Lighthouse

What it is

Lighthouse is a Seattle-based eDiscovery and legal-data services company. You buy an outcome from it, such as a reviewed production, a privilege log, or an early read on a dataset, rather than a software seat. The company says it has 30 years in the business. New York private-equity firm Lightyear Capital has held a controlling stake since 2019, and Ron Markezich, who spent about 24 years at Microsoft, has been CEO since October 2022. Lighthouse bought H5 in August 2021, which brought in about 250 people and its first document-review practice. That review practice is the base the AI products are built on.

The current product is LighthouseIQ, launched on 21 January 2026. It runs in Lighthouse’s private Azure environment on a data layer called IQ Fabric and has four applications:

• IQ Answers — natural-language questions against a collected dataset before formal review starts.
• IQ Case Strategy — chronologies, timelines, narrative development, and deposition and witness-interview prep.
• IQ Review — predictive AI for responsiveness, generative AI for issue coding, with summaries and rationales for each call. Results land in Relativity, and Lighthouse consultants guide the workflow.
• IQ Priv — privilege prediction plus drafted, editable log lines in standard, categorical, or hybrid formats.

Lighthouse also runs eDiscovery on Relativity, resells Relativity aiR, sells a self-service subscription called Spectra, and offers forensic collection and information governance. Cleary Gottlieb was named as an early LighthouseIQ user. Arnold & Porter, Kirkland & Ellis, and Baker Botts are quoted in a March 2026 release.

The question it answers

The real choice is managed AI review versus a self-serve platform. With Relativity, Everlaw, or DISCO, your team or your firm’s litigation-support group runs the AI and owns the validation. With Lighthouse, Lighthouse runs the models and stands behind the workflow. That matters most when a regulator is on the other side. Lighthouse says DOJ and FTC have accepted its review and privilege output, though it does not name the matters.

Every performance figure Lighthouse publishes is self-reported. They are still the most specific numbers in this segment:

• Microsoft privilege review. Microsoft’s Director of Discovery Programs is quoted on the Lighthouse homepage: a privilege set the team expected to be about 190,000 documents came in at just over 24,000, which Microsoft called a nearly 90% cut in time and cost.
• Privilege accuracy. IQ Priv’s log descriptions were 12% more accurate than contract attorneys’ in blind comparisons. In one 1.5M-document matter it was 6.6x more accurate than privilege search terms.
• Early insight. In one client benchmark, IQ Answers surfaced the same key documents in 3 hours that a continuous-active-learning review found in 3 weeks. Another matter analyzed more than 20M documents in weeks, with an estimated $7.7M avoided in review cost.

None of these numbers has been independently audited. Read them as evidence of what the workflow can do on a well-scoped matter, not as a figure for your contract.

Pricing reality

LighthouseIQ has no list price. What Lighthouse does publish is a rate card for services under its Spectra subscription agreement, labelled 2023 rates. It shows what a services-led eDiscovery bill is made of:

• Data ingestion $39/GB; export for review $175/GB; production $250/GB
• Online hosting $12/GB/month; Relativity users $89/user/month
• Responsive review $0.58/document, with a $97,500 project minimum
• Automated privilege log $3/document; project management $175/hour; forensic collection $295/hour

As an illustration at those rates, a 200 GB matter hosted for six months costs about $14,400 in hosting plus $7,800 in ingestion before anyone reviews a document. A 400,000-document responsive review at $0.58 adds $232,000. LighthouseIQ is pitched as reducing that last line, so ask for the IQ Review and IQ Priv price as a per-document or per-matter number compared directly with these rates. The IQ Answers trial is free, with no upfront commitment, for qualifying organizations.

Best for

In-house litigation and investigations leads at large companies, and the eDiscovery partners at AmLaw firms who support them, on matters measured in millions of documents: HSR second requests, regulatory investigations, and bet-the-company litigation where privilege errors are expensive and deadlines are fixed.

Not for you if your matters are under roughly 100 GB, or your team wants to run its own review. The $97,500 review minimum alone rules out most mid-market matters. A self-serve platform is cheaper there and keeps the work in-house.

Alternatives — and when to pick them instead

• Relativity — the market-share leader in review platforms. Lighthouse itself delivers on it. Pick RelativityOne with aiR directly when you have a litigation-support team that can run and defend its own AI review, and buy services matter by matter.
• Epiq and Consilio — the other large managed-review providers, and the right comparison quotes for a Lighthouse bid. Pick one when you already have a panel relationship and the pricing is comparable. Lighthouse’s edge is a newer AI layer with published privilege results.
• Everlaw — the cloud-native challenger with generative AI built into the platform. Pick it when you want modern self-serve UX and predictable platform pricing across many mid-size matters.
• DISCO — pick it for self-serve review with AI Q&A when your firm prefers software plus optional review services from one vendor.
• Logikcull — pick it for small matters and subpoena response where per-matter self-service beats any managed engagement.

Watch-outs

• Every benchmark is a vendor number. The 90%, 3-hours-versus-3-weeks, and 12% figures come from Lighthouse’s own releases and client testimonials. Guard: require a validation protocol in the SOW with sampled recall and precision on your own matter, reported before production, and make the AI-review fee conditional on meeting it.
• Minimums and per-GB lines dominate small matters. A $97,500 review minimum plus ingestion, export, and hosting per GB can exceed the value of a modest case. Guard: model total cost at the 2023 rate-card levels for your expected volume, and negotiate a hosting cap or a move to nearline storage at a set date.
• You are buying a service, not a system you control. There is no public API or MCP server, and IQ Review results are delivered through Lighthouse consultants into Relativity. Guard: put data export formats, coding history, and the privilege log in editable form into the contract as deliverables at matter close, so a second vendor can pick up the matter.
• AI output under scrutiny. Regulator acceptance of past output does not bind your opposing counsel or your judge. Guard: keep the IQ Priv rationales and QC sample for every logged document, and agree the AI review protocol with the other side or in an ESI order before you rely on it.