Syllo

What it is

Syllo is a litigation platform built around agentic document review. Instead of training a classifier on a seed set, as technology-assisted review (TAR) does, Syllo coordinates several large language models that split a review between them and make coding decisions within guidelines the case team writes in plain language. Issue codes are defined in natural language, there is no cap on how many a review carries, and a new issue can be added mid-review without reprocessing the whole set. The vendor says the same agentic system handles responsiveness, issue coding, privilege, confidentiality, and PII/PHI review.

Around the review engine sits the rest of the litigation stack: eDiscovery basics (raw data and load-file ingestion, search, threading, deduplication, redaction, production), case management (exhibits, transcripts in text, audio, and video, pretrial stamping), deposition preparation, chronologies, and a “Litigation Agent” for dialogue-based analysis and drafting against the case record. The vendor reports a SOC 2 Type II attestation (auditor: Prescient Assurance) and regional data-storage options.

Syllo was founded in 2019 by two former federal law clerks: CEO Jeffrey Chivers (Harvard computer science, Sullivan & Cromwell) and COO Theodore Rostow (Yale Law, Simpson Thacher). It ran in stealth for about five years, deployed agentic review with clients in February 2024, and raised $30M of growth funding from Venrock and Two Seas Capital in October 2025. Named customers include Quinn Emanuel, Pillsbury (multi-year agreement, September 2025), Mayer Brown, and Ballard Spahr, which rolled the Litigation Agent out to 400+ litigators firm-wide on 2026-08-05. On 2026-06-29, Kirkland & Ellis signed a multiyear partnership that gives Kirkland exclusive rights, “for an extended period,” to build proprietary AI solutions within and around the platform (Kirkland release, Bloomberg Law, Artificial Lawyer).

Why it shows up in legal-ops stacks

• It targets first-level review, the largest line in a litigation budget. In Syllo’s head-to-head at Royer Cooper Cohen Braunfeld, managed human review cost more than $2.00 per document and hit under 67% recall on a 16,000-document set; Syllo reached 93.4% recall at 69.8% precision on the same set.
• Published validation numbers. The March 2025 white paper, co-written with 25 litigators from seven firms and advised by Carnegie Mellon’s Jamie Callan, reports the last 10 completed responsiveness reviews: average estimated recall 97.8%, median 99.4%, lowest 93.4%; average precision 79.7%, median 85.9%. Those figures sit well above the recall levels courts have accepted for TAR productions.
• Unlimited issue codes without retraining. A TAR 2.0 workflow gets slower and less accurate as issue codes multiply. Syllo’s Ballard Spahr case ran 25+ issue codes over 100,000+ documents with about three hours of setup.

Pricing reality

Syllo does not publish prices. The pricing page, checked 2026-09-30, offers “enterprise basis” deals for in-house teams and “firm wide enterprise partnership” deals for law firms, both behind a demo request. One third-party directory lists a Workspace seat at $350 per month; we could not confirm that figure on any Syllo page, so treat it as a floor for the workspace alone, not for document review.

The comparison that matters is per document. The vendor’s own benchmark puts managed first-level review above $2.00 per document, which makes a 500,000-document production a seven-figure line item before second-level review. Syllo claims clients save 20–30% “and sometimes more,” without stating the baseline. Against that, Relativity has included aiR for Review in RelativityOne under a fixed fee with no metered AI charge since 2025-11-01 — so if you already pay for RelativityOne, the marginal cost of AI review there is close to zero, and Syllo has to beat it on recall and hours, not on the software line.

Best for

A litigation partner or litigation-support director at an AmLaw 200 firm, or an in-house litigation lead at a company with a standing docket, running productions of 50,000 documents or more with many issue codes — antitrust, MDL, merger enforcement, securities, patent, environmental — who wants first-level review done by agents and wants review, depositions, and trial prep in one record. It also fits plaintiff-side employment firms: Syllo partnered with NELA in 2024 to offer access to its 4,000+ members.

Skip it for a one-off matter under 10,000 documents; the enterprise-contract motion costs more than the review saves. Skip it if your firm has standardized on a Relativity or Everlaw review database and only wants AI coding on top.

Versus the alternatives

Relativity and Everlaw are the two largest review platforms. Pick Relativity when hosting partners, vendors, and opposing counsel already run on it and aiR for Review is included in your fee. Pick Everlaw when you want a cloud review database with Storybuilder chronologies and a transparent path from review to trial prep. Pick DISCO or Reveal when a corporate legal department wants a mature per-GB SaaS contract with AI review bundled in. Pick Logikcull for small, self-serve matters.

Pick Syllo when the review is large, issue-heavy, and time-boxed, and the case team would rather validate agent coding with elusion testing than staff a contract-review floor. The fastest-growing entrants in legal AI, Harvey and Legora, are not substitutes: pick them for research, drafting, and document Q&A across practice groups, and keep a review platform for productions. For fact chronologies on a matter without a review database, Mary Technology is the lighter option.

If none fit — a small firm with occasional productions — outsource review to an eDiscovery vendor running Relativity aiR and negotiate a per-document price instead of buying a platform.

Watch-outs

• The recall numbers are vendor-run, not independent. Law-firm partners validated the white-paper figures, and the reviews were zero-shot on Syllo’s own system. Guard: run a pilot on a closed matter with a known responsive set, and have your own team draw the elusion sample and compute recall with a confidence interval before you certify anything to a court.
• Marketing and white-paper precision disagree. The “Choose Syllo” page advertises 99% precision; the white paper’s median is 85.9% and the Royer Cooper head-to-head was 69.8%. Lower precision means more non-responsive documents reach second-level review or production. Guard: price second-level review hours into the business case at the white-paper precision, not the marketing figure.
• Kirkland holds exclusive build rights. Kirkland’s custom solutions built on Syllo are not available to other firms, and a vendor with its largest partner building on top may prioritize that partner’s roadmap. Guard: write roadmap-access, service-level, and data-export commitments into your contract, and ask which features ship to all customers.
• No public API and no MCP server. Syllo lists load-file ingestion and production, but no named integrations with practice-management, DMS, or other review platforms. Guard: test a full export (load files, coding, privilege log) back to Relativity or Everlaw during the pilot so you can leave or co-counsel without rework.
• Defensibility is still being written. Case law on TAR is settled; case law on multi-LLM agentic review is thin. Guard: disclose the methodology in the ESI protocol, agree validation metrics with opposing counsel up front, and keep the elusion-test documentation for any challenge.