# discover.legal > Legal technology consultancy for law firms and legal teams: selection (Buy), custom builds (Build), managed services (Run) and fractional technical leadership (Advise). Founded and run by a Law Society of Ontario licensee. Not a law firm; no legal advice. ## Services ### Buy: Choose the right tools, on your terms. Independent selection for firms buying legal technology: what you actually need, which vendors deliver it, and what the contract should say about your data. Price: Price on request. Details: https://discover.legal/work#buy ### Build: Build what the market does not sell. Custom intake, document automation, AI-assisted workflows with verification gates, and the integrations that tie them to the systems your firm already runs. Price: Price on request. Details: https://discover.legal/work#build ### Run: Managed services for the systems you rely on. Hosting, monitoring, updates, backups and support for what we build or what you choose, including open-source platforms like BigLaw, on infrastructure you control. Price: Price on request. Details: https://discover.legal/work#run ### Advise: Technical leadership without the headcount. Strategy, roadmap, AI governance and build-versus-buy decisions for firms and legal teams, on retainer or for a defined project. Price: Price on request. Details: https://discover.legal/work#advise ## Our own work ### discover.legal Documents (Sample walkthrough · Ontario) A guided interview turns a client’s story into organized Ontario family-law drafts, in the vocabulary the court uses. Shown here as a scripted sample with a fictional client. - Interview once; every document fills from the same record - Every fact keeps the words it came from - A missing fact becomes a Draft note, never an invention - A recorded demo: no accounts, nothing sent, nothing saved ### BigLaw (Free · Open source · Experimental) An open-source legal AI platform: research, drafting, redlining, e-signatures, docketing and billing, orchestrated by a bench of agents with a verification protocol between every finding and the page. - Apache-2.0, self-hosted, one static Go binary - Runs on local models or your provider of choice - Human gate before low-confidence findings reach synthesis - Aimed at solos, boutiques and small firms. Experimental: not yet independently audited. ## How we work - **Verification before output.** We design every AI step so its output is checked against the record before it reaches a page. We build it that way, and we buy it that way. - **Your data, your infrastructure.** Client data stays in an environment you control. Open models and self-hosting are always on the table. - **Open source where it counts.** Our own platform is public and free. When we build for you, you own the code. - **Fixed scope, written outcomes.** Engagements end with something you can read and act on: a recommendation, a running system, or a roadmap. ## Questions firms ask first **Who is this for?** Law firms and in-house legal teams that are choosing, building or running technology, from a solo practice standardizing intake to a firm replacing a vendor stack. **Can I use Documents today?** Not as a public product. It is shown here as a sample walkthrough with a fictional client: there are no accounts and nothing is saved. If you want intake like it for your firm, that is a Build engagement. **Is BigLaw really free?** Yes. BigLaw is open source under Apache-2.0. Paid work is the consultancy: selection, custom builds, managed services and advisory. **What does it cost?** Prices are on request. Every engagement is scoped in writing first, so you know what you get before anything is billed. **Do you resell software or take referral fees?** No. Selection work is independent. If we recommend a vendor, it is because it fits, and the contract we review is yours. **Can you run BigLaw for my firm?** Yes, as a managed service on your infrastructure. BigLaw is experimental and has not had an independent security audit, so a deployment starts with that review and authentication switched on. **Is any of this legal advice?** No. We are a technology consultancy, not a law firm. Nothing on this site or produced by our tools is legal advice. **How do we start?** Email us with the problem. We use a first conversation to understand it and come back with a scoped proposal. ## Who we are 17549903 Canada Ltd., Ontario, Canada. Horatiu Druma-Strugariu, licensed by the Law Society of Ontario. Every claim can be checked at https://discover.legal/verify. ## Contact Contact: h at discover dot legal. A first conversation is how every engagement starts. --- # Services Every engagement is scoped in writing first. We are a technology consultancy, not a law firm, and we do not give legal advice. ## Buy: Choose the right tools, on your terms. Independent selection for firms buying legal technology: what you actually need, which vendors deliver it, and what the contract should say about your data. A fixed-scope engagement that ends with a written recommendation you can act on. We take no vendor referral fees, so the shortlist is yours, not ours. What you get: - Needs assessment across practice groups - Market scan and shortlist - Scripted vendor demos on your matters - Contract, data and AI-terms review - Rollout and change plan Price: Price on request. ## Build: Build what the market does not sell. Custom intake, document automation, AI-assisted workflows with verification gates, and the integrations that tie them to the systems your firm already runs. We build the way we build our own products: code owns the workflow and the rules, the model supplies judgment and language, and dispositive statements need an explicit check before they reach a page. You own the code. What you get: - Client intake and interview flows - Document assembly from structured facts - AI workflows with human review gates - Practice-management, DMS and e-signature integration - Source code delivered to you Price: Price on request. ## Run: Managed services for the systems you rely on. Hosting, monitoring, updates, backups and support for what we build or what you choose, including open-source platforms like