AI Litigation Support Cost: A Per-Matter Breakdown of Lex Machina, Westlaw Litigation Analytics, Premonition, Everlaw, Relativity aiR, and Trellis (2026)

**Lex Machina** is the LexisNexis-owned analytics engine that built its reputation on federal patent litigation and expanded outward. The product answers questions like 'how does Judge Albright rule on motions to transfer venue in patent cases' and 'what is the median time to summary judgment ruling in EDTX antitrust matters.' It is not a review platform and it does not host your documents — it is a structured database of federal docket events, normalized by lawyers, queryable through dashboards and an API. Per the LexisNexis sales motion documented at https://lexmachina.com/, seats start around $5,000/year and climb past $25,000 for enterprise tiers with all practice-area modules unlocked. There is no public trial because Lex Machina sells to procurement, not to individual lawyers.

**Westlaw Litigation Analytics** is Thomson Reuters' answer, bundled into Westlaw Edge and Westlaw Precision AI subscriptions at https://legal.thomsonreuters.com/en/westlaw. The pitch is simple: if your firm already pays $200–500/seat/month for Westlaw research, the analytics layer is included or near-included, and your associates already live in the Westlaw UI. The coverage is broader than Lex Machina on state courts but historically thinner on the deep federal practice-area normalization (patent, securities, antitrust) that Lex Machina built its moat on. For firms with 200+ Westlaw seats already, the marginal cost of analytics is effectively zero — that is the strategic threat Lex Machina has been managing for a decade.

**Premonition** is the outlier in this comparison. It is not a research analytics platform in the Lex Machina sense — it is an attorney-and-judge win-rate database scraped from PACER and state dockets, available at https://premonition.ai/. The use case is plaintiff-side case selection (which firm has the highest win rate in front of this specific judge for this specific cause of action) and defense-side counsel evaluation. Pricing runs $10K–50K/year depending on API access and seat count, and the product is sold heavily into litigation funders, insurance carriers, and contingency-fee firms underwriting case economics. It is a 'who wins' tool more than a 'what happens' tool.

**Everlaw** at https://www.everlaw.com/ and **Relativity aiR** at https://www.relativity.com/ are the two review platforms in this stack. Everlaw is the modern challenger — built cloud-native on AWS, priced at $3K–12K/seat/year, with EverlawAI for summarization, deposition prep, and document Q&A. Relativity is the incumbent, and aiR is the GenAI module bolted onto RelativityOne with per-document pricing on top of $10–25/GB/month hosting. The two are not directly substitutable: Relativity is the de facto AmLaw 50 standard for complex matters; Everlaw is winning the mid-market and increasingly punching up. For the head-to-head, our Relativity aiR vs Everlaw vs DISCO comparison breaks down the feature and pricing math in detail.

**Trellis** at https://trellis.law/pricing/ is the cheapest tool in this lineup and the most interesting strategic story. State-court dockets are a mess — every county runs its own e-filing system, formats are inconsistent, and Lex Machina historically didn't bother because the unit economics don't work at enterprise pricing. Trellis built the ingestion pipeline, normalized state-court data across all 50 states, and prices Insights at $99–499/seat/month. For a state-court litigator, insurance defense practice, or any firm whose matters never see the inside of a federal courthouse, Trellis is a near-instant ROI. For AmLaw federal litigators, it is a complement, not a replacement.

The stack matters because the per-matter cost depends on which tools you're running in parallel. A typical mid-sized federal litigation matter at an AmLaw 200 firm looks like this: case intake and judge research in **Westlaw Litigation Analytics** or **Lex Machina** (analytics seats already paid for), document collection and review in **Everlaw** or **Relativity** (per-matter hosting and seat fees), deposition prep and motion strategy back in the analytics tool, and final case-economics validation in **Premonition** if the firm underwrites contingency exposure. The analytics tools are seat-priced and amortize across matters; the review platforms are usage-priced and bill per matter.

