recent ruling from a California federal court in Kim v. Cushman & Wakefield makes the point clearly. The court found that keyword searches designed for email are inadequate when applied to Microsoft Teams — because Teams messages are shorter, more informal, and less likely to contain the kind of formal identifiers, full names, or precise terminology that keyword protocols are designed to catch. That’s not a technology critique. It’s a structural observation about how humans communicate when they’re not writing an email.
What’s at Stake When the Search Protocol Doesn’t Match the Data
The exposure here isn’t hypothetical. A discovery response that misses relevant Teams communications — because the search terms were calibrated for email behavior — creates real downstream risk. Opposing counsel can challenge the completeness of the production. Judges can question the defensibility of the process. And in a world where sanctions for discovery failures are increasingly tied to the reasonableness of methodology, “we ran our standard keyword list” is not a sufficient answer if the standard keyword list wasn’t designed for the data source it was applied to.
Courts are increasingly skeptical of productions from collaboration platforms that don’t account for the format and context in which the data was originally created. The technical obligation isn’t new — ESI has always included chat data — but the judicial tolerance for generic search protocols applied to non-generic data is eroding, and it’s eroding faster than most discovery workflows have adapted.
The Real Problem Isn’t the Technology — It’s the Mental Model
What Kim v. Cushman & Wakefield surfaces is a failure of framing, not just a failure of execution. Legal teams were trained to think about discovery in terms of documents — formal, authored, stored. Email approximated that model well enough that keyword search became the default assumption. Collaboration platforms break it entirely.
A Teams conversation doesn’t begin with a greeting and end with a signature. It uses abbreviations, reactions, threads that branch mid-context, and references that only make sense to the people in the channel at that moment. Someone who types “did we move forward on that?” in Teams is not going to be found by a search for “project authorization” or “contract execution” — even if that’s exactly what they were discussing.
Courts have increasingly compelled production of surrounding messages for context, not just the individual messages that match a search string — a recognition that the atomic unit of a Teams conversation is the thread, not the line. Running a keyword search across raw Teams data and treating the output as a complete production is, at this point, a position that’s hard to defend.
AI-assisted search can significantly improve culling and prioritization, but collaboration data introduces conversational context, shorthand, and threaded discussions that still require human judgment to assess relevance, privilege, and meaning. That’s not an argument against using AI in eDiscovery. It’s an argument for being precise about what AI does well, and where human judgment has to own the result.
How We Think About This Problem
At LDM Global, we’ve been doing document review across more than 6,100 projects and 210 million documents over 30 years. In that time, the definition of “document” has expanded significantly — and so has the complexity of what it means to run a defensible review.
When a matter involves collaboration data, our approach starts before the first document is touched. Attorney-led review teams define the search strategy with the specific data source in mind — not a generic keyword list applied uniformly across custodians. For Teams and similar platforms, that means accounting for informal language, contextual threading, and the absence of formal identifiers. Our AI-enabled workflows accelerate first-pass classification and surface high-probability responsive content — but every judgment call about relevance, privilege, and context is owned by an experienced attorney, not delegated to an algorithm.
Our teams are certified on leading eDiscovery platforms, enabling us to collect collaboration data using the tools that best fit your environment—not the other way around. We operate under ISO 27001, SOC 2 Type II, and HIPAA certifications, with no work-from-home and client-separated project rooms. When defensibility is the standard, the environment matters as much as the methodology.
The Bottom Line
Kim v. Cushman & Wakefield is a signal, not an outlier. The legal system is catching up to a reality that anyone who’s run a matter involving Teams, Slack, or similar platforms already knows: the data is different, the search logic has to be different, and “we ran keywords” isn’t a methodology — it’s a starting point that requires human expertise to be worth anything.
If you’re re-evaluating how your organization approaches eDiscovery for collaboration data — or if you have a matter where that question is live right now — reach out to sales@ldmglobal.com. We’re happy to walk through what a defensible search and review protocol looks like for the data sources on your next matter.