LiveIn production with banks and fintechs across PH, ID, MX, and US.See markets

Underwrite borrowers anywhere in the world in minutes.

Kita is the AI-native platform for modern underwriting. We turn fragmented borrower data, manual review, and hard-to-verify documents into decision-ready credit analysis — built for emerging markets and underserved communities in the U.S., with final credit decisions kept with your team.

Live inPH·ID·MX·ZA·US
Lending typesSME lendingConsumer lendingCDFIsCommercial lending
100K+borrower files processed
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
IMB
Kontempo
N90
Trusting Social
beloz by amiloz
Cashalo
Where credit teams get stuck today

Files sit incomplete.

Good borrowers walk.

Analysts chase missing pages, re-key statements, cross-check IDs against tax filings, and watch for forgery line by line. The credit thinking comes last.

Sharpest where tickets are large and documents are local: SME, asset finance, commercial. Just as real where borrowers are thin-file and inputs are messy. Same root cause: borrower documents don't fit the systems built to read them.

01

Files arrive incomplete

Borrowers send the wrong documents, miss pages, or upload what you can't read. Each round of follow-up adds days to the file.

02

Documents don't fit your tools

Local formats, handwritten records, photos at angle. Templates break. Generic OCR drops fields.

03

Reconciliation is manual

Statements cross-checked against IDs, tax filings, and applications by hand. Hours per file before any credit work begins.

04

Verification is uneven

Forgery and tampering are easy to miss without consistent, document-by-document checks.

The platform

AI-native from the ground up.

Adapts to every underwriting scenario.

One stack, three layers

From intake to signed memo — three connected layers, no tools bolted on.

01 ·Handles the documents in lending

Kita Capture

Vision-language understanding for messy financial documents.

Reads GCash screenshots, M-Pesa statements, Mexican comprobantes, payslips, financial statements, and the long tail of local formats your borrowers actually send. Runs standalone for lenders that just need extraction, or powers AI Credit Officer and AI Underwriter on top.

  • Line-level and field-level extraction.
  • Validation and reconciliation built in.
  • Works on photos, scans, screenshots, originals.
  • Handles non-standard, local, and lender-specific formats.
LiveVision-language extraction
1.4s
G
GCash
9:42 AM
Available balance₱4,218.50
Sari-sari salesMar 14 · 09:42
+1,200
Send to ate LizaMar 13 · 18:11
−800
Sari-sari salesMar 13 · 14:02
+950
Cash-in 7-ElevenMar 12 · 11:08
+2,000
Vision readPhoto at angle.Read perfectly.
Structured fields12 of 12 ✓
account_holderMaria Santos
available_balance₱4,218.50
period2024-03-01 → 2024-03-31
transactions42 entries · 28 in · 14 out
monthly_inflow₱82,400 median (3 mo)
tamperingnone detected · score 0.04
Account matchBalance reconciled1 cross-doc flag
View JSON →
LiveConversational document collection
3m 12s
MS
Maria SantosWhatsApp · Active
File · APP-2419
Kita AI10:42
Hi Maria! I'm helping with your ₱200,000 loan application. I just need your last 3 months of GCash statements to keep things moving.
Maria10:44
Hi po! Pwede ba screenshot na lang? Hindi ako marunong mag-export.
Kita AI · Tagalog10:44
Sige po, screenshots ok lang! Kailangan ko 3 buwan — mula Enero hanggang Marso. I-send niyo isa-isa, aayusin ko na.
Maria10:51
gcash-jan.jpg+ 2 more
Kita AI10:52
Got all 3 screenshots — readable, no edits detected. Last thing: a photo of your government-issued ID, front side.
Borrower file6 of 7
borrowerMaria Santos
channelWhatsApp · TL/EN
loan_request₱200,000 · 12 mo
application_formsubmitted
gcash_statements3 mo · readable
proof_of_addressverified
government_idpending
File completeness
78%
Identity verified3 docs collected1 pending
Hand off to AI Underwriter →
02 ·Top of funnel

AI Credit Officer

The 24/7 agent that closes the loan file.

Engages every borrower the moment they apply, on WhatsApp, SMS, or whatever channel they actually use. Speaks their language. Chases documents until the file is complete, then hands a 100% complete file to AI Underwriter.

  • WhatsApp, SMS, email, web chat, same agent, same memory.
  • Multilingual: English, Spanish, Bahasa, Tagalog, +12 more.
  • Fluent in your lending policy, asks the right follow-ups.
  • Hands off a 100%-complete file to AI Underwriter.
03 ·Decision engine

AI Underwriter

Drafts the memo. Your credit officer decides.

Reads every document in the file. Checks for fraud. Structures the numbers against your credit policy. Then drafts a recommendation your credit officer can question, override, or sign, every claim traceable to a source line. Kita does not approve loans. Your team does.