**Lex Machina** integrates with Lexis+ AI and into the broader LexisNexis legal research workflow, which matters because the alternative is asking associates to switch tabs between research and analytics. The API documented at https://lexmachina.com/ supports outbound feeds to firm knowledge management systems, which AmLaw firms use to keep deal/matter databases warm. **Westlaw Litigation Analytics** has the equivalent integration story into Westlaw Edge and Precision AI, including new GenAI summarization features that Thomson Reuters has been aggressively shipping in 2025–2026.

**Everlaw** and **Relativity** integrate at the data layer with Microsoft 365, Google Workspace, Slack, and the major archiving systems (Microsoft Purview, Mimecast, Smarsh) for source collection. EverlawAI and aiR both ship REST APIs for connecting to firm workflow tools. The architectural difference is that Everlaw is single-tenant SaaS only — you cannot self-host — while Relativity Server (the legacy on-prem product) is still licensed for firms with data-residency constraints that even RelativityOne's regional clouds don't solve. The on-prem Relativity license is six figures before you've ingested a single document and is a different conversation from RelativityOne pricing entirely.

**Premonition** sits sideways to the workflow. It's a research-and-underwriting tool, not a production tool, and the integration story is mostly export-to-spreadsheet or API call from a case-intake form. Trellis fits a similar pattern at the state-court level — pull the judge profile and motion-outcome data into the case file at intake, then revisit before filing or settlement decisions. Neither tool is in the daily-driver workflow for most litigators; they're consulted at high-leverage decision points.

The practical integration question for procurement: does your firm want a single-vendor strategy (Westlaw + Relativity = Thomson Reuters ecosystem, or Lex Machina + Everlaw = the Lexis-adjacent answer) or a best-of-breed stack? Single-vendor gets you volume discounts and one throat to choke; best-of-breed gets you Trellis state-court coverage that the incumbents don't match and Premonition underwriting data the analytics platforms don't surface. There's no universally right answer — it depends on your matter mix.

The analytics seats are the easy math. **Lex Machina** at $5K–25K/seat/year and **Westlaw Litigation Analytics** at ~$200–500/seat/month bundled with Westlaw Edge are amortized across every matter a litigator touches — call it $40–200/month of analytics cost per active matter for a litigator running 5–20 simultaneous matters. **Premonition** at $10K–50K/year per seat lands in the same per-matter zone for firms that use it as a default underwriting check. **Trellis** at $99–499/seat/month is the cheapest of the four and the lowest per-matter incremental cost. These costs as of June 2026 — verify at lexmachina.com/pricing and the equivalent vendor pricing pages before signing.

Review platform pricing is where things get interesting and where the per-matter delta blows up. **Everlaw** seats are $3K–12K/seat/year, and a typical matter team might be 3–10 seats — call it $10K–120K of seat cost annualized, plus per-GB hosting on top. **Relativity aiR** doesn't bill by seat the same way. RelativityOne is hosting-priced at $10–25/GB/month per https://www.relativity.com/, which means a 500 GB matter costs $5K–12.5K/month just to sit there, before processing fees and before aiR per-document AI fees layered on top. For a 12-month matter with 500 GB, you're at $60K–150K in hosting alone.

The aiR add-on is the new line item that catches buyers off guard. Relativity prices aiR for Review (binary coding via GenAI), aiR for Privilege, and aiR for Case Strategy on a per-document basis. Public benchmarks from RelativityFest 2025 sessions put it in the $0.05–0.25/document range depending on volume commits, which on a 2-million-document matter is $100K–500K of aiR spend on top of hosting and seat costs. EverlawAI is more often bundled into enterprise seats or sold as a flat per-firm add-on at $1K–3K/user/year — different commercial model, often cheaper at modest volumes, sometimes more expensive at very large ones.

The hidden-cost line is processing, which is the step between ingestion and review. Both Everlaw and Relativity charge processing fees — Relativity historically at $20–50/GB processed, Everlaw bundled more aggressively into the per-seat fee at lower tiers. For a matter where collection is 2 TB and the responsive set after culling is 200 GB, processing fees alone can be $40K–100K depending on vendor and contract terms. This is where the 'per-matter cost' conversation really happens, and where pre-deal volume commits drive the actual unit economics.

