A working cost comparison for a typical 100 GB commercial matter — traditional per-GB/per-document pricing stacked against AI-augmented all-inclusive pricing — anchored to real 2026 market rates.
The matter parameters used throughout this analysis are drawn from a typical mid-size commercial dispute: 100 GB of collected data, approximately 250,000 documents after deduplication and near-deduplication, a six-month active review window, 15,000 documents identified as responsive, and 2,000 documents requiring privilege review and privilege log entries. These numbers are not aggressive. They are conservative mid-market parameters for a commercial breach of contract or trade secret matter in a US federal court.
All dollar figures are mid-market US rates as of early 2026. The traditional vendor column reflects what a mid-market AmLaw 100 firm working with a national eDiscovery vendor would actually receive in a quote for this matter today — not a worst-case scenario constructed to make the incumbent model look bad, but the rates that appear in standard vendor engagements: $100/GB for processing, $25/GB/month for hosting, $1.50/document for first-pass review. The AI-augmented column reflects DecoverAI's published all-in rate of $60/GB/month with no per-document fees, no per-entry privilege log charges, and no processing surcharges. The methodology is the benchmark. The numbers are the findings.
The table below stacks every material line item for this 100 GB matter across both pricing models. The traditional vendor column is built from publicly available rates and standard engagement terms; the AI-augmented column reflects DecoverAI's all-in per-GB pricing with every downstream service included. Review the line items carefully — the aggregate is less surprising once you understand where the money is going.
| Line Item | Traditional Vendor | AI-Augmented |
|---|---|---|
| Processing & ingestion (100 GB) | $10,000 ($100/GB one-time) | Included |
| Hosting (100 GB × 6 months) | $15,000 ($25/GB/mo) | $36,000 ($60/GB/mo, all-in) |
| First-pass responsiveness review (250,000 docs) | $375,000 ($1.50/doc, contract review) | Included (AI classification + human QC) |
| Privilege review (2,000 docs) | $12,000 ($6/doc) | Included |
| Privilege log generation (2,000 entries) | $20,000 ($10/entry) | Included |
| Production (Bates, TIFF, load files) | $8,000 | Included |
| Project management (~80 hrs) | $20,000 ($250/hr) | Optional ($150/hr LegalOps) |
| Total all-in | $460,000 | $36,000 |
The traditional-vendor column is not a strawman. It is what a mid-market AmLaw 100 firm working with a national eDiscovery vendor would be quoted for this matter today. The line items are standard. The per-document rate for contract review ($1.50/document) is actually on the lower end of current market rates, which run $1.00–$3.00 depending on the review platform, the contract review provider, and whether the documents are in English. The project management rate ($250/hour) is at the low end for a senior litigation support professional at a national vendor. If anything, this benchmark is conservative in the direction that flatters the traditional model.
KEY FINDING: On this benchmark matter, 92% of the cost differential between the two pricing models comes from a single line item: per-document responsiveness review. AI does not save money on the gigabytes — it saves money on the human hours that the per-document model exists to bill against.
The arithmetic is not subtle. The $424,000 cost differential between the two pricing models — $460,000 versus $36,000 — is driven almost entirely by one line item: first-pass responsiveness review. The $375,000 figure in the traditional vendor column represents 250,000 documents at $1.50 each. That single charge is 81% of the total traditional vendor spend and accounts for approximately 88 cents of every dollar of the cost differential. Add in privilege review and privilege log generation — both priced per-document and per-entry respectively — and the three document-count-indexed line items together account for over 92% of the gap.
AI does not save money on the gigabytes. The per-GB hosting rate in the AI-augmented model ($60/GB/month all-in) is actually higher than the traditional vendor's hosting-only rate ($25/GB/month). What AI eliminates is the labor-indexed billing that the per-document model is specifically designed to capture. The responsiveness pass that took a contract review team six weeks in 2015 — clicking through TIFF images in a hosted Concordance database, tagging documents one by one — can be performed in an afternoon by a multi-model GenAI classifier operating on extracted text. The classifier does not charge per document. It charges for compute, and the compute cost per document is a fraction of a cent at current cloud rates.
The implication is precise: traditional eDiscovery pricing is not a fair proxy for the cost of data management. It is a fair proxy for the cost of human document review. When AI replaces the human review pass, the traditional pricing model does not get cheaper — it gets exposed as structurally disconnected from the actual work being performed.
Several significant cost categories are excluded from this benchmark because they are not readily comparable across pricing models, and including them would obscure rather than clarify the analysis. Outside counsel time to scope, direct, and quality-check the review is excluded. Partner-level privilege adjudication on close calls — the documents where the GenAI classifier returns a confidence score that requires attorney review — is excluded. Deposition preparation and trial preparation costs are excluded entirely, as they are downstream of the review itself and incurred regardless of which eDiscovery platform was used.
Per-user platform costs are also excluded from the main comparison, but they deserve a separate note. Traditional eDiscovery platforms often charge per named user, per month, in addition to per-GB hosting and per-document review fees. Those seat fees can add $200–$800 per user per month for a mid-size review team, and they are additive to every line item in the traditional vendor column above. DecoverAI has no seat fees and no user limits; all users — reviewing attorneys, litigation support, co-counsel, client contacts — are included in the per-GB rate. On a matter with 15 active review users over six months, the seat fee differential alone can add $18,000–$72,000 to the traditional vendor total, which would push the gap above $500,000 for the same benchmark parameters.
None of these exclusions change the order-of-magnitude conclusion. The benchmark is $460,000 versus $36,000 before seat fees and outside counsel direction time. Adding those costs does not close the gap — it widens it.
A $460,000 review estimate on a $1.2 million dispute is no longer just expensive. Under the post-2015 FRCP 26(b)(1) proportionality amendment, it is potentially indefensible as a matter of law. The 2015 amendments moved proportionality from a backstop objection to a primary constraint on the scope of permissible discovery. The six-factor test — importance of the issues, amount in controversy, relative access to information, the parties' resources, the importance of the discovery in resolving the issues, and the burden-versus-benefit ratio — must be applied before discovery is demanded, not after the producing party resists.
The proportionality arithmetic on this benchmark matter is stark. A $460,000 review estimate represents 38% of a $1.2 million amount in controversy. Courts applying the amended Rule 26(b)(1) have been increasingly skeptical of discovery costs that approach or exceed a significant fraction of the amount in controversy, particularly when the producing party can demonstrate that a less expensive methodology is available. The question to ask opposing counsel — and the question to ask your own vendor — is not "how do we cut costs?" It is: "how would this expense survive a Rule 26(b)(1) proportionality challenge from the other side?"
An AI-augmented all-in cost of $36,000 represents 3% of the same $1.2 million amount in controversy. That is a cost structure that is proportionate on its face, that does not require extended judicial scrutiny, and that gives the producing party significant room to offer broader scope without triggering the proportionality objection that a $460,000 estimate would invite. For a deeper treatment of the Rule 26(b)(1) framework, the six-factor analysis, and the burden-shift the 2015 amendments created, see Proportionality in eDiscovery: How Courts Are Redefining "Reasonable".