K

Knovos Discovery

Knovos Discovery is a full eDiscovery platform first. Processing, deNIST filtering, deduplication, email normalization, early case assessment, review batches, redaction and production with Bates ranges and privilege logs all run without a language model. AI drives a growing share of the review. Knovos Insight tags documents as responsive, non responsive or by issue from a natural language case prompt. It answers questions over the review set with cited excerpts, summarizes in the reviewer's language, transcribes audio and video, and finds personal and health information across more than 300 file formats.

Technology assisted review, sentiment and theme detection, image analytics and name normalization sit alongside it. The same Insight engine powers the other Knovos products, Manage, Rooms, GRC and Arbitrate, rather than being sold as a separate AI product.

Insight is built on retrieval augmented generation over the customer's own documents, and its answers come back with cited source excerpts that reviewers can trace. Review predictions carry a rationale and the uncertainties the model flagged. Knovos goes further in its claims, describing output as never invented and never hallucinated, and says Knovos Discovery reviewed 4.3 million documents at 90 percent accuracy. No test set, method or error breakdown sits behind the 90 percent figure, and no precision or recall result is released for prompt based review or technology assisted review.

Human judgment is a stated design principle. Its AI, in Knovos's words, supports judgment while people make the decisions, always. In review, Insight predicts responsive, non responsive and issue tags across the full set, and each prediction shows its rationale, a confidence value and flagged uncertainties so reviewers can validate, refine and defend it. Privilege log drafting and personal data detection are framed the same way, with legal oversight kept in the loop.

A reviewer management console assigns batches and tracks progress in real time, quality control steps run through review and production, and every AI action is logged. The confidence level at which a prediction must go to a person, and what happens after a wrong call, are left to each team.

Every customer story is anonymous. One case study describes an elite litigation and regulatory firm with more than 3,500 legal professionals in over 30 offices that moved its eDiscovery project management onto Knovos Manage. Another describes a United States government law office running Knovos Discovery for several years in a dedicated cloud environment that Knovos builds and manages. A large research university is said to have replaced three tools with it for public records work, and a senior manager at an unnamed global AM 100 law firm recommends Knovos.

Outcome figures come without a customer or a method behind them, including 4.3 million documents reviewed at 90 percent accuracy and review volumes cut by 60 to 80 percent with technology assisted review.

Confidentiality rests on architecture and on agreements signed outside public view. Each client signs a confidentiality and nondisclosure agreement with Knovos, which keeps engagement data under strict privacy rules because much of it is evidence. Its AI runs inside the same access controls as the documents, and with a customer's own model the processing stays inside the customer's environment, with no exposure to outside cloud AI.

Tenants on Knovos Cloud are isolated from each other. Nothing published commits Knovos itself to never training on client data, states a retention period for prompts or outputs, or addresses privilege and work product handling.

The terms of use state that Capital Legal Solutions, LLC, doing business as Knovos, is not authorized to practice law or provide legal services. Its services cover only the non legal, administrative side of document review and discovery projects. Its AI principles add that people make the decisions and that AI extends professional expertise without overruling it. The platform is sold to law firms, legal departments, government offices and investigators rather than to the public. No ethics opinion, supervision duty or jurisdiction limit is named.

Knovos publishes five AI principles. AI is native to every product, grounded in the customer's content, defensible through audit trails and confidence values, sovereign through the customer's choice of model and infrastructure, and led by people. Customers who bring their own model can evaluate, monitor and govern it inside their own environment, and every AI decision is logged. No testing done before release is described, and no one inside Knovos is named as accountable for AI.

Nothing addresses uneven output across languages, custodians or document types, although the product is sold on review in more than 150 languages.

Controls are described in detail. Data is encrypted with AES-256 at rest and in transit, and Knovos Cloud tenants are isolated. Role based access defines who does what on which data set, administrators can block unapproved IP ranges, and audit logs record who changed what and when. Customer managed encryption keys and SIEM integration are available in every deployment model. Subprocessor disclosures, data processing addenda and standard contractual clauses are available through procurement.

The privacy policy covers marketing data under the Data Privacy Framework rather than case data. Discovery Lite keeps everything on one machine with no cloud dependency and writes a tamper evident audit entry for each action. No retention period, deletion process or breach notice practice for customer data is published.

The only published terms are the website terms and conditions, last updated 18 January 2023, which also apply to transactions for Knovos products and services unless another contract says otherwise. They cap Knovos's liability at the total price paid, exclude indirect damages, provide everything as is, and have the customer indemnify Knovos. The product license and any service level agreement are signed separately and are not published, and nothing addresses who bears the loss when an AI tag or summary is wrong.

Connections are described at the infrastructure level. Onboarding to Knovos Cloud covers single sign on and user provisioning against the customer's identity stack, every deployment can feed a customer's SIEM, and BYAIM connects outside or private models through standard APIs. Processed data from other platforms imports through load files with metadata field mapping, and chat exports from Slack, Teams and WhatsApp load for review.

Legal holds, matter management and billing live in Knovos Manage, a separate Knovos product. No documented connector to a document management system such as iManage or NetDocuments, or to a third party matter or billing system, is offered.

Four ways to run the platform are set out side by side. Knovos Cloud is a managed SaaS with tenants isolated from each other and a region chosen at provisioning from a regional footprint. Bring Your Cloud runs it in the customer's own AWS, Azure or Google Cloud account or private cloud, in the regions the customer picks. Bring Your Data Center installs it on premises, air gapped where required, and hybrid setups mix the three with migration between them.

A comparison table shows what changes in each model, from who owns the infrastructure to security configuration and audit access. AI processing follows the data, since a customer's own model runs inside the environment where the documents sit. Discovery Lite installs on one laptop for offline and field work. The regions in the SaaS footprint are not listed by name.

Knovos Cloud holds ISO 27001:2022, SOC 2 Type 2 and Cyber Essentials certification, and Knovos is certified under the Data Privacy Framework for marketing data. No auditor, report date or scope is given, and no trust center offers the reports. In customer hosted and on premises deployments, the customer's own controls govern the environment.

Knovos Insight is described as a proprietary engine built on retrieval augmented generation and multi LLM orchestration. Customers can swap in their own model from OpenAI, Google, Anthropic, an open source project or a private deployment. When a customer brings a model, that model runs where the data lives and the customer controls its version. The language models behind the native Insight engine are not named, and nothing says how customers are told when they change.

Three licensing paths are published with their units and terms. Standard licensing is a fixed annual license per user or per case, with no per gigabyte hosting fee, unlimited storage and processing, and 24/7 support and training included. Pay as you go is usage based with no long term commitment, aimed at firms with uneven caseloads. Enterprise agreements carry custom, multi year pricing with discounts for high volume service providers and large organizations.

The managed cloud follows the same subscription or pay as you go choice, with onboarding and data migration included. No price appears for any path.

Buyers are named in some detail. Law firms use it for litigation and arbitration review, and service providers can white label it. Corporations use it for investigations, breach review and data subject requests, government agencies for FOIA and investigations, and law enforcement and forensic teams run Discovery Lite on seized devices. Use cases extend to regulatory inquiries, M&A diligence, IP and pharmaceutical discovery and healthcare breach review.

Review works across more than 150 languages, and personal data detection is tuned to regional identifiers such as Aadhaar and CURP. The SaaS option is pitched at teams without on premises mandates, with sovereign or classified work steered to the other models. No practice area or matter size is excluded.