1. Why you need a sharper checklist for “income AI”

The market is crowded. Document AI vendors, OCR platforms, LOS add-ons, and niche startups all claim to “automate income.” Without a clear framework, it’s easy to:

• Buy a tool that only extracts data but doesn’t make defendable decisions.
• End up with an “AI layer” that your underwriters don’t trust and always override.
• Invest in a platform that can’t keep up with guideline changes or niche products.

The goal of this checklist is simple: help you find a vendor that behaves like a dedicated income & Smart Conditions engine, not just a fancy scanner.

2. Question 1 – Are you a vertical income engine or a generic doc AI?

What to ask:

• “Which industries do you serve?”
• “How much of your business is mortgage and consumer lending?”
• “Do you support income for FHA, VA, GSE, non-QM, reverse, DSCR, etc.?”

What you’re looking for:

• A clear focus on lending and mortgage, not 10 unrelated verticals.
• Evidence they’ve built domain-specific logic, not just templates for a few forms.
• Language that sounds like your underwriters, not generic “AI for documents” marketing.

A true income automation vendor talks about FHA, VA, GSEs, and overlays as first-class citizens, not edge cases.

3. Question 2 – How do you handle FHA, VA, GSE, and product guidelines?

What to ask:

• “Where does your guideline logic live? Can you show it to us at a high level?”
• “How do you update FHA, VA, Fannie, Freddie, and investor overlays when they change?”
• “Can we configure lender-specific overlays and see where they apply in the calculation?”

What you’re looking for:

• Concrete description of guideline packs or rule libraries.
• A clear process for staying current with agency changes.
• Support for your own overlays with traceability (what rule came from whom).

If a vendor can’t clearly articulate how they encode and maintain guidelines, your underwriters will end up treating the system as “just another opinion,” not a source of truth.

4. Question 3 – How transparent and auditable are your calculations?

What to ask:

• “Can we see the full math behind each income decision, not just the final number?”
• “How are lookback periods, averaging logic, and stability checks documented?”
• “Can we export the calculations for audit or investor reviews?”

What you’re looking for:

• A clear, human-readable report that shows:
  • each income source,
  • how it was averaged,
  • which months/years were used,
  • and why some sources were excluded.
• Ability for underwriters to drill into assumptions and adjust if needed.
• Export options for QC, audit, and investor conversations.

If the system is a black box, your best people will not trust it. Transparency is non-negotiable for both adoption and risk management.

5. Question 4 – Do you generate Smart Conditions or just flags?

What to ask:

• “What does the system output beyond the income number?”
• “Do you generate underwriter-style conditions, or only generic warnings?”
• “Can conditions be tailored to our credit policy and investor expectations?”

What you’re looking for:

• Conditions that sound like a UW wrote them:
  • “Provide 24 months of bonus history to support inclusion of bonus income,” not “Warning: low history”.
• Coverage for common issues: missing docs, gaps, declines, inconsistencies, high-risk patterns.
• Ability to map conditions into your LOS and pre-UW checklists.

Flags are easy to generate and easy to ignore. Smart Conditions directly change file quality, suspense rates, and investor confidence.

6. Question 5 – What does your human-in-the-loop and QC model look like?

What to ask:

• “How do you handle messy, non-standard, or low-quality documents?”
• “Do humans review files? If so, when and how is that process controlled?”
• “What SLAs and accuracy guarantees can you commit to?”

What you’re looking for:

• A clear explanation of human-in-the-loop for edge cases and low-confidence outputs.
• Defined QC checks that run before the file is returned to you.
• Service-level expectations for turnaround time and quality.

Pure automation is a myth for complex income. The question is whether the vendor has an organized, measurable QC layer — or whether all exceptions spill back onto your team.

7. Question 6 – How do you integrate with our LOS, POS, and doc flow?

What to ask:

• “How do docs get from our LOS/POS into your system?”
• “How do results get back — PDF reports, data fields, conditions, all of the above?”
• “What does implementation look like for a mid-size lender vs. a large IMB?”

What you’re looking for:

• Multiple integration paths: direct LOS integration, SFTP/API, and secure web portal.
• Minimal IT lift for initial rollout, especially for pilot programs.
• Support for different channels (retail, wholesale, broker, reverse) without separate projects.

A strong vendor will meet you where you are: starting with a simple workflow and growing into deeper integrations once value is proven.

8. Question 7 – Can you prove ROI with real metrics and references?

What to ask:

• “What impact have you had on time per file, suspense rates, or loans per UW?”
• “Can you share anonymized before/after data or case studies?”
• “Can we talk to existing customers that look like us?”

What you’re looking for:

• Specific before/after metrics, not vague statements.
• Case studies with complex income profiles, not just simple W-2 scenarios.
• Referrals from organizations with similar products, volume, and channels.

If a vendor can’t point to measurable improvements in capacity, suspense, or risk, you’re likely paying for a tool your team will quietly work around.

9. How to run a practical vendor evaluation (without derailing ops)

Once you’ve shortlisted vendors using the seven questions, the next step is a controlled trial on real files. You don’t need a six-month project.

Here’s a simple pattern:

• Step 1 – Pick a cohort of files
  30–50 loans with a mix of W-2, variable, self-employed, and niche products. Include some that caused internal debate or investor pushback.
• Step 2 – Run in parallel
  Have your team run their normal income process while the vendor runs the same files through their engine.
• Step 3 – Compare outputs
  Look at:
  • Time to get a result.
  • Differences in final income numbers.
  • Conditions raised or missed.
  • Clarity of reports for underwriters and QC.
• Step 4 – Involve underwriters and QC
  Let them review results and provide feedback on trust, clarity, and usability.
• Step 5 – Decide using data
  Use a small set of metrics (time, touches, suspense, defects) to make a decision — not just slideware.

This approach gives you a realistic view of vendor performance without disrupting your pipeline.

10. How Rapidio answers these seven questions

Rapidio is built as a vertical income & Smart Conditions engine for mortgage lenders and brokers. Here’s how we line up against the checklist:

• Vertical focus: Mortgage income only — FHA, VA, Fannie, Freddie, reverse, DSCR, and more.
• Guideline logic: Dedicated guideline packs with configuration for your overlays and transparent updates.
• Transparency: Full income reports with step-by-step math, lookback windows, and rationale per source.
• Smart Conditions: Underwriter-style conditions for missing docs, gaps, declines, and inconsistencies.
• Human-in-the-loop QC: Structured review to ensure 100% checked, guideline-aware output.
• Integration: Flexible intake (portal/API) and LOS-friendly outputs, starting with fast pilots.
• ROI focus: We help you measure time per file, touches, and suspense to prove value quickly.

Choosing an income automation vendor is not just a tech decision. It’s a decision about how your entire organization will treat income, risk, and file quality for years to come. These seven questions are designed to make that decision clearer — and safer.