Spread every statement in one pass.
Kita reads the borrower’s statements, general ledgers and bank data and returns a normalized period-over-period spread. Debt service coverage, margin and trend are computed in one pass, each figure cited to the line it came from.
Spreading is the slowest part of a credit file.
It is also the part with the least judgment in it, which is why it is worth automating first.
The rekeying is the job
An analyst retypes the same tax return, P&L and bank statement into a template before any judgment happens. The work is slow, and it is the part most likely to introduce an error.
Every borrower files differently
Periods do not line up, charts of accounts disagree, and the same line is named three ways across three businesses. Normalizing that by hand is where the hours go.
Templates break on real documents
Template-based extraction expects a layout. Scans, photos, foreign statements and anything non-standard fall out and go back to manual.
The spread is stale by the time it is read
When a spread takes days, the borrower waits and the credit committee reviews numbers that have already moved.
What Kita reads to build the spread.
PDFs, scans, photos and screenshots. No template per layout, which is what usually sends a non-standard file back to manual review.
Tax returns
Form 1120 and 1120-S, Schedule C, K-1s, 1065 partnership returns, quarterly 941s
Profit and loss
Monthly, quarterly and annual statements, management accounts, year-to-date figures
Balance sheets
Assets, liabilities and equity, with period-over-period movement
Bank statements
Business checking and savings from national banks, regional banks and credit unions
General ledgers
Account-level detail, journal entries and trial balances
Receivables
AR aging, concentration by customer, days sales outstanding
What the spread gives your analyst.
Computed from the spread, not retyped from a prior memo.
Period over period, with the direction of travel made explicit.
Existing obligations read from the file, not asked for again.
Every figure links to the document line that produced it.
The spread is the first step, not the whole file. What the numbers do not say is handled by credit analysis, and the memo your underwriter signs is drafted by the AI Underwriter.