Privilege review is the most expensive line item in most eDiscovery budgets — attorney rates applied document by document, plus a second manual deliverable in the log. Here is where AI actually removes the cost, and what keeps the result defensible.
Attorney-client privilege review is the most expensive line item in most eDiscovery budgets, and it does not have to be. For a typical 2,000-document privilege pass, firms are paying senior contract attorneys $4 to $8 per document to read for privilege, then another $5 to $15 per entry to build the privilege log — before a single page has been produced. On a 100 GB matter, that combination of privilege review and log generation alone can run $30,000 or more, layered on top of a first-pass responsiveness review that often costs ten times as much.
DecoverAI’s 360° Legal Brain™ was built to attack that specific cost center. By combining contextual AI classification with attorney-directed workflows, it cuts the number of documents a human ever has to read for privilege, automates the log, and does it in a way that holds up when opposing counsel challenges the methodology. Litigation teams that have moved privilege review onto the platform report cost reductions in the 80–98% range on comparable matters. What follows is where those savings come from, mechanism by mechanism, rather than as a headline percentage.
The short version: the dominant cost in a privilege pass is attorney time, not software. Anything that reduces the number of documents an attorney has to open, or the number of log entries they have to draft by hand, moves the total. Everything else is rounding.
Privilege review costs more than ordinary responsiveness review for a simple structural reason: it requires a lawyer rather than a contract reviewer, and it requires a second output — the privilege log — on top of the tagging decision itself. A responsiveness call is binary and can be delegated broadly. A privilege call turns on the role of every participant, the purpose of the communication, and whether the protection has been waived by onward circulation. That judgment carries a licensed-attorney rate, and it is applied one document at a time.
The log compounds it. Every withheld document generates an entry naming the author, the recipients, the date, a description precise enough to be meaningful without disclosing the privileged content, and the basis asserted. That is a drafting exercise rather than a coding decision, and it is billed separately from the review itself.
Industry benchmarks for 2026 put the components at roughly the following:
Scroll table horizontally →
| Cost component | Typical rate | Basis |
|---|---|---|
| First-pass responsiveness review | $0.50–$3.00 per document | Contract attorney, linear or TAR-assisted review |
| Attorney-client privilege review | $4–$8 per document | Senior/licensed attorney rate, higher scrutiny |
| Privilege log entry generation | $5–$15 per entry | Manual drafting of basis, date, author, description |
| Data processing & hosting | $25–$100 per GB | One-time ingestion plus monthly hosting |
| Project management overhead | ~$250/hour | Typically 60–100 hours on a mid-size matter |
Those numbers compound quickly. On a 100 GB matter — roughly 250,000 documents — a traditional review vendor stack covering processing, hosting, responsiveness review, privilege review, privilege log, production and project management lands around $460,000 in total legal review fees, according to a 2026 document review pricing benchmark. Privilege review and logging alone typically account for $30,000 to $40,000 of that, even though privileged documents are usually a small fraction of the overall population.
Zoom out and the pattern holds across the industry. The American Bar Association estimates document review consumes over 80% of total litigation spend — roughly $42 billion a year in the U.S. — and the RAND Institute’s oft-cited 2012 figure of $18,000 per gigabyte to fully review a dataset has not disappeared so much as shifted shape into per-document and per-entry attorney fees. Any benchmark of eDiscovery costs for a matter of real size has to treat privilege as its own line item rather than a rounding error inside general document review pricing. Our breakdown of what a privilege log entry actually costs covers why that per-entry rate has stayed stubborn for two decades.
AI-driven privilege review does not eliminate attorney judgment — courts still require a lawyer to make the final privilege call — but it collapses the number of documents that need that judgment applied manually, and it removes the manual labor of building the log itself. Three mechanisms drive most of the reduction.
Contextual triage is the single biggest lever, because attorney time — not software — is the dominant cost in a privilege pass. The classification step is not keyword matching against a list of attorney names, which is what most privilege screens have historically been. It looks at who is on the thread and in what role, whether the communication was made for the purpose of seeking or giving legal advice, and how comparable documents in the same matter were already coded by counsel. DecoverAI’s platform reports roughly 80% less document volume requiring human review as a result of that step. The documents that survive triage are the genuinely ambiguous ones, which is exactly where attorney judgment is worth paying for. Our note on what the AI classification step itself costs to run puts the machine side of that pass in perspective, and the model selection study for privilege review publishes precision, recall and cost per model.