BigLaw, on infrastructure you control. A monthly service with a defined scope and response times. Your data stays in your environment; we operate it, you keep the keys. What you get: - Deployment on your cloud or on-premises - Monitoring, patching and backups - Model-provider and cost management - Access control and audit trail - Support desk for your staff Price: Price on request. ## Advise: Technical leadership without the headcount. Strategy, roadmap, AI governance and build-versus-buy decisions for firms and legal teams, on retainer or for a defined project. Retained or project-based. Useful when the firm needs someone technical in the room for decisions that will outlast any single vendor. What you get: - Technology roadmap and budget - AI use policy and risk review - Build-versus-buy analysis - Vendor negotiation support - Fractional CTO on retainer Price: Price on request. ## Contact Contact: h at discover dot legal. A first conversation is how every engagement starts. --- # Research Open experiments on how language models behave on legal tasks, published with their code, design and status. ## Does the language a model reasons in change its legal accuracy? Status: In progress · pilot validated · no full-scale results yet. A factorial experiment over LegalBench: reasoning-language conditions × models × closed-label legal classification tasks. Nineteen reasoning conditions (fourteen natural languages across nine families, plus formal logic, pseudocode, an emergent notation, a wildcard, and a no-chain-of-thought control) are run against nine LegalBench tasks on a set of low-cost cloud and local models, with the final answer always requested in English. - Nine closed-label LegalBench tasks: hearsay, personal jurisdiction, contract NLI, unfair terms of service and others - Identical seeded samples across every cell; temperature zero and no output cap where the API allows - Hidden "thinking" switched off wherever possible, and measured and reported where it cannot be - Fifty-six offline tests; a live pilot run completed end to end with zero errors In the pilot, one model returned hidden reasoning on every sample, and in the Mandarin condition its visible reasoning was mostly Chinese while the hidden channel was mostly not. A model that thinks in a hidden channel is not reasoning in the requested language, which is why the design measures it. Code: https://github.com/hordruma/MultiLingual-Reasoning --- # Verify discover.legal Each claim this site makes about who runs it, and where to check it independently. - The business is 17549903 Canada Ltd., a federal Canadian corporation. Check: Search the corporation name in the Corporations Canada federal corporation database. https://ised-isde.canada.ca/cc/lgcy/fdrlCrpSrch.html - Horatiu Druma-Strugariu is licensed by the Law Society of Ontario (Law Society number 73174O). Check: Search the name or number in the Law Society of Ontario Lawyer and Paralegal Directory. https://lsodirectory.lso.ca/ - Horatiu Druma-Strugariu has worked in law and legal technology for about 10 years. Check: The professional profile lists roles and dates. https://www.linkedin.com/in/hor-druma/ - BigLaw, our open-source legal AI platform, is public under Apache-2.0. Check: Read the code, license file and commit history. https://github.com/discover-legal/BigLaw - Our other public work lives in the discover-legal GitHub organization. Check: List the organization's public repositories. https://github.com/discover-legal - This site is discover.legal, served over HTTPS with HSTS preload. Check: Check the certificate and the Strict-Transport-Security header of any page. https://discover.legal ## Contact Contact: h at discover dot legal. A first conversation is how every engagement starts. --- # Privacy Policy Last updated November 2026. This site is an information site. It has no accounts, no forms and no sign-in, and it saves nothing you do. ## What this site collects Nothing you type or click is stored by us: the site has no forms, and the sample walkthroughs run entirely in your browser from scripted, fictional content. Like any website, the hosting provider receives the technical details of each request (such as IP address, browser and the page requested) and keeps them in server logs for security and operations. We do not use them to build profiles of visitors. The site sets no cookies and loads no advertising or tracking scripts. ## When you email us If you email us, we receive your address and whatever you choose to send, and use it only to reply and to scope any work you ask about. Please do not send confidential client information before an engagement is agreed. ## Links to other sites Links to GitHub or other sites take you to services with their own privacy policies. ## Questions and changes Email us with any question about this policy or to ask what we hold about you. If this policy changes, the date at the top changes with it. --- # Terms of Service Last updated October 2026. These terms cover your use of this website. Consultancy work is governed by the written agreement for that engagement. ## Not legal advice discover.legal is a legal technology consultancy, not a law firm. Nothing on this site is legal advice, and using the site does not create a lawyer-client relationship. For advice about your situation, speak to a licensed lawyer. ## The samples The Documents and BigLaw walkthroughs are samples: fictional clients and matters with scripted output in the style of the products. They show how the software works; they are not live products, and their content must not be relied on or filed. ## Open-source software BigLaw is published under the Apache License 2.0. That license, not these terms, governs your use of the BigLaw source code, and like the license says, it is provided without warranty. ## Site content The site is provided as is. We work to keep it accurate, but we do not warrant that it is complete or current, and we are not liable for loss arising from reliance on it. Our names and logo may not be used to suggest endorsement without permission. ## Changes We may update these terms; the date at the top shows the current version.