Lazarus AI
by Lazarus AI
On-premises enterprise AI for insurers and government: document intelligence, code security and COBOL modernization
Lazarus AI sells AI products that run on infrastructure the customer controls, on premises, air-gapped or in a sovereign cloud. It is built for insurers, government agencies and other regulated organizations that need document intelligence, vulnerability hunting and legacy-code modernization without sending data to a pay-per-token model provider.
Lazarus AI is a Boston company founded in 2017 and led by CEO Alex Panait. It sells a set of AI products that run on open-weight, switchable models inside the customer's own environment: a standalone on-premises appliance the company calls the Lazarus box, the customer's own cloud, an in-country deployment for governments or a fully air-gapped system for defense. The Applied Intelligence System covers document intelligence (the Applied Intelligence Engine, which replaced the earlier RikAI product), orchestration (ATLS) and agentic automation (ReCursor); Lazarus says it cites each finding to its source page and improved field-level accuracy by 12.5 points over standalone frontier models across 150 customer use cases. Clearwing finds vulnerabilities in source code, reproduces them, grades the evidence on six levels and hands engineers patch files instead of writing to repositories. It was released as an open alternative to Anthropic's Glasswing in April 2026. Constructor converts COBOL, JCL, CICS, VSAM and Db2 systems to Java and checks the rewrite against a built-in COBOL emulator; it is in pilots with two large mainframe makers. Scribe (spreadsheet and VBA mapping) and Hexis (model and agent governance) are slated for Q1 2027. The company charges no per-token fees and claims a modeled 78% cut in annual operating cost against pay-per-use frontier models. In November 2026 it announced $43M from investors including QBE, State Farm Ventures, Mucker Capital and Techstars.
A CIO or CTO at an insurer, reinsurer or government agency whose data cannot leave its own perimeter and who wants AI on documents, code or mainframe systems without per-token billing.
AI document, security and modernization workloads that run behind the firewall on swappable open-weight models, with the customer approving every model change.
At a Glance
- Category
- Industry & Government
- Pricing
- Contact for pricing
- Target Market
- CIOs, CTOs, Insurance Executives, Government IT Leaders
- Deployment
- Self-hosted, Hybrid, Edge-first
- Founded
- 2017
- Headquarters
- Boston, USA
Key Features
- ✓Applied Intelligence Engine
Document intelligence that extracts fields from complex and handwritten documents and cites each finding back to its source page.
- ✓Clearwing
Finds source-code vulnerabilities, reproduces them, grades evidence on six levels and delivers fixes as patch files engineers choose to apply.
- ✓Constructor
Converts COBOL, JCL, CICS, VSAM and Db2 code to Java and verifies equivalence with a built-in COBOL emulator.
- ✓Customer-controlled deployment
Runs on an on-premises appliance, the customer's cloud, in-country or fully air-gapped, so sensitive data stays inside the perimeter.
- ✓Swappable open-weight models
Models can be replaced without rebuilding the system, and customers approve each model change themselves.
- ✓No per-token pricing
Removes usage metering from the cost model, which matters most for high-volume document and code workloads.
Capabilities
Use Cases
- •Reinsurance M&A due diligence
A large reinsurer used the Applied Intelligence System during an M&A review and surfaced over $30M in under-reserved liabilities.
- •Prior authorization triage
A digital health provider reports a 92% cut in response time for urgent prior-authorization requests using Lazarus document intelligence.
- •Mainframe modernization
Constructor is being applied to a 4-million-line production COBOL system to produce verified Java equivalents.
- •Military talent matching
US Army Talent Management reports role-to-person matching about 1,000 times faster than its earlier process.
Ideal For
Best For
- ✓Insurers and reinsurers extracting and reviewing data from claims, underwriting and M&A documents
- ✓Government and defense agencies that need air-gapped or in-country AI deployments
- ✓Enterprises modernizing COBOL and mainframe applications to Java with equivalence checking
- ✓Security teams that want AI vulnerability hunting on source code without sending code to an external model
- ✓High-volume document workloads where per-token pricing from frontier APIs becomes expensive
Not Ideal For
- ✗Teams that want a self-serve SaaS API with a published price list; every deal starts with a sales conversation and a pilot
- ✗Buyers who need verified third-party evidence today, since the 78% cost saving is a company model and reviews are absent from G2-style sites
- ✗Organizations that want a single focused product, as Lazarus spans documents, cybersecurity and code modernization with two more products still in trial
Deployment
Market Analysis
Pros
- ✓Deployable on premises, air-gapped or in-country, which suits insurers, defense and government buyers
- ✓Strategic insurance investors (QBE, State Farm Ventures, Nippon Life-backed AllegisNL) and named outcomes with a reinsurer and the US Army
- ✓No per-token fees and swappable open-weight models reduce lock-in to one frontier provider
- ✓Clearwing drew Hacker News attention in April 2026 as a model-swappable open alternative to Anthropic's Glasswing
Cons
- ✗Headline economics (78% saving, ~10% of frontier TCO) are company-modeled; Dealroom says the round's stage, terms and performance claims were not independently verified
- ✗No published certifications such as SOC 2 or ISO 27001 on the website
- ✗Broad product spread across documents, security and code modernization, with Scribe and Hexis still in enterprise trial until Q1 2027
- ✗No public pricing, and engagements run through pilots and co-builds rather than self-serve
- ✗No user reviews found on G2-style sites and the Lazarus-AI GitHub organization shows no public repositories
Pricing
Pilot (S0)
Contact for pricing
- ✓Scoped pilot on customer data
Co-build (S1)
Contact for pricing
- ✓Joint build with Lazarus engineers
Buy and self-operate (S2)
Contact for pricing
- ✓Customer runs the system with Lazarus support
- ✓No per-token fees
Via partner (S3)
Contact for pricing
- ✓Delivered through a partner
No price list is published. Lazarus says it charges no per-token fees, quotes hardware under $10K and claims about 10% of frontier-model total cost of ownership, plus a modeled 78% annual saving in a high-volume scenario; Dealroom notes none of these figures were independently verified.
Security & Compliance
Sources
This page was written from 6 sources, 5 on domains other than lazarusai.com.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
geniosAI
Auditable AI agents for financial analysis that trace every figure to its source, for the office of the CFO and investment teams
Casap
AI agents for card disputes, chargebacks and first-party fraud at banks, credit unions and fintechs
Coram AI
AI physical security platform that adds natural-language video search and alerts to existing IP cameras
Kalepa
AI underwriting platform for commercial and E&S insurers, from submission intake to quote, bind and portfolio