Automated log drafting attacks the second deliverable. Author, recipients, date, subject and basis for privilege all come out of the same contextual analysis that produced the classification, so the entry is a by-product of the review rather than a separate drafting project. Published estimates put the reduction in privilege-logging labor at around 60%, with per-entry time dropping from roughly 30 minutes to well under 15. The attorney role changes shape rather than disappearing: instead of composing several thousand descriptions from scratch, counsel reviews and corrects drafted entries. We have written separately on automating privilege log creation and on what generative AI does and does not change about log drafting.
The compounding effect is what turns two respectable percentages into a headline number. Modeled scenarios across matter sizes show the pattern consistently: a 250,000-page litigation matter drops from roughly $285,000 to $142,500, a 50% saving; a 1-million-page investigation from $1.15M to $512,000, a 55% saving; a 400,000-page M&A due diligence review from $320,000 to $128,000, a 60% saving. DecoverAI’s own case data goes further on some matters: a 30,000-document set processed in three days at a 98% total cost reduction against the traditional vendor quote for the same population. For how this behaves as volumes climb, see privilege review at scale.
Applied to the matter described above, the arithmetic is straightforward. A traditional stack priced around $460,000 compares to an all-in DecoverAI 360° Legal Brain™ engagement in the neighborhood of $36,000, at a flat $60/GB/month with no seat fees, no per-document surcharges and no separate privilege-log line item. The gap is not primarily a software discount — it is the difference between paying attorney rates on 250,000 documents and paying them on the fraction triage actually surfaces, plus the removal of a per-entry log charge that otherwise scales with how much you withhold.
Cost reduction only matters if the review survives a challenge, and privilege is exactly where courts scrutinise AI use most closely. Two things matter for defensibility in 2026, and neither of them is the model’s accuracy score.
Attorney direction, not autonomous AI. Courts are applying existing privilege doctrine to AI output rather than writing new rules, and the deciding factor is attorney control. Output produced from attorney-crafted prompts and reviewed by counsel is treated as reflecting counsel’s mental impressions and gets stronger work-product protection; unaided AI classification without attorney direction generally does not. 360° Legal Brain™ is built around that distinction — the model surfaces and drafts, a licensed attorney makes the final privilege call, and that human decision point is what a court will look for. Being able to show those decision points is part of what defensible means in practice.
Vendor data handling, not the AI itself. The real waiver exposure in recent case law is not the model’s judgment — it is what happens to the documents after they are uploaded. Courts have flagged the risk that public AI tools may retain submitted data or use it to train other models, which is functionally impossible to unwind once discovered material has been ingested. Any platform used for privileged material needs contractual guarantees against model training on client data, clear deletion capability, and restrictions on downstream disclosure — the same enterprise-security posture, SOC 2 Type II certification and matter-level data isolation, that should be table stakes before privileged documents ever leave your control. DecoverAI’s posture is documented on our security page.
Legal teams evaluating privilege review options today are generally choosing among a few models, and they price the work in structurally different ways.
Scroll table horizontally →
| Approach | How it prices privilege work | What to check |
|---|---|---|
| Per-document managed review | Flat rate per document, commonly $0.75–$0.90, with tiered minimums by volume | Privilege review and log creation are typically add-ons on top of the base rate |
| Enterprise AI privilege engines | Enterprise programme pricing, positioned for large firm and corporate review programmes | Contextual and multi-model analysis with human oversight built in |
| Traditional review platforms | Per-GB hosting or per-seat subscription | Privilege tagging is one feature inside a broader review workflow, not a priced outcome |
| DecoverAI 360° Legal Brain™ | Flat $60/GB/month, all-in | No per-document privilege surcharge and no separate log-drafting fee |
Per-document managed review — the Trustpoint.One model, among others — charges a flat rate per document, commonly $0.75 to $0.90, with tiered minimums by volume. These services are predictable, but they typically price privilege review and log creation as add-ons on top of the base rate, and they require hosting the data on the vendor’s platform, so the headline per-document rate is a floor rather than a total.
Enterprise AI privilege engines such as Consilio’s AI PrivDetect apply contextual and multi-model analysis with organisational learning across matters, paired with human “AI Guide” oversight. That is a similar philosophy to 360° Legal Brain™’s attorney-in-the-loop design, generally positioned for large enterprise and law-firm review programmes.
Traditional review platforms such as Logikcull and similar eDiscovery tools handle privilege tagging as one feature within a broader review workflow, generally billed through per-GB hosting or per-seat pricing rather than a flat, privilege-specific cost model. Privilege work still happens there; it is simply not what the pricing is organised around, which makes the cost of the privilege pass hard to see until the invoice arrives.