Add it up for a single mid-sized federal matter, 6-month duration, 500 GB hosted, 5-person review team: analytics layer ~$2K, **Everlaw** path roughly $50K–80K (seats + hosting + processing), **Relativity aiR** path roughly $120K–250K (hosting + processing + aiR fees). The 2–4x delta is real, and it's why mid-market firms keep migrating to Everlaw and why Relativity keeps its hold on AmLaw 50 work where matter complexity justifies the spend. Premonition and Trellis don't materially move that per-matter number — they're decision-support tools, not workflow tools.

A federal patent infringement case in EDTX or NDCA: **Lex Machina** is non-negotiable for judge analytics, **Westlaw Litigation Analytics** is duplicative if you also have Lex Machina, **Relativity aiR** for review if document volume is north of 250 GB, **Everlaw** if the matter is leaner. **Premonition** matters if you're contingency-fee or running on litigation funding economics. **Trellis** is irrelevant for federal patent work because state courts don't hear those cases.

Insurance defense practice handling state-court personal injury, premises liability, and similar matters: **Trellis** is the right primary analytics tool at $99–499/seat/month per https://trellis.law/pricing/ because the matters live in state court where Lex Machina coverage is thin. **Everlaw** for review on the meatier cases. **Westlaw Litigation Analytics** if you already pay for Westlaw — which most insurance defense firms do. **Lex Machina** is overkill for this practice profile. **Relativity aiR** rarely justifies the per-matter cost on individual insurance defense files.

Plaintiffs' contingency-fee firm doing case selection on mass torts: **Premonition** earns its $10K–50K/year fast because the case-economics underwriting question (who's the judge, who's the defense counsel, what's the realistic win probability) is the entire ROI question. **Lex Machina** for federal MDL strategy. **Everlaw** is the default review platform for plaintiffs' work because of pricing and modern UI. **Westlaw Litigation Analytics** if the firm is also a Westlaw shop. **Relativity** is rare in plaintiffs' practice; **Trellis** is useful for the state-court companion litigation.

Regulatory investigation (DOJ, SEC, FTC) at AmLaw 50 firm: **Relativity aiR** is the standard because government productions and second requests are still routinely 1+ TB and the workflow tooling matters at that scale. **Westlaw Litigation Analytics** and **Lex Machina** both get used for any litigation that spins out of the investigation. **Everlaw** is increasingly competitive at this tier but still loses RFPs against Relativity on the very largest matters. **Premonition** and **Trellis** don't fit this matter type.

Mid-market commercial litigation at a 100-lawyer firm: this is where the question gets hardest. **Everlaw** + **Westlaw Litigation Analytics** (assuming firm-wide Westlaw subscription) is the modal answer at the lowest total cost. Adding **Lex Machina** is justified if the practice mix includes patent, antitrust, or securities work where Lex Machina's normalization beats Westlaw's. **Trellis** is the cheap state-court add-on. **Relativity** and **Premonition** are situational — Relativity for the rare large matter, Premonition only if the firm does contingency or litigation finance work.

**Relativity** leads on certifications, with SOC 2 Type II, ISO 27001, ISO 27017, ISO 27018, FedRAMP Moderate authorization, and HIPAA alignment documented at https://www.relativity.com/. RelativityOne offers regional hosting in the US, EU, UK, Australia, Canada, and Japan, which is the deepest data-residency footprint in this comparison and the main reason Relativity holds the big regulatory and cross-border investigation work. The Relativity Server on-prem option remains licensable for firms whose CISOs reject any cloud answer, though it carries a separate six-figure license and significant ops overhead.

**Everlaw** is SOC 2 Type II and ISO 27001 certified, with FedRAMP Moderate authorization achieved in 2023 and hosting regions across US, EU, AU, and CA documented at https://www.everlaw.com/. The architecture is multi-tenant AWS with customer-managed encryption keys available on enterprise tiers. For most US-based commercial litigation, this is more than sufficient; for matters with EU GDPR exposure or Australian privacy law requirements, the regional hosting matters and is part of why Everlaw has been winning mid-market RFPs that previously went to Relativity by default.

