Sino Lanka / STS Group · Vertical AI TeaserSESSION 01 · FINANCIAL SERVICES
Financial services, under quiet attack
For the leadership of LankaBangla Finance and LankaBangla Securities. The pressure on NBFI margins is not a cycle. It is a structural shift, and it is accelerating. This session is about what the disciplined response looks like.
Audience
CEO and Senior Operating Leaders
Entities
LankaBangla Finance & Securities
Duration
30 Minutes · 25 + 5 Q&A
You Leave With
A Working Morning Briefing
The Arc · 30 Minutes
How The Next 25 Minutes Run
Five beats. Tightly run. You leave with something you can use tomorrow morning.
0:00 — 3:00The Honest OpenWhere the gap actually is in lending right now, and what it costs a regulated NBFI to be slow.
3:00 — 8:00One Industry, A PatternTonik as the modern playbook at your scale, First Hawaiian as the proof, one efficiency kicker. Three is a pattern, not an anecdote.
8:00 — 15:00Three Strategic PlaysBuilt on the data and infrastructure you already have. Mapped to Finance and Securities.
15:00 — 25:00Built LiveA morning briefing for the CEO, configured in front of the room. Yours to keep.
25:00 — 30:00ConversationReal questions. A clear path to a deeper engagement when you are ready. No hard ask.
Beat 01 · The Honest Open
Six months ago the gap between lenders who had moved on AI and those who had not was a matter of efficiency. Today it is a matter of who the customer reaches first.
This is not a pitch about technology. It is an honest read of where your industry already is, what your competitors are already doing, and what the disciplined response looks like for an institution carrying regulatory weight and a real balance sheet.
Beat 01 · The Shift
What has changed in lending: underwriting is moving from a five-day decision to a thirty-second one.
The competitor who approves a small-business loan in thirty seconds, using mobile money flows, behavioural data, and a risk model, does not need to outprice you. They need to be in front of the customer first. Five days into your process, that customer is already gone.
bKash, Nagad, and digital-first lenders are not your future competition. They are your current competition, and they are compressing NBFI margins in a way that does not reverse. LankaBangla has the data, the Tier II infrastructure, and early AI already in motion. What is missing is the strategic frame that turns those into a coherent advantage.
Beat 02 · One Lender, Your Scale · Tonik
Tonik — a BSP-licensed digital bank in the Philippines — rebuilt its entire loan origination on AI scorecards, fusing 1,000+ signals to underwrite borrowers a national model could not see.
100%
Loan origination on AI scorecards
Real-time + batch decisioning
~2.2×
Predictive power vs. legacy scorecards
Same applicants, better separation
1,000+
Data points fused per decision
Device, network, repayment context
This is not a tier-one bank with a thousand-person AI division. It is a regional digital lender serving underbanked Filipinos — the same shape of customer, the same shape of balance sheet, the same regulator-facing posture as an NBFI. The method is what travels.
The result is the shape of the curve, not the size of the number: 10× the loan portfolio in three years, with fewer ops staff, not more.
10×
Loan portfolio growth
Last three years
−20%+
Direct ops servicing staff
Over the same period
>US$20M
Projected ops cost savings
Next three years, ops-AI program
Operating leverage is the tell. When AI is doing real work, every new peso of lending drops more to margin — Tonik's own COO words. That is the line a regulated NBFI board can underwrite.
The reason this matters is not that they had a large budget. Modern AI has collapsed the cost of doing this.
Tonik · Operating Spine
AI is giving us tangible savings — not hype. It lets us scale profitably without adding staff, so every new peso of lending drops more to margin.
Tomasz Borowski · COO, TonikOfficial statement, November 2025Source ↗
What required a dedicated data science team and a multi-year build two years ago is now configurable in weeks, at a fraction of the cost. Foundation models, managed feature stores, and packaged decisioning platforms have moved the floor.
The barrier is no longer resources. It is deciding to start — and then running it with the discipline of a regulated balance sheet rather than a startup growth bet.
PartnerZEST·AI
Beat 02 · Proof At Your Scale · First Hawaiian Bank
First Hawaiian — Hawai'i's oldest bank, 48 branches and $24.9B in assets, where nearly half of customers are Asian or Pacific Islander — was being underwritten by a national scoring model built for someone else's borrower.
>90%
Applications sent to manual review
Layered credit policies to cover the model's blind spots
18–24mo
To build a custom model in-house
Bank-only data · still blind to half the customer base
0
Room left to scale safely
Every new application meant more manual work, not more growth
A national scoring model is tuned for the average American borrower. FHB's borrowers are not the average — and the model could not tell a safe applicant in Honolulu from a risky one. So underwriters had to. Less nuanced data in, more manual review out, and a hard ceiling on how fast — and how safely — the bank could grow.
Beat 02 · Proof At Your Scale · First Hawaiian Bank — with Zest AI
FHB partnered with Zest AI for a custom ML model trained on Hawai'i, Guam, and Saipan data. Six months to production. Automated decisioning went from 4% to 55%. Approvals up 25%.
