Legaleey

An AI powered document search platform that reduces the hours lawyers spend finding and locating items within contracts.

B2B SaaS Legal Tech Desktop First AI Search
Legaleey search results screen with the persistent filter panel, date and percentage sliders, and clause aware results in an eight column table

At a Glance

Role: UX and Product Designer, end to end
Timeline: June to November 2023
Platform: Desktop first web application
Type: B2B SaaS · Legal Technology · AI Search · Demo Release
Outcome: Time to find relevant clauses dropped 80 percent compared to Acrobat in testing

Search that works the way lawyers think

Legaleey is an AI driven document search service built for legal firms. The goal: reduce the time lawyers spend finding and locating items within contracts, so they can focus on judgment rather than document retrieval.

The demo release was designed to validate three core questions: whether AI driven search meets customer needs, how users interact with search options and filters, and whether the structured output is understandable and genuinely valuable to an experienced legal professional.

9 Flows Designed
15 Participants
B2B SaaS Platform
Desktop Primary Target

Hours lost to the wrong kind of search

Contract review is one of the most time intensive tasks in legal practice. Lawyers at all levels spend a disproportionate share of their working hours not reading contracts but hunting within them.

The tools available, primarily Adobe Acrobat and Microsoft Word, were built for document creation and reading, not structured retrieval. Their search functions return raw text matches with no understanding of legal structure, clause type, or semantic context.

Core Problem

Experienced lawyers spend 40 to 50% of contract review time on manual search, pulling up keyword matches that lack structure, context, or clause awareness. There is no way to search across documents, filter by legal concept, or export findings in a usable format.

User Profile

The target user is an experienced lawyer who is deeply familiar with complex Google searches and proficient in Adobe Acrobat and Microsoft Word. They are not a beginner. They have high expectations for precision, speed, and information density.

Senior Associate
Large law firm · Contract review and due diligence

Goal: find specific clauses, parties, and dates across dozens of contracts without rereading every document.

Partner
Midsize firm · Client advisory and deal management

Goal: quickly surface risk exposure across a portfolio of contracts before client meetings.

What legal search actually needs

Interviews and observation sessions with 15 legal professionals and non lawyers, supported by secondary research, revealed a consistent pattern of workarounds, frustration, and lost time in existing search workflows.

Key Pain Points

No Structural Awareness

Acrobat returns raw keyword hits. It cannot distinguish a mention of a party name in a recital from a material obligation in a covenant.

No Cross document Search

Lawyers managing portfolios of 10 to 100 contracts must open each file individually. There is no way to query across a set.

No Exportable Output

Even when the right text is found, getting it into a memo, spreadsheet, or summary requires manual copying. There is no structured export path.

Cognitive Overload

Reviewing dense legalese while also managing a search UI creates significant cognitive load, especially under time pressure before hearings or closings.

Research Insight

Lawyers do not want a smarter Ctrl+F. They want a tool that understands the structure of a contract well enough to answer questions like: who are the parties, what are the termination conditions, and are there any governing law clauses that deviate from the norm?

Four principles that shaped every screen

Before any wireframes were drawn, four design principles were established. Every subsequent UI decision was evaluated against them.

Simplicity

Avoid overloading the UI. Let the ML surface the right results. The interface should recede so the content can lead.

Structure

SERP output must be easy to read. Clear information architecture so lawyers can scan results at speed, not read them line by line.

Familiar Patterns

Use standard UI elements. Lawyers are not product explorers. Friction from unfamiliar patterns destroys trust and adoption.

Desktop First

Contract review happens on large monitors. Design for horizontal space, keyboard navigation, and dense information displays.

Search onboarding modal with an embedded screenshot of the SERP

Onboarding embeds a screenshot of the actual SERP so users arrive at search with formed expectations. Showing the real interface during orientation reduces the learning curve on first use without adding a separate tutorial step.

Search onboarding modal explaining that each added filter triggers a new search

Explaining that each added filter triggers a new search prevents users from applying all filters and waiting for a single result. This behavioral cue was added after testing revealed users were stacking filters before running any search at all.

Search results with the persistent filter panel for party, people, and county

Filters for party name, people, and county map directly to how legal professionals think about clause retrieval. These categories were derived from user interviews and reflect the actual mental model lawyers use when searching across contracts.

Nine flows, end to end

The full product scope covered nine primary user flows, from first sign up through daily document search. Each flow was designed to feel complete and self contained, with clear feedback at every step.

  • 05

    Document Management

    Table view with folder and file listing, date created, last modified, and file type. Context menu with Download, Make a copy, and Delete. Filter pills for Uploaded, Verified, and Problems states.

  • 06

    Search and SERP

    An always visible search bar anchors every screen. A left filter panel covers file, party, date, and percentage parameters, with results surfaced in an 8 column table with Export to Excel.

  • 07

    Search Options Modal

    Toggles for Case Sensitive, Include Inflections, and Maintain Word Order. A single select group for Search Anywhere vs. Search Within N Words, with a number input for the distance parameter.

Also designed

  • 01

    Onboarding and Sign Up · email, password, phone, and verification, kept lightweight

  • 02

    First Time Dashboard · four orientation tiles instead of a walkthrough

  • 03

    Returning User Dashboard · recent work surfaced for mid project returns

  • 04

    Document Upload · circular drop zone with full error handling

  • 08

    Search Help Modal · Boolean operator reference with concrete examples

  • 09

    Settings Panels · account, notifications, billing, and privacy

Upload screen with circular drag and drop zone and Add files button

The circular focal point was chosen over a rectangular zone to reduce upload hesitation in testing. A single centered target creates a clear call to action without requiring any instructional copy.

