RFPX: submission intake that scales without adding people.
For carrier and MGU intake teams. Agents read every submission as it lands, structure it into a case, and tell you what you are looking at before anyone keys a field.


The problem
Intake only scales by hiring.
In busy season an intake team takes hundreds of emails a day, each carrying a different set of unstructured attachments. Someone opens every one, reads it, works out what is actually there, then pivots to the underwriting system and keys it in — most of an hour per submission. Volume rises, headcount rises with it, and margin goes the other way. When the team can’t keep up, good business never gets quoted. Worse, nobody knows whether a submission is worth quoting until the keying is already done.
AI at work in RFPX
Agents act. Workbenches show. People decide.
Every check is designed by people who know stop loss, not borrowed from a generic document reader.
Agents act
Agents sweep the inbox, classify every attachment, and capture values exactly as they appear on the page — then normalize them into the fields your systems expect.
Workbenches show
The case arrives assembled: every field traced to its source document and colour-coded by confidence, with NIGO gaps and guideline exceptions surfaced on arrival.
People decide
Your team validates the AI results, confirms the exceptions, and reviews DTQ flags — in the first hour, not the third day.
Review goes to the person who used to key the data, not the underwriter. Every AI action is logged, reviewable and subject to human approval.

Capabilities
Everything the intake team needs, before keying.
Human-in-the-loop review
Low-confidence fields route to the person who used to key the data, not to an underwriter.
Extract & normalize
Values captured exactly as they appear, then normalized into the fields your systems expect. Both are kept, because no two brokers present data the same way.
Document classification
Every file typed and indexed. We never split, edit or re-save your documents — classification is metadata, not alteration.
Deficiency tracking
NIGO gaps surfaced to the intake team the moment a submission lands, not after it has been keyed.
Email sweep
Submissions ingested straight from the inbox, attachments and all.
Appetite & guideline checks
Every submission measured against your underwriting guidelines, so quote-or-DTQ happens early instead of after the work.
RFPX on IPAS
Open by design.
Open APIs
Every screen in the intake workbench runs on APIs you can call yourself.
Your data and your documents
Cases, extracted fields, decisions and the source files, on demand. Your data leaves as easily as it arrives.
Connected data
Claims history enriched through Gradient and Curv before the case reaches underwriting.
Built on IPAS · Powered by Amazon Bedrock
- Submission receivedSystem
- Then, in parallel:
- AI Agent
Classify documents
Types and indexes every attachment on arrival
- AI Agent
Extract & normalize
Captures values as they appear, then normalizes them
- AI Agent
Check guidelines & gaps
Flags NIGO gaps and appetite exceptions
- AI Agent
- Human
Intake workbench
Validate AI results, confirm exceptions, review DTQ flags
- Case ready for underwritingSystem
- System step
- AI agent step
- Human step
Part of MSLX
Works on its own. Better together.
Start with RFPX on its own, or pair it with UWX so every validated submission lands in underwriting ready to quote.
- RFPXSubmission intakeYou are here
- UWXUnderwriting
- ClaimsXClaims
- ClinicalXClinical review
- MarketXBroker marketing
All MSLX solutions run on IPAS · Powered by Amazon Bedrock
There were times when we couldn’t even get to every submission coming through the door. Now we have the capacity to look at every opportunity without staffing up for busy season.
Days to hours
time for a submission to be ready for underwriting
See MSLX at work in your process.
A demo built around your underwriting, claims, clinical or distribution team.