AI Powered QC Workflow

The AI Quality Control (QC) toolkit in Jaro enables our customers to perform AI-powered checks, calculations, and decision-making chains based on QC rule sets. This empowers QC users and appraisal desk teams to streamline workflows, enhance delivery speeds, and quickly resolve issues. Customers using this toolkit have reported reduced backlog, the ability to identify potential errors before underwriter revision requests, and improved lender allocation due to better delivery rates.

AI in QC

Artificial intelligence has the potential to significantly enhance QC accuracy and workflows for appraisal desk and AMC teams. When properly managed, AI can execute complex QC rule sets and workflows rapidly, serving as a powerful copilot for QC teams.

Why It Matters

Cross-checking an appraisal report involves numerous rule sets. These range from basic checks like: “Does the borrower’s name in the sales contract match the appraisal report?” to more intricate, multi-layered questions with various dependencies.

Adding complexity are requirements based on loan types or sources, such as FHA loans, or lender-specific directives to meet internal compliance or process standards. The specific questions depend on appraisal standards, government regulatory organizations like HUD, the FHFA/GSEs, and client preferences.

The new AI features in the QC tool allow Jaro users to create multi-step QA flows with AI, ensuring responses are grounded in factual data about the report or property details. This enhances QC workflows while maintaining quality outcomes, providing a reliable AI copilot to appraisal desk and QC teams who remain in the loop for final approval and checks.

Question Structure

The Jaro QC tool organizes questions based on standard QC checklists. These include:

• Out-of-the-box Jaro standard questions and sections.
• Customizable questions tailored for specific product types (e.g., 1004, 1079 appraisals, 1004D reports, etc.).

Watch this workflow in action in this detailed product demo.

How It Works

The QC process flow in JaroDesk involves predefined checks that a QC user must review before approving a report or returning it to the vendor for clarification. Once these checks are marked as passing, the report can be finalized.

Appraisal QC rules are categorized as either Passing or Failing. In Jaro, results are presented to QC users, who can accept and confirm the AI outcome or reject it as incorrect. This process is illustrated in the following example:

1. A question about the report is passed to the AI tool along with relevant data and images.
2. The AI provides a pass/fail response, supported by evidence.
3. A QC team member either confirms or rejects the result based on the information or further investigation.

This simple evaluation covers basic checks typically required for an appraisal report. For more complex checks, Jaro employs calculations, evaluations, and conditional logic to determine next steps. Below, we explore these features in detail.

AI Evaluation

AI evaluates data in the report against QC checks, determining true or false outcomes. For instance, it can verify whether the address on the report matches the sales contract.

AI Calculation

This feature asks AI to perform calculations based on report data. These calculations can analyze subject data or verify accuracy, reducing the need for manual checks.

By combining these capabilities, Jaro enables complex QC workflows involving cascading calculations and evaluations.

Example Workflow:

[
 {
   "prompt": "Evaluate the range of adjusted sales prices as a percentage of price, please show the formula used for evaluation",
   "type": "calc"
 },
 {
   "prompt": "Is the range difference less than 30%?",
   "type": "eval"
 },
 {
   "prompt": "Does the appraiser make a comment that addresses why a range of greater than 30% was used?",
   "type": "eval"
 }
]

Cascading Checks & Logic

Jaro introduces chains of logic to address scenarios where multiple conditions must be evaluated sequentially, as you can see from the the example workflow above.

This allows for combining calculations, evaluations, and conditional logic to meet diverse QC requirements across various loan and product types and can be done by non-technical users in the QC builder.

Logical operators include:

• Manual verification.
• Continued evaluation.
• Pass/fail determinations based on true/false statements.

Using these combinations, any logical flow can be tailored to meet QC needs, and the results of each step can be passed in the chain for further evaluation or calculation. 

Example Use Case: Are Utilities Connected and Working?

The AI QC feature automates checks by evaluating if the occupancy status on page one aligns with the photos. It also examines adverse conditions, external factors, or land use changes and whether the appraiser addressed them.

For instance, Jaro’s tools can check if the appraiser indicated that all utilities were on and working. This workflow accounts for exceptions, such as new construction where utilities might not yet be active but are accounted for in the valuation. Multi-step prompts enable AI to assess scenarios like these accurately, leading to nuanced QC checks.

Why Jaro QC?

Jaro’s human-in-the-loop AI design ensures guardrails are in place, improving QC outcomes for AMCs, vendors, and lender teams. This results in:

• Increased accuracy.
• Reduced turnaround times.
• Higher confidence in the appraisal process.

For a detailed demonstration, watch the in-depth product demo: Jaro AI QC Toolkit Demo.