  • Calibrated on your risk policy, not a generic LLM prompt.
  • Fraud detection across forgery, tampering, and impersonation.
  • Every data point cited to its source document.
  • Your team makes every call. Kita just gives them the cleanest possible file to decide from.
Draft recommendationfor credit officer to decide
ApproveKita's draft. Your team signs or overrides.
Confidence 92%·Policy 11 / 12·Fraud low · 0.04
Pending decisionMaria Santos · Senior Credit Officer
Credit-officer summaryMaria Santos · APP-2419

Borrower demonstrates stable monthly income (₱82,400 median, 3 mo) with low cash-flow volatility. DSCR 1.84× sits comfortably above the 1.5× policy floor. No tampering or impersonation signals across submitted documents.

Tenure of 11 months falls 1 month short of the 12-month policy minimum. Recommend approving as a credit-officer exception.

Supporting evidence4 sources · 12 cited claims
DSCR
1.84×Above 1.5× policy floorgcash-feb.jpg · line 14
Inflow
₱82,400Stable across 3 monthspayslip-mar.pdf · pg 1, ln 8
Identity
VerifiedName, DOB, selfie all matchid-front.jpg + selfie.jpg
Tenure
11 months1 month short of policyapplication.pdf · section 3
Decision rests with credit officer. Always.
Open full memo →
Underwriting scenarios

One stack. Every scenario you underwrite.

01 · Use case

Microfinance & small-ticket consumer

High-volume payroll, cash, BNPL, and merchant credit. Reach every borrower fast, validate documents in real time, decide in seconds.

  • Borrower outreach in any language, any channel
  • Sub-second decisioning APIs with fraud signals
  • Webhook delivery into your pipeline
POST /v1/decisions920ms
payslip_aug.jpg✓ EXTRACTED
bank_stmt_q3.pdf⚑ TAMPERED
Built with: AI Credit Officer + Capture
02 · Use case

SME & commercial lending

Spread financials, narrate the story behind the numbers, draft a defensible memo from page one to sign-off.

  • Cash-flow modeling from any statement format
  • Memo drafted in your template, cited to source
  • Underwriter signs in minutes, not hours
CREDIT_MEMO.pdfDRAFTED
Built with: Capture + AI Underwriter
03 · Use case

CDFI & community lending

Underwrite the borrowers bureau scores miss. Non-standard documents from immigrant-owned and unbanked businesses.

  • Reads tax returns, P&Ls, bank statements in any format
  • Cash-flow story when scores fall short
  • Audit-ready memo for the loan committee
COMPLIANCEIn-region
SOC 2 Type IIIN PROGRESS
VPC deploymentAVAILABLE
Built with: Full stack: Officer + Capture + Underwriter
What Kita produces

Decision-ready files,

not raw extractions.

Every borrower file comes out as a structured applicant profile. Documents cross-checked for consistency. Cash flow modeled against your policy thresholds. Fraud and tampering signals tested. A credit memo drafted in your team's voice.

Each finding cites the exact line of the source document it came from, so your credit officer can read a complete file in minutes, override anything the model got wrong, and sign off with the audit trail intact.

Per file, in minutes

Kita completes the application profile so your team can decide. Every claim traceable to a source line.

BORROWER FILE
DECISION-READY
m
Maria SantosSME loan · ₱500,000 requested · Quezon City, PH
FILE #PH-22418
Document review
Information consistent across all 8 documents
Address consistent across declared form, bank statements & utility bills
No fraud or tampering detected on any submitted file
Financials & capacity
12 months of cash flow modeled, ₱218,500 avg monthly inflow
DSCR 1.84, comfortably above your 1.5 policy threshold
Income matches declared figures and tax filings
For your review
One minor inconsistency: declared revenue ₱2.4M vs. AFS ₱2.38M
Kita's draftCredit memo drafted, your officer signs or overrides.
Why now

AI-native underwriting in emerging markets

just became possible.

For decades, underwriting was the part of lending that resisted automation, especially in emerging markets. Documents too varied. Too messy. Too local. Vision-language models change that.

Software can now read a Tagalog payslip, an Indonesian e-wallet screenshot, and a Mexican AFS with the same fluency as a human analyst, at thousands of files an hour, with citations.

Combine that with agents that talk to borrowers, and the full underwriting workflow, intake, validation, analysis, memo, becomes software. Lenders that adopt this layer first will outpace the market on speed, cost, and risk.

Built by a team from
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Stanford University
Apple
Microsoft
Stanford AI
Tesla
NVIDIA
Built for lenders

Backed by Y Combinator.

Built out of Stanford AI.

Enterprise-ready by default

A platform that fits inside your credit policy, your systems, and your compliance posture — without slowing the decision.

i.

Customer-specific deployments

Your tenant. Your keys. Isolated infra.

ii.

Provider allowlisting

Choose the model providers you trust.

iii.

Lender-controlled decisioning

Kita recommends. You decide.

iv.

SOC 2 · ISO 27001

Type II and ISO 27001 in progress.

v.

Borrower data stays in-region

SEA, LatAm, Africa, never leaves the region.

Get started

Underwrite more borrowers.

Faster. With cleaner data.

Talk to us about deploying Kita inside your lending stack, pilots usually launch in under a month.

kitaTAGALOG
adj.
seen; visible; obvious; easily understood
n.
earnings; wages; income; revenue
Book a demo →View documentation