AI Tools

The AI suite Polly can't ship.

63 agent tools in GA. Natural-language pricing. A config optimizer that explains every change. Anomaly narratives for your capital-markets desk. LO copilot for every loan officer.

📊 Pricing Explainer

Every best-ex result comes with a plain-English narrative: which LLPAs fired, which investor won, which margin stack applied.

"At 80 LTV + 740 FICO, Investor X wins by 8 bps because their 720+ FICO LLPA is -12.5 bps vs. Investor Y's -4 bps."

🎯 Config Optimizer

AI-suggested pricing-config changes based on your pull-through history. Every suggestion comes with a projected margin impact and an undo button.

"Lowering your 60-day extension fee from 12.5 bps to 10 bps is projected to boost lock-to-fund by 3.2%. Accept? [Yes] [Modify] [Reject]"

🤖 LO Copilot

A per-LO agent that knows the LO's pipeline, comp plan, and lock queue. Answers scenario questions, drafts borrower emails, escalates lock exceptions.

"Hey, what's my 30-year fixed for 740 / 80 / 500K in CA with AHL's overlays?"

🚨 Anomaly Narratives

When an investor's rate sheet diverges from the cohort, we flag it with a story: "Investor Y's 15-year fixed is 18 bps rich vs. FHLB 15Y — either their margin widened or there's a parse error."

"⚠️ 2026-04-14: Investor Z's 7/6 ARM is 22 bps rich vs. cohort. Reason: pricing sheet dated 4/11 — likely stale upstream."

💬 NLI Assistant

Natural-language pricing queries. Type (or say) "30-year fixed, 740 FICO, 80 LTV, California, conventional" and get a structured best-execution result in JSON.

"what's my best rate for conforming 30yr, 740, 75 ltv, cali, 400k?"

🔧 63 Agent Tools

Full function-calling catalog exposed to GPT-4o (and any LLM): get_best_execution, list_investors, explain_llpa, simulate_scenario, submit_lock_request, and 58 more.

GET /api/v1/admin/internal/agent/tools → {"tools":[...63 items...]}

Try it live

The chat-demo loads our agent inside an iframe so you can try it without signing up.

Demo iframe placeholder — launches to /agent/chat-demo