Automated decisioning
4%→55%
13× in twelve months
Approvals
+25%
More yes — to the right borrowers
Time to production
6 mo
Initial development to live decisioning
13× more decisions automated. 10× more applicants getting an instant yes. Underwriters freed for the complex cases, not removed. Accounts approved on Zest scores outperform score-exception accounts 4× on delinquency.Same pattern as Tonik — modern AI, configured in months, with the regulated balance sheet kept intact. Zest is a partner LankaBangla can evaluate directly.
JPMorgan's COiN reviews 12,000 commercial credit agreements in seconds. The manual equivalent: 360,000 lawyer hours a year.
Manual · per year
360,000
lawyer hours
≈ 180 full-time lawyers · 12,000 contracts
→COiN
With AI
30 Seconds
to review all 12,000
Fewer servicing errors · lawyers freed for negotiation
Annual time to review 12,000 commercial credit contracts · ABA Journal, JPMorgan ChaseSource ↗
Point AI at the document bottleneck every lender carries and the throughput math stops being marginal. It becomes categorical.
Beat 02 · One Anecdote vs A Pattern
One bank is an anecdote. Three is a pattern. Here is the same result at a fraction of the size.
01
The modern playbook · Tonik, Philippines
A BSP-licensed digital bank, not a global one. 100% of loan origination on AI scorecards fusing 1,000+ signals, with ~2.2× the predictive power of legacy models. 10× loan portfolio in three years with fewer ops staff, not more. Built in the era where AI is configurable, not custom-built.
A 48-branch bank, not a global one. AI underwriting lifted loan approvals by 25%. Automated decisioning went from 4% to 55%, a 13x jump. Instant approvals went from 4% to 40%. Live in production in six months, with underwriters freed for the complex cases instead of removed. Same pattern, different geography.
An AI document platform reviewed 12,000 commercial credit agreements in seconds. The same work by hand was roughly 360,000 lawyer hours. One line, but it lands: this is what happens when AI is pointed at the document bottleneck every lender carries.
The thin-file borrower is not a limitation. In your market it is the opening.
A 2025 World Bank study of 27 emerging-market financial authorities found that institutions in markets like yours are more likely to use AI for credit scoring and underwriting, precisely because they need to serve customers without formal credit histories. The data gap that looks like a constraint to a Western bank is, here, the reason to move first.
Every digital lender already understands this. The question for this room is not whether the method works. It is whether LankaBangla applies it before bKash and Nagad finish doing it to the same customers.
Beat 03 · Three Strategic Plays
Three strategic plays, built on what you already have.
Not productivity tips. Strategic plays the leaders in this room can recognise as their own.
01
AI-Assisted Underwriting
An underwriting assistant that synthesises an applicant's history, behaviour, and external signals into a risk view your credit officer reviews, not one it replaces. Decision time compresses from days to hours. The credit officer's judgment becomes the differentiator instead of the bottleneck. This is precisely the Tonik pattern, configured for your book.
Where it landsLankaBangla Finance
02
Early-Warning Portfolio Surveillance
Most defaults signal months in advance. The signals are already in your data: payment patterns, account activity, sector developments, public news. AI synthesises them into a weekly portfolio risk brief for the CRO. The exposure gets intervened on at month two, not month nine. The ceiling on this is high — Tonik and FHB both proved the operating leverage is real.
Where it landsFinance & Securities
03
Cross-Portfolio Customer Intelligence
The LankaBangla and STS co-branded Mastercard already connects finance to the education portfolio. The data layer linking those customers across deposits, securities, fee payments, and lifetime value is the most undervalued asset in the group. AI is what turns a card relationship into a working customer platform.
Where it landsGroup-Wide
Beat 04 · Built Live
Before we close, I am going to build one of these in front of you.
Not a dashboard. Not a project that lands in three months. A short briefing in the CEO's inbox before the day starts, that reads in ninety seconds and tells you the three things that have moved that matter to this portfolio. Configured live, in this session, and yours to keep when we close.
The point is not the tool. The point is that the distance between curiosity and something useful is ten minutes, not a transformation programme.
The Take-Home · Your Morning Briefing
The Financial Services Daily Briefing. In the CEO inbox. Before the huddle.
Configured live, tailored to your portfolio. The three things that moved. Ninety seconds to read. The forty-five-minute manual scan becomes a two-minute review before the leadership huddle.
Daily Briefing · Illustrative Structure
01Risk signals since the last briefing, read against the shape of your portfolio and top exposures.
02Digital lender and mobile money moves: bKash, Nagad, and digital-first competitors.
03Regulatory, rate, and currency developments relevant to the book.
Illustrative structure. Real signals configured to your portfolio, built live in this session.
Closing · Session 01 · Financial Services
If today was useful, the next conversation is not a bigger version of today. It is a structured roadmap.
If You Want To Go Deeper
A focused risk-modelling pilot inside one product line, scoped to one quarter.
If You Want To Move The Group
The AI Transformation Roadmap: governance, enablement, deployment, then the flywheel your team owns.
If You Want To Think About It
Run the briefing for a week. Let it earn its place. We talk when you are ready.
Scott Walker · UpShift Collective · scott@upshiftcollective.comSino Lanka / STS Group · May 2026 · Confidential