Cost Analysis
A comparison with manual review and TAR, based on a real 20,000-document matter
and why the savings don’t come at the cost of accuracy or defensibility.
Why Document Review Is the Largest Cost Driver in Litigation
In most civil litigation, the single largest line item in discovery is document review, the process by which attorneys examine electronically stored information (email, files, communications) to determine what is relevant to the case, what is privileged, and what must be produced to the other side. On a matter with 20,000 documents, that work would traditionally require a team of contract attorneys reviewing files one by one, generating substantial cost before any case strategy work even begins.
This matter involves 20,000 documents requiring relevance review, 5,000 of those documents requiring deeper case strategy analysis, and 1,000 documents requiring privilege screening and log preparation. Under a traditional manual review model, the total cost across all three workflows is estimated at $38,214. Trustpoint.One is deploying an alternative approach that reduces that figure to $14,875, while improving the accuracy and defensibility of the results.
What the Technology Does
The platform being used is RelativityOne, the industry’s leading document review environment, used by the largest law firms and legal departments in the world. Built into that platform is a suite of AI tools, collectively called Relativity aiR, that performs document analysis using the same large language model technology that underlies tools like ChatGPT, but purpose-built for legal review and operated under a zero-data-retention security agreement with Microsoft.
There are three tools in use on this matter, each handling a distinct phase of the workflow:
aiR for Review
Analyzes all 20,000 documents against a set of legal criteria written in plain English by an AI specialist attorney and assigns each document a relevance determination along with a written rationale and specific citations from the document’s text. It also identifies Key documents within a single pass, which would otherwise require a separate review phase. This process takes hours rather than weeks, applies the same standard to every document without fatigue or drift, and produces recall rates of 85–95%, meaningfully higher than the 60–80% typically achieved through human review alone.
aiR for Case Strategy
Takes the 5,000 documents identified as responsive and extracts facts, builds chronologies, maps witness relationships, and produces draft materials for deposition preparation and case narrative development. Work that would otherwise require senior attorney hours, organizing facts from thousands of documents into a usable case picture, is done automatically, with attorneys reviewing and validating the output rather than building it from scratch. Early adopters report completing this work up to 70% faster than traditional methods.
aiR for Privilege
Screens the 1,000 documents flagged for potential privilege, predicts privilege status using both content analysis and social network mapping (identifying which individuals are attorneys and how they interact across the document set), and drafts the privilege log descriptions required for production. In comparable matters, this tool has saved hundreds of attorney hours and substantially reduced the risk of inadvertent disclosure, including catching privileged documents that human reviewers missed.
What This Means for Cost
The figures below reflect the actual cost model Trustpoint.One applies to this matter, with specific assumptions for each approach.
First-Level Review: Manual vs. TAR vs. aiR for Review
Under manual review, all 20,000 documents are reviewed at a rate of 35 documents per hour. At a bill rate of $40/hr for first-level review, $40/hr for quality control (10% of total hours), and $150/hr for project management (10% of total hours), the first-level review total comes to $33,714.
Under TAR (Technology-Assisted Review), predictive coding reduces the document population by 40%, bringing the review universe to 12,000 documents. The same rate and fee structure applies to that smaller population, producing a first-level review total of $20,228, a meaningful reduction, but still dependent on a full human review team working through a large document set.
Under aiR for Review, the AI analyzes all 20,000 documents directly, no per-document attorney review is required at first pass. The cost is driven instead by: (1) per-document AI processing fees across the review universe plus an additional 1,500 documents used for prompt iteration and validation ($0.25/doc, totaling $5,375); and (2) 20 hours of AI specialist attorney time at $300/hr for prompt development, validation oversight, and QC ($6,000). Total: $11,375.
Privilege Review: Manual vs. TAR vs. aiR for Privilege
All three approaches require some form of privilege review on the same 1,000 potentially privileged documents.
Under manual review and TAR, privilege review is conducted eyes-on at a rate of 10 documents per hour at $45/hr, totaling $4,500 in each case. TAR offers no advantage here because predictive coding does not perform privilege analysis or generate privilege log descriptions.
Under aiR for Privilege, the AI screens all 1,000 documents, predicts privilege status, flags relevant individuals via social network analysis, and drafts privilege log descriptions for attorney review. The cost is: per-document AI processing ($0.50/doc across 1,000 documents = $500) plus 10 hours of AI specialist attorney time at $300/hr ($3,000). Total: $3,500, saving $1,000 against the manual approach while producing more consistent results and a substantially faster log.
Case Strategy: aiR for Case Strategy
The 5,000 responsive documents (25% of the review universe) are run through aiR for Case Strategy. Per-document AI processing applies across those 5,000 documents plus 500 additional documents used for prompt configuration ($0.25/doc = $1,375), plus 5 hours of AI specialist attorney time at $300/hr ($1,500). Total: $2,875.
This work, fact extraction, chronology building, witness mapping, deposition preparation, has no equivalent in either the manual or TAR cost models. Under traditional methods, it would be billed entirely as senior attorney time at rates well above the contract review rates used for first-level review, and is not reflected in those totals.
What Is Not Being Traded Away
A reasonable question is whether these cost reductions reflect a reduction in quality or legal rigor. The answer is no, for three reasons.
First, accuracy improves. Human document review is not the gold standard it is often assumed to be. Studies consistently show recall rates of 60–80% for manual review, meaning a meaningful share of relevant documents are simply missed. The AI tools deployed here achieve 85–95% recall, a higher standard, while applying uniform criteria to every document without the inconsistency that comes from reviewer fatigue, varying interpretations, and team turnover over the life of a project.
Second, defensibility is maintained. The validation process for AI-assisted review follows the same statistical framework courts and regulators have accepted for technology-assisted review for over a decade: elusion rate, recall, and precision are measured against a statistically valid sample at a 95% confidence interval, and results are documented with full audit trails. Every document also carries a written rationale and citation, more documentation than manual review typically produces.
Third, human oversight is preserved. Attorneys make the final determinations. AI provides predictions, rationale, and draft outputs. AI specialist attorneys at Trustpoint.One curate the data sets, develop the criteria, validate the results, and QC the output before anything is acted upon. Technology accelerates work; it does not replace the professional judgment required to do it correctly.
Frequently Asked Questions
What is the role of AI in eDiscovery?
AI helps corporate legal teams manage growing data volumes more efficiently by classifying documents, identifying patterns, generating summaries, and surfacing relevant information earlier in the discovery process.
Why are corporate legal teams adopting AI for eDiscovery?
AI can help legal teams find relevant facts faster, reduce manual review volumes, control discovery costs, and gain earlier insights for case strategy and risk assessment.
Does AI replace lawyers in eDiscovery?
No. AI is intended to augment legal professionals by handling data-intensive tasks and surfacing insights, while legal professionals retain responsibility for strategy, judgment, and final decisions.
How does aiR compare with TAR?
TAR can reduce the document population that requires human review, but still relies on a human review team for the remaining documents. In the article’s 20,000-document model, TAR reduced the review population to 12,000 documents, while aiR analyzed the full population directly at a lower modeled first-level review cost.
What is aiR for Case Strategy used for?
aiR for Case Strategy analyzes responsive documents to extract facts, build chronologies, map witness relationships, and create draft materials that can support deposition preparation and case strategy development.