Kolena

Kolena sells an AI automation platform for document-heavy knowledge work. The product ingests source files from systems such as Google Drive, SharePoint or email, detects document types, extracts and validates fields against customer-defined business rules, generates outputs in customer templates, and exports results to downstream systems. It handles PDFs, spreadsheets, emails, scans and audio, and is designed to run hundreds or thousands of documents in parallel rather than operate as a chat interface. Every extraction is presented with reasoning, confidence scores and citations to the source document.

The platform is organized around configurable agents. Kolena ships a catalog of pre-built, "verified" agents for tasks including lease abstraction, CAM reconciliation, rent roll analysis, loss run analysis, loan onboarding and compliance testing, which customers can launch and customize without AI expertise, and it recommends further automations. The rules and policies encoded in agents also function as living process documentation that can be queried in plain language, with the stated aim of keeping written policy and executed work aligned. A quality-assurance layer benchmarks newly released models each month so customers can switch models with one click; the site advertises 99.2% extraction accuracy.

Earlier, at the time of its 2023 funding announcement, Kolena was positioned as a model-quality framework for testing, benchmarking and validating AI models — offering scenario-level and unit testing, end-to-end testing of AI/ML products, test-coverage gap analysis and deployment risk tracking, with an architecture in which customers did not need to upload their data or models and only test results were stored.

Founding story

Mohamed Elgendy launched Kolena in 2021 with Andrew Shi and Gordon Hart. The three had worked together for around six years in AI divisions at companies including Amazon, Palantir, Rakuten and Synapse, and set out to build a "model quality framework" providing unit and end-to-end testing for AI models in an enterprise-friendly package, motivated by the view that AI lacked trust from both builders and the public.

Business model

Enterprise software sold directly to companies in real estate, financial services, insurance and banking, with a two-week free trial, no setup fees, prebuilt workflows that can go live in hours, and enterprise support and integrations.

Traction

The website cites named customer references including Essential Properties Realty Trust, EAH Housing, SFA Partners, Paragon Loan Advisors, Milestone Bank, Real Capital Solutions and Loffa, with reported results such as roughly four hours saved per lease and nearly $100,000 in efficiency gains across 58 leases, an 85% reduction in lease review time, a 96% reduction in UCC filing review labor with funding turnaround cut from five days to hours, and turnaround-time reductions of 95%+. Agent usage counts listed in the catalog range from several hundred to 1,400 uses. At the time of the 2023 funding round the company had 28 full-time employees and declined to disclose customer counts.

Latest developments

The current product line centers on end-to-end document workflow automation with an agent catalog, centralized process documentation and retrieval, an automation hub for accuracy monitoring, and continuous model benchmarking, offered with a two-week free trial and SOC 2, HIPAA and PCI compliance claims.