Crayon Data
Tangram AI
LEANM AI for Financial Services
Accenture FILAP 2026
The Story1 / 11
120 days ago, I walked into a bank in Kochi to conduct an AI workshop.
We identified a universal problem in banking.
Today, it is built, solutioned, sold, contracted, under deployment.
AGL (Accelerated Go-Live) is the future present of agentic AI in banking.
120 · 12 · 122 / 11
120 · 12 · 12
120 daysFrom problem to live deployment
12 monthsBuilding agentic AI for regulated industries
12 yearsDeploying heuristic AI inside large banks
1 thread that connects these.
The Thread3 / 11
My Journey
A 40 year backward progression —
sales to marketing to advertising to media to....
RedPillFounded 2000, sold to IBM, 2009
→
IBMGlobal analytics leader, 2 years
CrayonFounded 2012, 3 patents, graph-AI pioneer
12 Years4 / 11
12 Years
Most AI companies sell a future.
Crayon Data provides a track record of numbers.
The maya.ai personalisation platform.
3 patents. TasteGraph with a billion entities.
HDFC Bank65M+ customers personalisation · Incremental impact last year of $600M spends and $6M revenues
ADIB50+ agents across 3 suites
KBZ PayZero to 450M transactions/month · two-thirds of Myanmar GDP
12 Months5 / 11
12 Months
Customer experience and maya.ai built Crayon Data.
Agents and tangram.ai are building what's next.
Operations and Productivity as focus.
100+ manual controls made agentic at ADIB.
200+ agentic AI in the Tangram AI store, 150+ live.
The Problem6 / 11
The Problem
1000s of notices a month. Five days to respond.
Personal liability if you don't.
VerifyIs the notice even genuine?
InterpretMap it to the regulation and bank policy
ResolveRoute to the right team, follow through
Today: spreadsheets, CRM, and a wall of people. Every bank. Every month. 5,000 to 50,000 notices.
120 Days7 / 11
120 Days, Start to Deploy
7 days to prototype.
120 days to solution, sell, contract, and deploy.
Demo
→
Co-develop
→
Sandbox
→
Security
→
Contract
→
Deploy
Our library was the difference.
Domain knowledge
Data models
Agentic capabilities
Security credentials
Business Engineers
LEANM AI8 / 10
The Solution, Live
LEANM AI
Ingest → Extract → Classify → Enrich → Generate → Route → Resolve → Track
Next 12 Months9 / 10
The Next 12 Months
We didn't just rethink the product.
We rethought the company.
~$5M annual revenue
Profitable, EBITDA+
~200 agents live
80 business engineers
The Close10 / 10
We sell to the most regulated industry on earth, and we like it.
We like to handle complexity.
We want to accelerate go-live.
Simplify success in enterprise AI. One agent at a time.
Crayon Data / Tangram AI · Accenture FILAP 2026 · Singapore | India | Dubai