Upload screen with the first iteration rectangular drop zone

The rectangular drop zone was the first iteration, replaced after side by side testing showed the circular variant reduced time to first upload.

Login screen for existing users in a distinct visual environment

A separate visual environment for login signals the transition into a secure, credentialed workspace. The distinct treatment reinforces that document search happens inside a protected context, not a public facing surface.

Returning user dashboard with recent searches

Returning users are dropped directly into their four most relevant task states, skipping any navigation step. The tile layout was chosen over a list view because lawyers returning mid project need to act immediately, not orient themselves.

Why we built it this way

Six decisions defined the character of the product. Each one was grounded in a specific user need, not visual preference.

Why is the filter panel always visible?

Lawyers refine searches iteratively. Hiding filters behind a toggle forces them to interrupt their scanning rhythm every time they want to adjust a parameter. A persistent panel reduces navigation overhead and respects how legal professionals actually work.

Why is Apply always a sticky button?

On long filter panels, users scroll past the button. A sticky Apply keeps the action visible at all times, reducing the chance of a lawyer setting filters and forgetting to run the search, which wastes time and creates confusion.

Why a table layout for results?

An early version used card based results. User feedback showed lawyers could not scan across multiple results quickly, so we switched to a table. The format maps directly to how legal professionals already process information and allows Export to Excel, which fits naturally into how findings get documented and shared within a firm.

Why a circular drop zone for upload?

A single, clear focal point on the upload screen reduces hesitation. The circular form creates visual hierarchy without adding text instructions. It signals: this is where the action begins.

Why no native app?

Lawyers work across firm devices, client environments, and court facilities. A browser based platform requires no IT approval or installation and is accessible from any machine with login credentials.

Why a Boolean reference modal?

Power users want advanced queries but do not always remember operator syntax under pressure. An in product reference reduces support requests and builds confidence without requiring a separate documentation visit.

Design Tension

The hardest balance in this product was between simplicity for new users and power for experts. The solution was progressive disclosure: simple by default, with advanced options always one step away but never in the way.

File sorting tutorial showing the Download, Make a copy, and Delete actions

File management actions were shown in context within onboarding rather than discovered through trial and error. Surfacing the Download and Delete options early reduces support requests from users who cannot find core file actions.

Upload progress with per file status and unsupported format errors

Upload status was tracked per file so lawyers could identify unsupported formats before committing to search. Per file feedback was added after testing showed users had no way to diagnose why certain documents were not appearing in results.

Document management table view with file actions menu

A table layout for document management mirrors the file paradigm lawyers already use in their daily workflows. Familiar patterns reduce the learning curve and increase the likelihood that new users complete their first upload without assistance.

Search results with date range and percentage slider filters

Date range and percentage filters address the most common clauses lawyers need to surface across contract portfolios. Adding sliders rather than text inputs for these fields reduced input errors and sped up filter application in testing.

The marketing site around the product

The demo release shipped with a small marketing site alongside the product: a homepage, a Who We Are page, and a contact page. Their job was to establish the AI value proposition and give firms a human touchpoint before account creation.

Legaleey marketing homepage with sign up form and demo call to action

The marketing homepage pairs the value proposition with a sign up form and a demo call to action, the public entry point for firms evaluating the product.

Who We Are section of the marketing site explaining the AI value proposition

The marketing site was designed to establish the AI value proposition before any account commitment was required. Lawyers evaluating new tools need to understand the core benefit before engaging with a sign up flow.

Contact page for connecting with a Legaleey team member

Direct firm outreach was prioritized because legal procurement decisions require a human touchpoint before account creation. The contact page was built into the product rather than linked externally to keep evaluation friction low.

Validating AI search for legal work

The demo release was used to validate product market fit with early law firm customers. The results confirmed that AI driven search addresses a real, high priority pain point in legal practice.

Search Quality

Users confirmed that structured, clause aware results were significantly more useful than raw keyword matches from Acrobat. The study ran with 15 participants, lawyers and non lawyers, each running the same searches on the same documents in Legaleey and in Adobe Acrobat. Time to find relevant clauses dropped by 80%.

Filter Adoption

Users actively used the date sliders and party filters, confirming the decision to keep the filter panel always visible rather than hidden.

Export to Excel

One of the most used features in early testing. Lawyers immediately exported results into existing workflows, validating the table based output format.

Upload Experience

The drag and drop upload with real time progress feedback received consistently positive responses. The upload flow scored the highest satisfaction rating of any flow in user testing.

Upload error modal with specific cause and View File and Retry recovery actions

Error messages surface the specific cause and provide a recovery action rather than generic failure copy. Lawyers under time pressure need to know immediately whether an error is recoverable and what to do next.

Future Directions

Annotation Layer

Allow lawyers to annotate and tag search results directly in Legaleey, creating a persistent record of their review findings.

AI Confidence Scores

Surface a confidence indicator alongside each AI result so lawyers can quickly identify which findings need manual verification.

Key Finding

The demo validated that structured AI search output is not just useful but immediately adoptable by legal professionals when presented in familiar UI patterns. The biggest barrier to adoption was not learning curve but trust in the AI output accuracy. That finding directly shaped the next phase of design, surfacing confidence scores alongside every AI result.

OutcomeTime to find relevant clauses dropped 80 percent compared to Acrobat in testing.

Next project Lumi An AI social robot that monitors anger patterns, calms arguments in real time, and builds empathy through light, sound, and voice.

Open to full time roles and select freelance work. Say hello, find me on LinkedIn, or use the contact page.