In 2006, a senior partner at a major law firm reviewed a litigation budget for a mid-size commercial dispute. The e-discovery line item — the cost of collecting, processing, and reviewing electronic documents — was $180,000. The partner laughed. “That’s more than the entire case was worth ten years ago,” he said. He approved it anyway. He had no choice.
In 2024, that same type of case carried an e-discovery budget of $2.3 million. The partner didn’t laugh. He’d stopped being surprised years earlier.
E-discovery is the fastest-growing cost in civil litigation, and it isn’t close. What began in the early 2000s as a niche specialty — a few tech-savvy lawyers figuring out how to handle email in lawsuits — has become the dominant expense in document-intensive cases, routinely consuming 30 to 50 percent of total litigation spend. In large matters, e-discovery costs can exceed $10 million before a single deposition is taken. The data volumes driving this are almost incomprehensible: a single corporate custodian’s email, chat messages, and files can easily exceed 100 gigabytes. A case with 20 custodians generates terabytes. Reviewing it all, even with AI assistance, costs a fortune.
How we got here
The story starts with the 2006 amendments to the Federal Rules of Civil Procedure, which explicitly addressed electronically stored information for the first time. Before those amendments, e-discovery was the Wild West — some courts ordered broad production, others barely acknowledged email existed. The amendments created a framework, but they also created an expectation: if the data exists, it must be found, preserved, and produced.
What the rulemakers didn’t anticipate was the explosion in data volume that followed. In 2006, corporate communication meant email and shared drives. By 2016, it meant email plus Slack, Teams, text messages, WhatsApp, Salesforce, Jira, Zoom recordings, and dozens of SaaS applications — each a potential source of discoverable information. By 2024, the average enterprise used over 100 SaaS applications, according to product analytics firms, and employees generated data across all of them simultaneously.
The legal obligation expanded with the data. Courts have held that companies must preserve and produce information from personal devices used for business, from ephemeral messaging apps (with serious sanctions for auto-deleting messages after litigation holds issue), and from collaboration platforms that didn’t exist when the rules were written. The SEC’s sweeping enforcement actions against Wall Street banks for off-channel communications — resulting in fines exceeding $2 billion collectively since 2021 — sent an unmistakable message: if your employees are texting about business, those texts are discoverable.
Each new data source multiplies cost. Collecting from Slack isn’t the same as collecting from email — different tools, different formats, different review challenges. A matter involving 15 data sources doesn’t cost 15 times a single-source matter. It costs more, because the complexity is multiplicative.
The review bottleneck
Collection is expensive. Review is ruinous.
Traditional document review — human attorneys reading documents for relevance and privilege — costs $200 to $400 per hour for contract attorneys, or $50 to $150 per hour for managed review operations in lower-cost markets. A case with one million documents, at an average review speed of 50 documents per hour, requires 20,000 hours. At $100 per hour, that’s $2 million. For review alone. Before a single legal argument is made.
Technology-assisted review (TAR) and predictive coding promised to solve this. The idea: train an AI model on a sample of reviewed documents, then let it categorize the rest. Courts have broadly accepted TAR since the landmark Da Silva Moore decision in 2012. In practice, TAR reduces review volumes by 60 to 80 percent in suitable cases — a massive saving.
But TAR has limits. It works best on large, homogeneous document populations. It’s less effective on the short, context-dependent messages — Slack threads, text chains — that now dominate corporate communication. And it requires upfront investment: training the model, validating its accuracy, defending the methodology to opposing counsel. For smaller matters, the overhead of TAR can exceed the savings.
The newest wave — generative AI for document summarization, privilege identification, and issue coding — is showing promise. Early adopters report 40 to 60 percent reductions in review time for certain tasks. But the technology is young, courts are cautious, and the ethical obligations around AI-assisted review are still being defined. The ABA’s 2024 guidance on AI in legal practice emphasized that attorneys remain responsible for work product regardless of the tools used — which means someone still has to verify the AI’s output.
The proportionality fight
The legal system’s primary tool for controlling e-discovery costs is proportionality — the principle, enshrined in Federal Rule 26(b)(1), that discovery must be proportional to the needs of the case. In theory, courts weigh the burden of production against the amount in controversy, the importance of the issues, and the parties’ resources.