**Lex Machina** and **Westlaw Litigation Analytics** are read-only research tools and don't host client documents, which dramatically simplifies the security review. The data flowing into both platforms is public docket information; the data flowing out is research queries. SOC 2 Type II is standard for both, and SAML SSO is standard at enterprise tiers. Privacy concern centers on query logs (which judges are you researching, which attorneys are you scoping) and both vendors offer enterprise contracts that lock down query log retention.

**Premonition** sits in the same category as the research analytics tools from a CISO perspective — public docket data in, queries out, no client documents hosted. The product is sold predominantly into the US market and US data residency is the default. SOC 2 reports are available under NDA per the standard Premonition enterprise sales motion. **Trellis** is also SOC 2 Type II per https://trellis.law/, US-only hosting, and SSO available on Pro tier and above.

The data-residency wildcard is cross-border matters. If your firm is regularly doing UK Competition Appeal Tribunal work, EU regulatory investigations, or Australian Federal Court matters with privacy obligations attached, **Relativity** is the only vendor in this comparison whose regional hosting story is mature enough to satisfy most outside-counsel guidelines from Fortune 500 clients with strict data-handling addenda. Everlaw is closing this gap fast in 2026; the others aren't trying to compete in this market segment.

Trap one: 'starting at' pricing on **Lex Machina** and **Premonition** is almost never what your firm will pay. Both vendors structure pricing in modules — Lex Machina by practice area (patent, antitrust, securities, employment, etc.) and Premonition by API access and seat counts. The published $5,000 starting seat for Lex Machina is a single practice area, single seat; the realistic mid-firm deployment is multi-practice, multi-seat, and lands in the $15K–25K per-seat range at scale. Get the multi-year pricing in writing before you sign year one.

Trap two: **Relativity** hosting overage is brutal. A matter that ingests 500 GB and ends up with a 250 GB review set still pays $10–25/GB/month on the full 500 GB until and unless you actively cull the index. Firms get caught at end-of-matter when the data is no longer being actively reviewed but is sitting on hosting bills for compliance retention. Negotiate a 'cold storage' or 'archive' tier into the contract upfront — Relativity will offer it, but only if asked.

Trap three: **Everlaw**'s mid-tier pricing is genuinely competitive but the per-GB hosting overage on the small-firm tier is not. The $3K/seat/year entry point includes a hosting allocation, and exceeding it triggers per-GB fees that erode the price advantage versus Relativity. For any firm that anticipates 100+ GB matters as a regular pattern, the mid or enterprise Everlaw tier is the right purchase and the small-firm tier is a trap.

Trap four: **Westlaw Litigation Analytics** is 'included' in Edge and Precision AI, but the bundle pricing has crept up substantially over the past three years. The marginal cost of analytics is low only if you were going to buy Westlaw anyway. If you're a Lexis shop considering switching to Westlaw to get the analytics 'free,' run the math on full Westlaw subscription cost minus your current Lexis cost — it is often more expensive than just buying **Lex Machina** as a standalone analytics seat.

Trap five: **aiR** per-document pricing on Relativity. The pricing pages list ranges, the contract lists committed minimums, and the actual bill at end of matter includes overage charges if your aiR usage exceeded the commit. For a firm that's doing aiR-heavy review (and that's the point of buying it), the committed minimum is often unrealistically low and the overage rate is materially higher than the committed rate. Either commit at realistic volumes from day one or accept that you'll pay overage rates — the marketing-grade pricing on the website doesn't reflect this. The published numbers as of June 2026 — verify at relativity.com/pricing — should be treated as a starting point for negotiation, not a final answer.

The headline of 2025–2026 in litigation tech has been GenAI features shipping at every layer of the stack. **Relativity aiR** moved from a beta module to a production product line with aiR for Review (first-pass binary coding), aiR for Privilege (privilege identification with rationale), and aiR for Case Strategy (a research-and-drafting copilot that overlaps with Westlaw and Lexis features). **EverlawAI** has expanded from summarization into deposition prep, document Q&A across matter corpora, and chronology building. The features genuinely change the per-matter cost equation — but only if the matter is large enough that the AI usage offsets associate hours.

