Expert Decision: your best underwriters' judgement on every case
Build your credit policy once, without code. Send applicant data by API and get approve, decline or refer back in seconds, with the limit, the grade and every reason behind it.
- ~10 min
- from policy change to go-live
- < 1 week
- to launch a new loan product
- 24/7
- decisions, after hours and on holidays
The challenge
Credit decisions still depend on who picks up the file
Policy lives in manuals and in the heads of senior underwriters, so every case waits for a person and every change waits for IT.
- Consistency
- Decisions vary by reviewer
- Two officers can read the same policy differently, which leads to inconsistent approvals and weaker portfolio quality.
- Speed
- Every case waits in line
- Checks run one after another during office hours, so hire-purchase customers can wait 1–3 days for an answer.
- Agility
- Policy changes wait for IT
- A new rule or product becomes a release project, and launches take 1–4 months on a traditional system.
What it does
Your credit policy, as rules your team owns
The credit team builds and changes the policy itself. Every version is simulated on past cases before it goes live.
Policy rules
Pre-screen, hard and soft policy rules, set up without code and versioned from draft to publish.
Scorecards
Internal and alternative-data scorecards, banded and weighted the way your risk team designs them.
Decision and limit
Approve, decline or refer, with the credit limit and risk grade calculated in the same call.
Every reason returned
Each factor, band and point comes back with the result and is logged for audit.
Proven in production
Same-day hire purchase at a top-5 HP lender
Results from the bank's post-implementation study, with Expert Decision running inside Thinker LOS.
- 5 min
- to approval, down from 1–3 days
- 4 hrs
- to disbursement
- 67%
- fewer tasks per loan officer
A new approval model
Approve first. Verify automatically, in parallel.
Low-risk cases are approved at once and sampled afterwards by risk. All the data checks your policy requires still run before approval.
Traditional
Office hours; every case waits
- 1 · Apply
- 2 · Verify every case
- 3 · Approve
With Thinker
Instant, 24/7
- 1 · Apply
- 2 · Approve by rules
- 3 · Post-audit by risk
People on hard cases
High-risk or out-of-policy cases go to a person in the Work Pool, under maker-checker.
The same criteria every time
Your best underwriters' rules apply to every application, whoever is on shift.
Explainable afterwards
Risk-based post-audit, with every decision traceable to its rules and data.
AI inside, experts in control
AI does the busywork. Your experts make the call.
Thresholds you set decide what goes straight through and what goes to a person. An officer can review and override any automated result.
01
AI reads and checks
AI-OCR and eKYC capture the data, and AI asks the applicant for anything missing from their profile.
02
AI suggests a decision
It reads the case against your rules and recommends approve, decline or refer, and shows why.
03
Your team decides
Unclear or high-risk cases go to a person, who can override any result under maker-checker.
04
Every decision logged
Automated or human, the audit trail shows who decided, what changed and why.
Works without AI, too
Every AI step has an expert-rules mode, so lending keeps running.
Any AI provider
We plug in leading AI services rather than training our own, so there is no lock-in.
Where it runs
Inside Thinker LOS, or behind any channel by API
The same engine decides for every channel, so the policy lives in one place.
With Thinker LOS
Decide inside the workflow
Comes with Thinker LOS: the workflow calls NCB and AMLO in parallel, then the engine approves returning customers automatically.
By API
Answer your own channels
Any app, web journey or existing LOS sends applicant data by API and gets back the decision, the stage results and every reason.
AI Assistant
Configure rules by chat
Product and risk teams set up rules with the AI Assistant, simulate them on past cases, and approve each change before it goes live.
Buyer questions
How this is different
What risk and credit teams ask when they compare a decision engine with their current process.
01
How is this different from a generic rules engine?
It is built for Thai credit: NCB, AMLO and alternative data arrive through Databridge, scorecards, limits and grades are part of the same policy, and every reason is returned for audit. Business teams change it without code.
02
Is the AI a black box?
No. The decision comes from your own rules and scorecards, and the AI suggestion is shown with its reasons. Every AI step can run on expert rules alone, and a person can override any result.
03
Can we use it without replacing our LOS?
Yes. It answers any channel or existing LOS by API. Running your policy on sample cases and comparing the results with today's decisions is the fastest proof of concept we run.
Frequently asked questions
What credit teams ask first
The practical questions risk, credit and technology teams ask before a demo.
Run your own policy on sample cases
Bring your credit policy and a set of past applications, and see how the decisions compare with today's.