Gives each team member a personal AI agent that coordinates with teammates' agents automatically.
Using agentic workflow orchestration, contextual knowledge retrieval from CRM and docs, personalized message drafting in each user's style, and document classification.

|
Workflow Automation
|
YC W26

Last Updated:
March 19, 2026

Builds personal AI agents that automate team coordination and lending workflows. Each user gets a dedicated assistant that collaborates with teammates' agents for scheduling, handoffs, and document collection, with deep specialization in mortgage and lending.
Deepening lending workflow automation for mortgage brokers and loan officers. Expanding agent-to-agent coordination. Integrating with email/calendar/CRM/document management. Goal of becoming AI operating system for financial services.
GitHub shows C++ language server and VSCode extension development. 'Human-in-the-loop data factories' signal proprietary data flywheel. Deliberate stealth posture ahead of YC W26 Demo Day launch.
<p>AI agents automate scheduling, document collection, and borrower follow-ups across the entire mortgage origination pipeline.</p>
An AI assistant handles all the back-and-forth scheduling, document chasing, and status updates so loan officers can focus on closing deals instead of playing phone tag.
Clice AI deploys a dedicated AI agent for each loan officer that autonomously manages borrower communication, document collection reminders, and internal handoffs to processors and underwriters. The agent learns each officer's communication style and borrower preferences, drafting personalized follow-up emails and texts. When a borrower submits a document, the agent automatically routes it to the correct processor, flags missing items, and updates the loan file status. Agents coordinate with each other across the team—if a processor is overloaded, the system can redistribute tasks or escalate to a manager. Over time, the platform builds a proprietary dataset of successful loan workflows, enabling increasingly accurate predictions of bottlenecks and proactive interventions before delays occur.
It's like giving every loan officer a hyper-organized executive assistant who never sleeps, never forgets a document, and somehow gets along with everyone else's assistant too.
<p>Personal AI agents answer routine internal questions by querying teammates' agents, eliminating repetitive Slack messages and email threads.</p>
Instead of pinging your coworker for the same status update for the fifth time, your AI agent just asks their AI agent and gets you the answer instantly.
Each Clice AI agent maintains a contextual model of its user's current projects, availability, and domain expertise. When a team member asks a question—such as "Where is the Smith appraisal?"—their agent first checks its own knowledge base, then communicates with relevant teammates' agents to locate the answer. The system uses retrieval-augmented generation (RAG) to pull from shared documents, CRM records, and prior conversation history. Agents learn which questions are routine versus sensitive, routing appropriately and respecting permission boundaries. Over time, frequently asked questions are cached and surfaced proactively, reducing the total volume of human-to-human interruptions and creating an ever-improving institutional knowledge layer.
It's like having a personal librarian who also happens to be best friends with every other librarian in the building—and they all gossip productively.
<p>AI agents draft and send personalized borrower follow-ups in the loan officer's voice, keeping deals moving without manual effort.</p>
Your AI writes follow-up emails to borrowers that sound exactly like you—except it never procrastinates and never forgets.
Clice AI's agent observes each loan officer's historical email and messaging patterns—tone, vocabulary, sign-off style, timing preferences—and builds a personalized generative model. When a borrower hasn't responded to a document request or rate lock decision, the agent automatically drafts and (with configurable autonomy) sends a follow-up in the officer's authentic voice. The system tracks borrower engagement signals (open rates, click-throughs, response times) and adjusts messaging cadence and content accordingly. For high-value or sensitive communications, the agent surfaces a draft for human review before sending. This creates a closed-loop system where every borrower interaction is timely, on-brand, and data-informed, directly accelerating pipeline velocity.
It's like cloning yourself—but only the version of you that's really good at writing emails and never gets distracted by lunch.
<p>AI agents automatically collect, classify, and validate borrower documents to prepare complete underwriting packages with minimal human intervention.</p>
The AI checks every document in a loan file before it goes to underwriting, catching missing pages and mismatched numbers so humans don't have to.
Clice AI's document intelligence module ingests borrower-submitted files (pay stubs, tax returns, bank statements, IDs) and applies ML-powered classification to sort each document by type, extract key fields (income, employment dates, account balances), and cross-validate data against the loan application. The agent flags discrepancies—such as income on a pay stub not matching the stated amount—and automatically requests corrections or additional documentation from the borrower. Once the package is complete and validated, the agent assembles it in the format required by the specific underwriting system and notifies the processor. This dramatically reduces the back-and-forth cycle of conditional approvals and re-submissions, compressing the timeline from application to clear-to-close.
It's like having a perfectionist intern who reads every single page of every document, highlights every mismatch in red, and never once complains about it.
Lance built prediction-market infrastructure at Kalshi and founded one of Canada's top university AI research orgs (wat.ai). Zachary is a YC alum who co-founded Traverse. Rare blend of production ML engineering and academic rigor for messy lending workflows.