EchoFelix · Custom Legal AI
Legal Q&A, case law search, and contract analysis — built on verified South African legal data, your firm's documents, and authoritative public sources. Pre-built modules or fully bespoke systems designed around how your practice actually works.
Your attorneys already use Copilot and ChatGPT daily. The problem is not whether AI works — it is that generic tools hallucinate citations, leak privileged information, and know nothing about South African courts. This is the alternative built for legal practice.
83% of lawyers already use AI for research. The question is not adoption — it is whether the tool you are using is safe, accurate, and built for the work you actually do.
Nearly 1,000 documented instances of AI-fabricated legal citations have been filed in courts across six countries since 2022. Attorneys have been sanctioned, fined, and referred for professional misconduct. ChatGPT generates convincing case names, docket numbers, and judge names for cases that do not exist — and presents them with total confidence. A custom AI system grounded in verified case law cannot invent what it retrieves from an authoritative database.
When you paste a client instruction into ChatGPT or ask Copilot to summarise a privileged document, that content enters an environment you do not control. Retention policies, training data usage, and third-party access are governed by the provider — not by your professional obligations under the Legal Practice Act or POPIA. A custom system keeps privileged matter data inside your environment with retention and access controls you define.
ChatGPT was not trained on the Constitutional Court's full jurisprudence, the latest SCA decisions, or the 2024 amendments to the Companies Act. It does not understand the hierarchy between a High Court division ruling and an SCA judgment. It cannot distinguish ratio from obiter. A purpose-built legal AI system draws from structured South African legal databases — not from a general internet scrape that treats a blog post and a Supreme Court judgment with equal weight.
Every module retrieves information from verified sources, generates answers with traceable citations, and presents output an attorney can verify before relying on it. This is not a chatbot that guesses — it is a retrieval system that finds and reasons over real legal data.
Ask a question. Get a cited answer.
Ask any legal question in plain language and receive a structured answer grounded in case law, legislation, or your uploaded documents. Every response includes the source — judgment citation, section reference, or document page — so you verify before you rely on it.
Example queries
Find authority. Understand how it fits.
Search across court hierarchies with natural language — not just keywords. The system understands judicial hierarchy, distinguishes between ratio and obiter, and surfaces how later courts have treated a judgment. Find the leading case, the dissent, and the most recent application in your division.
Example queries
Upload a contract. Query its terms.
Upload any contract and ask questions about its terms, obligations, risks, and gaps. The system extracts clauses, maps obligations to parties, identifies missing protections, and compares terms against your standard templates or statutory minimums. Built for review, not just extraction.
Example queries
The difference between a generic AI tool that invents citations and a legal AI system that retrieves real authority comes down to architecture. Here is the short version.
In plain language — the way you would brief a junior associate. “What are the requirements for a valid cession under South African law?” or “Find me the latest SCA authority on contractual penalties.”
Instead of generating an answer from memory (which is how ChatGPT fabricates cases), the system searches across its legal databases — case law, legislation, your documents — and retrieves the actual passages that answer your question. It finds before it writes. This is why citations are real and traceable.
A structured response with the reasoning, the authority it draws on, and links to the source material. You read the judgment, check the section, verify the reasoning — and then use it in your advice. The AI accelerates the finding; you apply the professional judgment.
This retrieve-then-generate architecture is what separates a legal AI system from a general chatbot. Every answer is grounded in source material that exists and can be checked. The system cannot cite a case that is not in its database — because it only cites what it retrieves.
A legal AI system is only as good as the data it draws from. EchoFelix is built on authoritative South African legal sources — not internet scrapes, not Wikipedia, not legal blogs. Here is what feeds the system and why each source matters.
Constitutional Court, Supreme Court of Appeal, High Court divisions, Labour Court, Labour Appeal Court, Land Claims Court, Competition Tribunal, and specialised courts. Historical and current judgments structured for search, cross-referencing, and citation.
Includes: Full-text judgments, headnotes, citation metadata, judicial hierarchy mapping.
National Acts, provincial legislation, government gazettes, regulations, and notices — structured in machine-readable format with amendment tracking so the system always references the current operative version.
Includes: Companies Act 71 of 2008, Labour Relations Act, POPIA, National Credit Act, Immigration Act, Income Tax Act, and subordinate regulations.
Academic analysis, case notes, and practitioner commentary that adds interpretive depth beyond what the bare judgment or statute provides.
Includes: South African Law Journal, De Jure, Stellenbosch Law Review, Industrial Law Journal, and specialist publications.
Contracts, opinions, memoranda, precedent files, and matter documents uploaded into your private instance. These never leave your environment and are never used to train shared models.
Includes: Standard contracts, precedent opinions, clause libraries, internal knowledge base documents, and matter-specific uploads.
Curated contract templates and clause libraries covering common South African commercial transactions — structured so the AI understands clause purpose, not just clause text.