**Westlaw Precision AI** and **Lexis+ AI** (the research-side AI products from the same parent companies as **Westlaw Litigation Analytics** and **Lex Machina**) have continued to mature. The integrations between research AI and litigation analytics are still loose — you can't yet ask Westlaw Precision AI 'how often does this judge grant motions to dismiss in cases like mine' and get a single answer. Both vendors are roadmapping this integration; neither has shipped it as a production feature as of June 2026.

The GenAI math that actually matters for procurement: aiR-driven binary review at $0.05–0.25/document replaces contract-attorney review at $40–60/hour. For a 1 million-document matter where contract review would have cost $400K–800K, aiR at $50K–250K is a clean win. For a 100K-document matter where contract review would have cost $40K–80K, aiR at $5K–25K is also a win, but a smaller one. Below that threshold, the GenAI tooling cost can exceed the human cost it's replacing.

**Trellis** has shipped GenAI brief summarization and judge tendency prediction features that compete with the lower end of Lex Machina's product on state-court matters. For firms that don't have federal-court analytics needs, the Trellis GenAI features at $99–499/seat/month genuinely substitute for $5K–25K Lex Machina seats. This is the most interesting strategic shift in 2026 — Trellis is no longer just the cheap state-court alternative; it's a credible analytics tool at one-tenth the price for the right practice profile.

**Premonition** has been slower on the public GenAI features front, focused instead on data quality and coverage expansion. The Premonition value proposition (attorney win-rate underwriting) is data-quality-bound more than AI-bound, so the vendor's choice to invest in coverage over chatbot features is defensible. The question is whether that holds in 2027–2028 when GenAI-driven case-outcome prediction becomes table stakes — at which point Premonition's data moat becomes more valuable, not less.

Build the TCO model in three layers. Layer one is fixed annual seat costs: analytics seats (**Lex Machina** at $5K–25K, **Westlaw** at $2.4K–6K, **Premonition** at $10K–50K, **Trellis** at $1.2K–6K) times the number of seats your firm needs. For a 50-lawyer litigation firm, this layer might run $50K–500K annually depending on which combinations you select.

Layer two is variable per-matter cost. Estimate matter mix in three buckets: small matters (under 50 GB review data), mid matters (50–500 GB), and large matters (500+ GB). For each bucket, calculate **Everlaw** versus **Relativity aiR** cost given seat counts, hosting volume, processing, and aiR usage. The output is a per-matter cost per bucket and an annual matter-mix volume. Multiplied together, this is the variable spend that dominates total TCO for active litigation practices.

Layer three is the AI usage line item, which is the new and least predictable component. Build sensitivity around aiR usage scenarios: low usage (privilege screening only), medium usage (privilege plus first-pass review on responsive docs), high usage (full aiR-driven review with human QC). The high-usage scenario can add 50–150% to total review cost in absolute dollars while reducing contract-attorney hours by 70–90%. Whether that math works depends on what your associate billing rates and contract-attorney rates look like.

Run the TCO twice — once with your current matter mix and once with the matter mix you expect in 24 months. Litigation practice mixes shift, and the tooling decision you make today should accommodate a reasonable forecast of where the practice is heading. AmLaw firms expanding into regulatory investigations need to weight Relativity heavier; firms expanding into state-court class actions need to weight Trellis heavier; firms expanding into contingency-fee work need to weight Premonition heavier.

Pressure-test the analysis with two scenario questions. First: what happens to your tooling cost if matter volume doubles next year? Per-seat tools scale linearly with seats; per-GB tools scale linearly with data; aiR scales with documents reviewed. The mix matters. Second: what happens if matter volume halves next year? Are you stuck on annual seat commits that don't right-size? Both Everlaw and Relativity offer flex hosting models — get them in writing before signing the multi-year deal.