Four fields, one sentence each - the model finds the balance, the settlement amount, the payment schedule and the expiry date, and points to the sentence behind each figure.
Forbear's AI reads creditor mail and writes settlement terms into your existing CRM.
Letters, portal screenshots, faxed pages, offers buried in a paragraph: the AI turns each into settlement terms, and every figure cites the sentence it came from.
No card. Cancel in a click.
Four vendors will sell you an AI-native CRM this year. Each demo is good, and each one ends at the same question: what happens to nine years of custom fields, creditor mappings and state license logic.
No migration, and no second place a consumer exists. A negotiator accepts each proposed term before anything is written, and if you switch Forbear off, every file is still where it was.
The AI
The AI that reads what templates can't
A third of creditor mail never arrives the same way twice: a portal screenshot, a faxed page, an offer buried in a paragraph, a note typed by whoever took the call. No template fits it, so a model reads it.
A confidence floor - decides which proposals a negotiator reviews now and which documents wait in the queue. The model never writes on a guess.
Better with every correction - each accepted or corrected extraction teaches it a creditor's wording, so a new creditor needs no template written by hand.
Who it fits
This is for a firm that already has a system it does not want to leave
The first three lines are the buyer. The rest should close the page.
This is for you if
- ✓You run 2,000 or more open files in a CRM that has been customised for years by people who have since left.
- ✓Negotiators retype settlement terms out of letters and portal screenshots, and the typing is a real share of the day.
- ✓You have looked at an AI-native CRM, liked the demo, and stopped at what the migration would cost you.
This is not for you if
- ✕You are under a few hundred files. The typing is an hour a week and this will not pay for itself.
- ✕You want one vendor for sales, servicing and reporting. That is the thing we refuse to be.
- ✕Your creditor correspondence arrives as structured data already. There is nothing here for you to read.
The call we made
We will not hold a consumer record, and that costs us deals
A vendor that owns the record owns the migration, and the migration is why nobody bought.
The reason a nine-year-old CRM survives is that the custom fields encode decisions nobody wrote down. A vendor that replaces it has to relitigate all of them before the first useful day.
So the extraction writes into that system and keeps nothing of its own. There is no Forbear pipeline view, no Forbear inbox, no second place a consumer exists.
The cost is ours and it is real. We cannot show you a beautiful product screenshot, we cannot upsell you reporting, and we lose every buyer who wanted one vendor for the whole floor.
How it works
AI extraction, with rules where the format is already predictable
Rules handle the letters a large creditor lays out the same way every time. Everything else is read by the model.
Every step below runs on every document, not only on the first load.
The model reads the mail - Forbear's extraction models read creditor correspondence however it arrives: screenshots, faxes, PDFs with the offer mid-paragraph, call notes.
Rules for the predictable - where a large creditor's format never changes, a rule extracts the terms instantly and no model is needed.
Read from your CRM - documents and file fields come through your CRM's own API, on credentials you scope.
A named reviewer decides - below the confidence floor the document waits in a queue with the page beside the proposed fields.
Write back as the reviewer - an accepted extraction lands in your CRM under the name of the person who accepted it.
Connects to
The systems we connect to, and nothing we ask you to replace
Read on scoped credentials. Write only to the fields you map.
- Salesforce
- Debt Pay Pro
- Forth
- CUBS
- Custom REST
- SFTP drop
- Email intake
- Portal export
Scope
What this does not do, on purpose
What's in scope
- 01Nopayment handling. We never see a payment instrument and never touch a trust account.
- 02Nowrite without a named person. If nobody clears the queue, nothing reaches your CRM, which is the intended failure.
- 03Noguessed field mapping. You decide what we may write to and we will not infer it.
What we don't do
- 01Noconsumer record. There is no file, no pipeline and no inbox on our side to log into.
- 02Nocontact with creditors. It reads what arrived; it does not call, negotiate or accept.
- 03Noadvice. An extraction is what a document says, not a judgment that a settlement suits a consumer.
Numbers
The numbers that hold at any file count
$1,200
per connected CRM per month, flat
- 0234%of incoming documents match no creditor template
- 034fields extracted per document, each with its source sentence
- 0460 secondsfrom a reviewer accepting to the value landing in your CRM
- 057 yearsthe write record is exportable, including after cancellation
Compare
Against the AI-native CRM you were shown
One row goes against us and it is the row a floor manager weighs.
CriterionKeeps your existing CRMPrice you can see before a callOne vendor for the whole floorEvery written value traceable to a documentWorks the week you sign
ForbearYes, it is the only record$1,200/mo publishedNo, extraction onlyYes, with the source sentenceYes, no migration
An AI-native CRMNo, it replaces itDemo requiredYes, sales to reportingYes, inside their systemAfter the migration
A staffing agencyYesQuoted per seatNoNoYes
Pricing
$1,200per connected CRM, per month
Flat. Any file count, any document volume, any number of reviewers.
- Unlimited documents, files and creditor templates
- Unlimited named reviewers, at no per-seat cost
- Write-back to the fields you map, as the reviewer who accepted
- Exportable write record, kept for seven years
You are paying for the extraction and the write record. Not for seats, not for storage, and not for a screen to look at.
The comparison that matters is a negotiator's hour. If retyping settlement terms costs your floor more than fifteen hours a month, this is cheaper than the status quo, and if it does not, it is not.
Get started
- Free historical backfill across everything your CRM already holds, with no document cap
- Billing starts at your first accepted extraction, not at connection
- Month to month, and every value already written stays yours
FAQ
The questions a floor manager asks in the first ten minutes
What happens if the model reads a settlement figure wrong?
Nothing, unless a reviewer accepts it. The proposal sits beside the page it came from and a person confirms each field.
Where does the AI run?
In the United States, under zero-retention terms. Azure AI Document Intelligence parses scans and screenshots and Azure OpenAI reads them, in Microsoft's East US region, and our models are retrained on Azure Machine Learning. Your documents, intake and the review queue stay on AWS in us-east-1.
Do you become the system of record over time?
No. There is no consumer record here to grow into one.
What if we do migrate to a new CRM later?
Connect it and we write there instead. The extraction store and the write record come with you.
Who can accept an extraction?
Only users your firm names. An unnamed user cannot write through us at all.
Connect a CRM and read a year of mail for nothing
One work email. The backfill runs before anything is billed.
We'll come back within one business day with a straight yes or no.