The Future of Document Review Is Already Here. What Happens After?

Something changed in legal technology over the past 18 months, and it moved faster than most attorneys expected.

The ABA's 2024 Legal Technology Survey found that 30.2% of attorneys now use AI tools in their practice. That's nearly triple the 11% reported in 2023 (ABA, “2024 Artificial Intelligence TechReport,” published March 2025). Clio's 2024 Legal Trends Report found an even larger number: 79% of legal professionals reported using AI in daily work, up from 19% the year before.

The number one use case? Document review.

Thomson Reuters' 2025 Generative AI in Professional Services Report found that 77% of legal organizations ranked document review as their top anticipated AI use case, followed by legal research at 74% and document summarization at 74%. This is not hypothetical. The platforms have already moved.

Relativity launched aiR for Review using GPT-4o. By October 2025, over 250 customers had used its AI solutions to review more than 25 million documents. Throughput hit 3 million documents per day. Then Relativity made the move that tells you where this is going: at Relativity Fest 2025, they announced that aiR for Review and aiR for Privilege will be included at no additional cost in the standard RelativityOne offering starting in 2026.

Everlaw expanded with Deep Dive, a natural-language query tool running OpenAI's o3 model, cut pricing on Coding Suggestions by 40%, and made several AI features free. Everlaw also became the first e-discovery vendor with FedRAMP-authorized AI features.

DISCO's Cecilia AI now processes roughly 25,000 documents per hour. DISCO announced an all-inclusive platform bundling e-discovery, AI, depositions, and timelines at a single price.

When enterprise platforms start giving away their AI review features, it means the competitive value of AI-assisted review is approaching zero as a differentiator. The intelligence layer, the part that reads documents and classifies them, is being commoditized.

The economics of review were already under pressure

Document review has always been the most expensive part of litigation. The RAND Institute for Civil Justice's 2012 study by Pace and Zakaras, Where the Money Goes (Monograph MG-1208), analyzed 57 large-volume e-discovery productions and found that review consumed 73% of total production costs. The median cost of producing ESI was $1.8 million per case.

The numbers at the per-document level paint a clear picture. The ComplexDiscovery/EDRM Summer 2024 pricing survey found that 44.3% of respondents reported per-document remote review costs in the $0.50 to $1.00 range. The full review process, including initial review, quality control, and production, averages around $2.50 per document. Contract attorneys doing review work bill between $25 and $95 per hour depending on the arrangement, reviewing about 50 documents per hour at first-pass speed.

Compare that to AI-assisted review. The EDRM's Summer 2025 pricing survey found the emerging tier sits at $0.11 to $0.50 per document. That is a 50 to 90% cost reduction.

The accuracy case is just as strong. Grossman and Cormack showed in the Richmond Journal of Law and Technology (2011, XVII RJLT 11) that technology-assisted review can be more effective and more efficient than exhaustive manual review. Their work was cited by the court in Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012), the first federal case to approve technology-assisted review. The earlier Roitblat, Kershaw, and Oot study (2010, 61(1) JASIST 70-80) found that two human review teams looking at the same documents agreed on only 28% of their decisions. Computer systems scored higher than both teams.

AI is not just cheaper. It is often more consistent than people.

What this means for solo and small-firm attorneys

The ABA data shows a pattern worth paying attention to. At firms with 2 to 9 attorneys, 64% prefer general-purpose AI tools over specialized e-discovery software. At firms with 100 or more attorneys, the ratio flips.

Solo and small-firm attorneys are not buying Relativity. They never were. But they are using general-purpose AI to do what Relativity does: read documents, classify relevance, flag privilege, summarize content, and organize case materials. An attorney who uploads 500 case documents to an AI tool and asks it to find every document referencing a specific incident, flag privileged communications, and organize by relevance is doing document review. It is not Relativity's version. But for a practice handling dozens or hundreds of documents instead of hundreds of thousands, it works. And it costs a fraction of the price.

Thomson Reuters projected that generative AI could free up 12 hours per week per legal professional within five years, which works out to roughly $100,000 in additional billable capacity per attorney. For a solo, those hours are the difference between taking on another case or turning it away.

The e-discovery market is projected to roughly double over the next decade, from an estimated $15 to $19 billion today to somewhere between $31 and $46 billion by the early 2030s (Grand View Research; ComplexDiscovery; Fortune Business Insights). Growth is partly driven by data volume and partly by the fact that AI is making review accessible to firms that could never afford it before. Grand View Research specifically projects the SME segment will see the fastest growth rate through 2030.

More attorneys are reviewing more evidence, more efficiently, at lower cost. That is where things are headed.

The question nobody is answering

So the review happens. AI helps the attorney find the relevant documents, flag privilege, organize by issue and timeline. The expensive part just got a lot cheaper.

Now what?

That evidence needs to get to opposing counsel. Or an expert witness. Or co-counsel. Or the court.

How does that actually happen today?

The ABA's 2023 Cloud Computing TechReport found that 70% of attorneys use cloud storage, but fewer than 50% encrypt their file transfers based on earlier ABA survey data. Unencrypted email is still the primary way most firms share files. A 2020 ABA survey found 67% of attorneys used Dropbox for document sharing. ILTA's 2023 survey of over 500 law firms found that while 80% had moved email to the cloud, only 57% had done the same with document management.

