No, and the system prompt is configured to prevent it. Quoting a specific rate or estimating a borrowing figure for an anonymous visitor on the basis of casually stated income is exactly the kind of behaviour that creates regulatory exposure for a mortgage lender. The bot describes general program features and fee categories that the lender already publishes, and routes every rate-quote or borrowing-capacity question to a licensed MLO.

No. The bot describes what FHA, VA, conventional, jumbo, and state-specific programs are at a general level and what categories of borrower each one typically serves, but it does not recommend a specific product for a specific visitor. Product recommendation requires a licensed MLO making a personalised determination based on the borrower's full file, and the bot routes that decision to a real loan officer with the conversation context attached.

The bot is configured to identify itself as a non-licensed automated assistant and to surface the lender's NMLS ID and the assigned loan officer's NMLS ID when bookings are made. The lender's footer and loan-officer pages already carry NMLS information per regulation, and SleekAI can be configured to surface the same identifiers in the chat handoff so the disclosure trail is intact.

Yes. Add the fee categories your Loan Estimate disclosures already publish (origination fees, lender credits, third-party services) as a page, custom post type, or ACF fields. SleekAI reads them through your WordPress structure and passes them into the prompt as context. The bot describes the general fee categories without ever quoting a specific dollar figure or rate for a specific borrower.

Bring your own key for OpenAI, Anthropic, Google Gemini, or OpenRouter, plus any OpenAI-compatible endpoint such as Azure or self-hosted models behind a compatible proxy. Most lenders run GPT-4o or Claude Sonnet because those models follow strict no-quote, no-recommendation system prompts more reliably than smaller models, which matters when the cost of a hallucinated rate is regulatory and not just reputational.

Yes. SleekAI ships with 28+ display conditions so the lender can show the bot only on the program pages, hide it on the loan-officer bios, or trigger different bots per program. A common setup runs an FHA-focused bot on the FHA page, a VA-focused bot on the VA page, and a generalist bot on the homepage that triages to the right MLO after the first qualifying question.

Logs live in your WordPress database and you control retention. Most lenders export them to their CRM or compliance archive on a recurring schedule via webhook. That keeps chat transcripts inside the same archival workflow as other regulated communications, which is what state regulators and CFPB examiners expect to see in a record-keeping review. Retention windows are typically aligned with the lender's existing communications policy.

SleekAI runs in WordPress and logs every conversation there. Most lenders push the transcript and contact details into Encompass, BytePro, or a CRM like Velocify via Zapier or a webhook. The conversation summary plus the qualifying answers (purpose, property type, state, credit-range bucket) end up as a fully shaped lead record so the MLO sees the context before the consultation begins.