In practice, proportionality arguments often fail. Requesting parties ask for everything, citing the possibility that a smoking gun lurks in the unreviewed data. Producing parties object to the cost, citing the burden of reviewing millions of documents for a mid-size dispute. Courts, reluctant to limit discovery before knowing what’s in the data, frequently order broader production than strict proportionality would suggest.
The 2015 amendments to the Federal Rules tried to strengthen proportionality by moving it to the forefront of the scope definition. A decade later, practitioners are divided on whether it worked. Some report that courts are more receptive to cost-based objections. Others say the culture of “collect everything, sort it out later” persists because the sanctions for under-production — adverse inference instructions, case-ending penalties — are so severe that no rational attorney risks them.
This asymmetry is the core economic problem. The cost of over-preservation and over-production is high but predictable. The cost of under-production — sanctions, lost cases, malpractice exposure — is potentially catastrophic. Every rational actor over-collects. The system is designed to produce exactly the cost explosion we’re seeing.
What sophisticated clients are doing
The companies managing e-discovery costs best treat it as an information governance problem, not a litigation problem. The cheapest document to review is the one that doesn’t exist — or more precisely, the one that was properly disposed of under a defensible retention policy before litigation was anticipated.
Microsoft, JPMorgan Chase, and other large enterprises have invested heavily in automated retention and disposition systems. Emails older than a defined period are deleted automatically. Chat messages expire. The goal isn’t to hide evidence — it’s to avoid accumulating petabytes of low-value data that becomes a litigation liability. Courts have generally upheld reasonable, good-faith retention policies, provided they’re suspended promptly when litigation holds issue.
The second strategy is early case assessment done right. Rather than collecting everything and reviewing broadly, sophisticated parties use sampling, data analytics, and phased discovery to identify the highest-value sources first. If the key documents are in the CEO’s email, don’t spend $500,000 reviewing the entire sales team’s Slack history before looking there.
The third is insourcing. Large corporations are building internal e-discovery capabilities — hiring litigation support professionals, licensing review platforms directly, and reducing reliance on outside counsel for routine processing. The Association of Corporate Counsel’s surveys show steady growth in legal operations roles focused specifically on discovery management.
The counterargument: discovery works
For all the complaints about cost, it’s worth remembering what e-discovery replaced. In the paper era, document discovery meant warehousing millions of pages, with review costs that were arguably higher in real terms. Electronic discovery, for all its expense, is faster, more searchable, and more precise than the alternative.
More importantly, e-discovery has exposed misconduct that paper discovery never would have caught. The text messages, the deleted Slack channels, the “off-channel communications” that triggered billions in SEC fines — these are the product of electronic discovery’s reach. A system that couldn’t access digital communications would be blind to how modern business actually operates.
The question isn’t whether e-discovery is worth it in the abstract. It’s whether the marginal millionth document is worth reviewing. And on that question, the current system — driven by fear of sanctions rather than rational cost-benefit analysis — almost certainly overshoots.
Budgeting for the inevitable
For general counsel building litigation budgets in 2026, e-discovery can no longer be a contingency line item. It needs to be modeled from day one, with explicit assumptions about data volumes, custodians, sources, and review methodology. The budget should include a proportionality analysis — not just for the court, but for internal decision-making about settlement versus litigation.
The most important budgeting insight is also the simplest: e-discovery costs are incurred early and irreversibly. Once you’ve collected and processed terabytes of data, that money is spent whether the case settles or goes to trial. This front-loading changes settlement dynamics — defendants who’ve already spent $3 million on e-discovery have less incentive to settle early, because the sunk cost makes trial relatively cheaper. Smart plaintiffs’ attorneys know this and use early, aggressive discovery to force the economic question.
E-discovery didn’t just add a line item to litigation budgets. It changed the game theory of settlement itself. And until the legal system finds a better equilibrium between the duty to produce and the cost of production, that line item will keep growing.
E-discovery is eating litigation budgets — litigationbudgeting.com is available for acquisition by the platform built to control it.
The firms that price litigation well are the firms that budget it well. LitigationBudgeting.com — currently available — is the domain for the discipline that makes AFAs work.