None of these tools answer the questions that matter for litigation evidence. Who actually viewed the file? Can you prove it was not altered after you shared it? If the file gets leaked or forwarded to someone unauthorized, can you trace it back to a specific recipient? Is there a chain of custody log that would survive a challenge?

For the vast majority of solo and small-firm attorneys, the answer is no.

The authentication gap is growing

The 2017 amendments to the Federal Rules of Evidence added FRE 902(13) and 902(14), creating a path for self-authenticating electronic records through written certifications and hash value verification, without live witness testimony. The Advisory Committee Notes were direct about the purpose: “the expense and inconvenience of producing a witness” for authentication is “often unnecessary.”

Those rules created a framework for exactly the kind of documentation that evidence sharing should produce automatically. Cryptographic hash verification, access logging, certified declarations. But most attorneys sharing evidence through email or Dropbox have no way to generate any of it.

The authentication picture is also getting more complicated. The Advisory Committee on Evidence Rules considered a proposed Rule 901(c) in November 2024 addressing AI-generated and deepfake evidence. Professor Daniel Linna, Judge Grimm, and Dr. Grossman wrote in “Deepfakes in Court” in the University of Chicago Legal Forum (2025) that “there is no foolproof way today to classify text, audio, video, or images as authentic or AI-generated.”

Louisiana passed HB 178 (effective August 1, 2025), requiring attorneys to exercise “reasonable diligence to verify the authenticity of evidence” and disclose known AI manipulation. As more states follow, the ability to show file integrity from the moment of upload through every access event becomes less of a nice-to-have and more of an obligation.

Courts are not giving attorneys a pass on this. In DR Distributors, LLC v. 21 Century Smoking, Inc. (N.D. Ill.), Judge Johnston issued a 256-page sanctions order and a $2.5 million sanctions award, writing: “It is no longer amateur hour. It is way too late in the day for lawyers to expect to catch a break on e-discovery compliance.” In Industrial Quick Search, Inc. v. Miller, Rosado & Algios, LLP (S.D.N.Y.), the court allowed malpractice claims against a firm that failed to implement a litigation hold. The Maryland Supreme Court disbarred an attorney for discovery failures in Attorney Grievance Commission v. Parris (2023).

The duty of technology competence is now on the books in 40 states plus D.C. and Puerto Rico. ABA Model Rule 1.1, Comment 8 requires attorneys to stay current on “the benefits and risks associated with relevant technology.” ABA Formal Opinion 477R (2017) said unencrypted email is not always sufficient for sensitive information. ABA Formal Opinion 512 (2024), the first ethics guidance on AI, went a step further and suggested “there could come a time when lawyers will have to use generative AI to competently complete certain tasks.”

The bar is rising. The tools most attorneys use to share evidence are not rising with it.

The cybersecurity problem

Sharing evidence through insecure channels is not a theoretical concern. The ABA's 2023 Cybersecurity TechReport found that 29% of firms have experienced a security breach. Another 19% were unsure whether they had been breached. Only 34% have an incident response plan. Only 40% carry cyber liability insurance.

The costs are real. IBM's 2024 Cost of a Data Breach Report put the average breach cost for professional services at $5.08 million. Orrick, Herrington and Sutcliffe's 2023 breach exposed data belonging to over 637,000 individuals and settled for $8 million. Kirkland and Ellis was hit through the MOVEit vulnerability, leading to over 100 lawsuits.

When evidence leaves an attorney's hands through email or a consumer file sharing link, the attorney loses any ability to see who accessed it, whether it was forwarded, and whether the file is still intact. For evidence subject to preservation obligations under FRCP 37(e), that lost visibility creates spoliation risk that did not need to exist.

Infrastructure for the endpoint

AI is handling the review problem. General-purpose AI tools are putting document analysis within reach of solo practitioners. Specialized platforms are building AI into their review workflows and, in some cases, making it free.

None of that solves what happens at the endpoint. The moment reviewed evidence needs to leave the attorney's system and reach another party with documented integrity, verified access, and a paper trail that holds up in court.

That is the gap. Not in review tools, but in sharing infrastructure. What attorneys need during review is intelligence: classification, relevance scoring, privilege detection. What they need after review is different. They need cryptographic integrity verification, viewer identity attribution, access logging, and authentication certificates formatted under the rules that courts actually apply.

However you handle your review, whether you read every page yourself, have a paralegal organize the files, or run documents through AI, the question at the end is the same. Who viewed it? Can you prove it was not altered? If it gets leaked, who is accountable?

That is the problem Attested was built to solve. Every file gets a SHA-256 hash at upload. Every viewer's identity is embedded in the content through watermarking. Every access event is logged with IP, timestamp, and verified email. And every file can generate a certificate formatted under FRE 902(13) with a declaration under penalty of perjury per 28 U.S.C. 1746, built to support self-authentication without live testimony.

Whatever tools you use, the underlying question is the same for every litigator. The review is getting easier. The sharing needs to catch up.