Includes: Sale agreements, lease agreements, shareholder agreements, employment contracts, service level agreements, NDAs.
Legal data is only useful if it is current and correctly structured. Case law is ingested with full citation metadata, court hierarchy tagging, and judicial treatment tracking. Legislation is maintained with amendment history so the system references the operative version — not a repealed section from three years ago.
Your firm's private documents are indexed separately and never mixed with the public corpus. When you ask a question, the system can search across both — public authority and your own precedent files — but your documents are never visible to anyone outside your firm.
A legal AI tool that is technically impressive but impractical to use daily is not worth the licence fee. These are the three dimensions we optimise for.
A commercial litigator and a family law practitioner ask fundamentally different questions. The system is configured for the practice areas your firm operates in — the right statutes, the right court hierarchies, and the right precedent patterns for the work you do daily.
Due diligence Q&A across transaction documents, Companies Act compliance checks, shareholder agreement clause extraction, BEE verification against scorecard criteria.
Cross-reference pleadings against discovered documents, find precedent on disputed legal points across High Court divisions, build chronologies from voluminous records.
Lease clause comparison across portfolio properties, sectional title rule compliance, title deed condition extraction, municipal regulation lookups.
CCMA award pattern analysis, employment contract clause review against BCEA minimums, disciplinary hearing precedent search, retrenchment procedure compliance checks.
Antenuptial contract review, maintenance calculation precedent lookup, custody order comparison against Children's Act requirements, divorce settlement clause analysis.
Sentencing precedent search across magistrate and High Court decisions, bail application authority lookup, criminal procedure compliance verification.
Visa category requirement extraction from Immigration Act and regulations, permit application precedent search, cross-reference DHA directives against current practice.
RAF claim quantum precedent analysis, medical report extraction and cross-referencing, insurance policy clause comparison, prescription period verification.
Business rescue plan compliance checks against Chapter 6 requirements, creditor claim verification, sequestration precedent search, Companies Act section 129 procedure analysis.
Income Tax Act interpretation across SARS rulings and court decisions, VAT regulation analysis, transfer pricing documentation review, tax dispute precedent search.
NCA compliance verification, FICA due diligence document review, financial regulation cross-referencing, loan agreement clause extraction and comparison.
MPRDA compliance checks, environmental impact assessment review, water use licence requirement extraction, mine closure obligation analysis.
Multi-practice firms get a single system configured across all their departments. Solo practitioners get a system focused on their niche. The configuration adapts to how your firm is structured — not the other way around.
Some firms want to start using AI this month. Others need a system engineered around specific workflows, data sources, and integration points. We deliver both.
Firms that want to start immediately
Production-ready AI modules for legal Q&A, case law search, and contract review — configured for South African law and ready to deploy. You choose which modules your firm needs, we configure them for your practice areas, and you are operational within weeks.
Firms with specific workflow or data requirements
Bespoke AI systems designed around your firm's exact workflows, document types, and integration requirements. We build to your specification — from the data sources that feed the system to the interface your team works in daily.
Attorney-client privilege and data protection are not features you add later. They are architectural decisions made before the first line of code.
Data processed and stored on South African infrastructure. No cross-border transfers of client data without explicit consent and contractual safeguards.
Your documents and queries are never used to train shared AI models. Your matter-specific data improves your system only — and only within your private instance.
Role-based access so partners, associates, and candidate attorneys see what they should. Full audit logging of queries and document access for compliance and risk management.
Data retention periods aligned with your firm's policy and professional conduct requirements. Automated deletion schedules. No indefinite storage of privileged material in an environment you do not govern.
For detailed information on our security measures and data processing agreements, see our security and compliance page and POPIA compliance statement.
Custom legal AI integrates with the rest of the EchoFelix platform — and with the systems your firm already relies on.
Transcripts from dictation and transcription feed directly into the AI system — turning a client consultation recording into searchable, queryable matter data.
Matter documents processed through case analysis connect to the legal research and Q&A modules — so facts extracted from your bundle inform the legal questions you ask next.
Answers from law research export directly into document drafts — research memos, heads of argument, or opinion outlines that your team edits and signs off.
No AI system — no matter how well built — replaces an attorney's professional judgment. EchoFelix reduces the time between a legal question and a structured, sourced starting point. Every citation should be verified against the primary source. Every contract analysis should be reviewed before it informs advice. Every research output is a draft for attorney sign-off, not a final work product.
This is not a limitation — it is how responsible legal AI should work. The tool finds and organises. You analyse, advise, and take professional responsibility for the output.
Whether you need a pre-built module operational in weeks or a bespoke system engineered around your workflows — the conversation starts with understanding your practice areas, your document types, and what you need the system to do daily.
We scope every implementation to your firm's size, specialties, and confidentiality requirements. No generic demos — we discuss your actual use cases.