Where 360° Legal Brain™ differs is in collapsing all of it — responsiveness, privilege classification, log generation, redaction and production — into a single flat per-GB rate with no per-document privilege surcharge and no separate log-drafting fee. That is what produces the larger total cost reduction against itemised alternatives: fewer units billed, and fewer line items to bill them under.
Privilege review costs are high because they require expensive attorney time applied document by document, plus a second manual deliverable in the privilege log. AI does not remove the attorney from that process — it removes the requirement that the attorney touch every document to get there. For litigation-focused firms and in-house legal departments managing recurring discovery volume, that is the difference between privilege review as a five- or six-figure line item and privilege review as a controlled, defensible, largely automated step in the workflow.
If your team is scoping an upcoming matter and wants a side-by-side estimate of traditional versus AI-driven privilege review costs for your actual document volume, our 360° Legal Brain™ team can run the comparison against your current vendor quote.
The rate ranges in this article are market benchmarks rather than quotes. The per-component figures reflect industry benchmarks for 2026, the $460,000 total for a 100 GB matter comes from a 2026 document review pricing benchmark, the 80%-of-litigation-spend and $42 billion figures are the American Bar Association’s estimates, and the $18,000 per gigabyte figure is the RAND Institute’s 2012 number, cited here because it still shapes how the market talks about review cost rather than as a current price. Treat all of them as directional context for a negotiation, not as audited results.
The savings figures carry the same caveat, and it is a stronger one. The 50%, 55% and 60% scenarios are modeled across matter sizes, not audited engagements. The roughly 60% reduction in privilege-logging labor and the roughly 80% reduction in document volume requiring human review are published estimates and platform-reported figures, and the 80–98% range — including the 30,000-document matter at 98% — is reported by teams that have run comparable matters on the platform against a traditional vendor quote for the same population. Your own numbers will depend on your data, your privilege population and what your current vendor charges.
Why does privilege review cost more than responsiveness review?
Two reasons. It requires a licensed attorney rather than a contract reviewer, and it requires a second deliverable — the privilege log — on top of the tagging decision itself. Industry benchmarks for 2026 put first-pass responsiveness review at roughly $0.50–$3.00 per document, attorney-client privilege review at $4–$8 per document, and privilege log entry generation at a further $5–$15 per entry.
How much of an eDiscovery budget does privilege review consume?
On a 100 GB matter of roughly 250,000 documents, a traditional vendor stack lands around $460,000 in total legal review fees according to a 2026 document review pricing benchmark, and privilege review plus logging typically accounts for $30,000–$40,000 of that — even though privileged documents are usually a small fraction of the population. More broadly, the American Bar Association estimates document review consumes over 80% of total litigation spend, roughly $42 billion a year in the U.S.
Where do AI privilege review savings actually come from?
Three mechanisms. Contextual triage replaces linear reading, so a reviewer is not opening every document in a custodian’s mailbox — DecoverAI’s platform reports roughly 80% less document volume requiring human review after classification. Automated privilege log drafting generates the entry from the same contextual analysis rather than at $5–$15 per entry by hand, with published estimates putting the reduction in logging labor at around 60%. And because responsiveness and privilege review draw from the same reduced pool, the savings compound at scale.
Does AI privilege review remove the attorney from the process?
No. Courts still require a lawyer to make the final privilege call. AI removes the requirement that the attorney touch every document to get there. Attorney direction is also what makes the review defensible: output produced from attorney-crafted prompts and reviewed by counsel is treated as reflecting counsel’s mental impressions and gets stronger work-product protection, while unaided AI classification without attorney direction generally does not.
What is the real waiver risk when using AI on privileged documents?
Recent case law points at vendor data handling rather than the model’s judgment. Courts have flagged the risk that public AI tools may retain submitted data or use it to train other models, which is functionally impossible to unwind once discovered material has been ingested. Any platform used for privileged material needs contractual guarantees against model training on client data, clear deletion capability, restrictions on downstream disclosure, and an enterprise security posture such as SOC 2 Type II with matter-level data isolation.
How does DecoverAI price privilege review and privilege logs?
Flat, at $60/GB/month all-in. AI review, privilege logs, Bates numbering, redactions, semantic search and production sit inside that number, with no seat fees, no per-document privilege surcharge, no separate log-drafting line item, no contracts and no minimums. On the 100 GB example, that is a traditional stack priced around $460,000 against an all-in engagement in the neighborhood of $36,000.
This article is general information about legal technology and discovery practice, not legal advice for any particular matter. Rate ranges and vendor descriptions reflect publicly available information at the date of writing and may change; verify current terms directly with each vendor. Savings figures are modeled scenarios, published estimates and platform-reported results rather than audited outcomes, and DecoverAI pricing reflects published rates subject to change.