LEA deploys autonomous agents that process files and orchestrate workflows on your team's behalf, all within your firm's existing enterprise IT systems.
AI Platform for Wealth Management Growth & Operations
Autonomous document and data agents deployed in existing enterprise architecture that get critical jobs done and grow net flows.
Your AI partner,
built for the enterprise.
01
Enterprise-grade security
02
Deploys as a system application inside your enterprise IT environment
03
Works with your enterprise AI subscription
SOC 2 Type 1 and 2
Independently audited
SOC 2 is the industry standard for data security, privacy and operational integrity. In addition to adherence to these standards, LEA encrypts all data in motion and at rest, and provides customizable data retention.
LEA's infrastructure — compute, storage, and networking — runs on AWS regions covered by AWS's own SOC 2 certification, providing a layered attestation chain from the physical data center to the application layer.
LEA's controls map directly to the workflows your team uses: role-based access ensures only authorized users can view extracted data; every extraction event is logged with a tamper-evident audit trail; and data is logically isolated per tenant so one firm's documents never touch another's.
At renewal, LEA proactively shares the updated report and a gap summary so your compliance team can clear the vendor review with minimal friction.
Microsoft Entra ID
One app provisioned by IT. Zero new logins for your team.
LEA deploys as a Microsoft Entra ID app inside your admin console. This means straightforward IT management, rapid deployment into Microsoft 365 applications like Sharepoint and OneDrive, and deep integration via Microsoft Graph API into business applications like Azure, Power BI and Power Automate.
Because LEA lives inside Entra, your security team can audit LEA sign-in logs in the same Entra portal they already use for all other applications. Failed authentication attempts, unusual sign-in locations, and token refresh events are all visible.
Group-based access control is handled at the Entra layer. You can restrict LEA to a specific security group (e.g., 'LEA-Users') and manage membership without touching LEA's own admin panel.
Single sign-on tokens are short-lived and never stored by LEA beyond the active session. Refresh is handled through Entra's standard token rotation mechanism.
Egnyte
Documents stay in Egnyte. LEA reads, extracts, and moves on.
LEA integrates with Egnyte as a registered application inside your domain, reading documents directly from the folders you designate. This means no file migration, no parallel storage, and access scoped by your Egnyte administrator. Webhook triggers process new uploads automatically within minutes of arrival.
Extraction happens in LEA's processing pipeline, not inside Egnyte's compute environment. The file is fetched over an encrypted API call, processed, and the result is persisted. The original document is not copied to any LEA-owned storage.
LEA supports all major document types commonly found in Egnyte for financial services: PDF statements, Excel/CSV portfolio reports, DOCX correspondence, and scanned image PDFs (via OCR pre-processing).
Webhook-based triggers provide near-real-time processing: most documents are extracted within 60–90 seconds of landing in the watched folder, so downstream systems are updated the same business day without manual intervention.
LEA deploys as a Box platform application with service-level credentials scoped to specific folder paths. This means extraction happens inside your Box environment, Box Shield classifications and retention policies stay intact, and structured data is written back as metadata on each file for native search and filtering.
Box's event stream and LEA's webhook listener together provide a durable processing queue. If LEA's processing pipeline is temporarily unavailable, Box retries webhook delivery with exponential backoff, and LEA reconciles the backlog on recovery.
Extracted data is returned to Box as structured metadata on the file object, giving your team searchability and filter capabilities inside Box itself — for example, filtering all statements by client ID or reporting period.
LEA's Box integration is compatible with Box's enterprise key management (EKM) feature, which lets your firm control the encryption keys for Box content. Documents are accessed only during the processing window and only with keys that are live at that moment.
Salesforce
Extracted data lands in Salesforce FSC automatically. No copy-paste, ever.
LEA writes extracted data directly into Salesforce Financial Services Cloud through the REST API and a scoped API-only profile. This means no copy-paste or swivel-chair work. Documents arrive, LEA extracts, and structured values map straight to FSC objects like Financial Account, Holdings, and Client, with Duplicate Rules preserving record integrity.
Before writing any record, LEA performs a duplicate check using Salesforce's Duplicate Rules engine, so extraction events never create orphaned records or overwrite data that a human advisor has manually corrected.
LEA supports both upsert and create-only write modes. In upsert mode, re-processing the same document updates the existing record rather than creating a duplicate. In create-only mode, LEA flags the record for human review if a match is found, preserving the advisor's most recent edits.
All Salesforce API calls made by LEA are logged in LEA's audit trail with the Salesforce record ID, object type, field names modified, old values, and new values. Your compliance team can reconstruct the full data lineage for any FSC record back to the source document.
Anthropic Claude
The model purpose-built for careful, accurate reasoning on complex documents.
LEA deploys Claude skills and agents that work with your firm's enterprise Claude subscription.
For documents where extraction confidence falls below a configurable threshold (default: 95%), LEA flags the record for human review in a queue rather than automatically posting to Salesforce or your data warehouse. This catch-and-review loop means the system gets better over time as human corrections feed back into the prompt library.
LEA's Claude integration is model-version-pinned during onboarding. When Anthropic releases a new model version, LEA validates extraction accuracy on a holdout document set before promoting the new version to production for your tenant.
If your firm has an existing Anthropic enterprise agreement, LEA can be configured to route API calls through your organization's own API key, which means all usage accrues against your agreement's volume commitment and pricing tier.
OpenAI
What clients say
We've valued LEA's expertise and collaborative approach in tackling complex document AI challenges. Their speed and industry knowledge made them a natural fit for our team.
As a high-growth firm, we've evaluated productivity solutions for our financial planning team for years. LEA is the first one that meets our specific needs, and deploys directly inside our enterprise systems like Egnyte. Our team members level-up their contribution to the business, and I learn something new from the LEA team every time we speak.
Susan Brandeis Chief Financial Planning Officer, Pure Financial
We've valued LEA's expertise and collaborative approach in tackling complex document AI challenges. Their speed and industry knowledge made them a natural fit for our team.
As a high-growth firm, we've evaluated productivity solutions for our financial planning team for years. LEA is the first one that meets our specific needs, and deploys directly inside our enterprise systems like Egnyte. Our team members level-up their contribution to the business, and I learn something new from the LEA team every time we speak.
Susan Brandeis Chief Financial Planning Officer, Pure Financial
