The Impact
Zero
Natural
Zero
About the Customer
REGION
INDUSTRY
JUMP TO SECTION
Share this Article
Executive Summary
A leading US online legal document platform, offering a large catalog of legal templates across Personal, Business, and Real Estate categories, engaged Matellio to solve a specific discovery problem: visitors typically know their situation but not the exact legal term for the document they need, and both category browsing and name-based search assume they do. Matellio built a fully anonymous, conversational AI assistant that lets visitors describe their situation in their own words, asks clarifying questions only when needed, searches the platform’s live document catalog in real time, and recommends only real, existing documents — never an invented one. The assistant requires no login or signup, keeping the experience frictionless from the visitor’s first message through to the document link.
Customer Challenge
Visitors typically arrive at the platform already knowing their situation, but not the correct legal terminology for it. Someone looking to rent out a spare room to a friend is looking for a Residential Lease Agreement, but a keyword search on that phrase would likely return nothing relevant. This mismatch between how people naturally describe their need and how the catalog is structured for search was a real source of visitor friction and lost conversions:
Manual search and review created a throughput bottleneck that worsened as application volumes increased.
Risk scoring depended on individual attorney judgment, with potential inconsistency across attorneys reviewing the same case.
Manual analysis created exposure to missed matches and gaps in conflict evaluation.
Adding volume meant adding attorney headcount — there was no structural improvement available within the existing process.
Without a scalable solution, the company faced longer turnaround times, rising operational costs per application, declining customer satisfaction, and a constrained ability to compete in a fast-growing LegalTech market.
Why AWS
Matellio hosts the AI service’s production workload on AWS Bedrock AgentCore Runtime, running Anthropic’s Claude model through Amazon Bedrock as a managed, containerized service. Because the browser cannot call the AI service directly, all visitor traffic is relayed through the existing Laravel application, with AWS-side authentication handled through an EC2 instance IAM role rather than long-lived credentials stored in the application.
Matellio delivered the assistant as a new, lightweight AI service alongside the customer’s existing Laravel application, rather than a rebuild of the front-end experience. The existing Laravel application continues to serve as the single front door for all browser traffic, relaying chat requests to the AI service and streaming responses back byte-for-byte — allowing the new capability to be added without disrupting the platform’s existing architecture.
Partner Solution
The assistant has a short, natural-language conversation with the visitor, understands their situation, asks clarifying questions only when there isn’t enough detail to search yet, searches the real document catalog on their behalf, and recommends the right document with a direct link — entirely anonymous, with no login or signup required. The conversation continues naturally from there: the visitor can refine their request, ask follow-up questions, or ask a related general legal question.
At the core of the solution is a LangGraph-based AI agent hosted on Amazon Bedrock AgentCore Runtime. The agent orchestrates the complete screening pipeline:
The visitor sends a message, tagged with a fresh, anonymous session ID generated for that page load.
Receives the request, relays it to the AI service, and streams the reply back to the browser as it’s generated.
Loads recent conversation context, sends it to the Claude language model, and reasons about what the visitor needs.
If enough detail is available, the model searches the real, live document catalog and gets back real candidate matches.
The model writes a natural-language recommendation grounded only in the real search results – never an invented document.
The reply streams back through Laravel to the browser in real time, rendering as it’s generated.
Both the visitor’s message and the assistant’s reply are saved to the database, so the next message in the same session has context.
Results and Benefits
Quantified usage and conversion metrics are not yet available for this solution (see internal note above). The following benefits are structural, grounded directly in the design of the solution rather than measured outcomes:
Removing login and signup removes a barrier between the visitor and the right document, at the exact moment they need it.
Visitors can succeed without knowing the correct legal term for their situation, addressing the core mismatch between how people describe their needs and how the catalog was previously searchable.
The assistant never surfaces a document that doesn’t exist in the real catalog, avoiding a credibility-damaging failure mode common to less constrained AI assistants.
The assistant stays within document-finding and general legal information, declining personalized legal advice and directing visitors to a licensed attorney instead.
Let’s turn your next big idea into a scalable product, engineered for speed, reliability, and real-world impact.
