About PAND.AI
We were doing this when it was still called chatbots.
Since 2016, inside banks, insurers and law firms across Southeast Asia. Long enough to have made every generation of this technology work in production.
Our Journey
We were early, repeatedly.
Firsts, certifications and a licence. Each one is the reason the next was possible.
2016
Founded
2017
Launched TOMI
Singapore's 1st Insurance Chatbot, with Tokio Marine Life.
2018
Launched Schroders GO
Asia's 1st Asset Management Chatbot.
2019
ISO 27001 certification obtained
The international standard for information security management.
2020
Investment by Bualuang Ventures
The venture arm of Bangkok Bank, following our work with the bank in Thailand.
2021
First IMDA Open Innovation win
The first of three, working with public agencies and consumer bodies on live problem statements.
2022
MAS broking licence
Granted after GINA graduated from the MAS FinTech Regulatory Sandbox, making us a licensed broker rather than a vendor to one.
2023
WhatsApp Business Solution Provider
Accredited by Meta to build and operate on the WhatsApp Business Platform, the channel GINA and much of our client work runs on.
2024
OSPAR certification obtained
The Association of Banks in Singapore audit standard for outsourced service providers to financial institutions.
2025
Launched A&GEL
Our first on-premise sovereign AI deployment, with Allen & Gledhill.
2026
IMDA Technology Partner
Under the GenAI × Digital Leaders initiative.
Why Us
Two things you cannot buy quickly.
01
Craft
10 years of shipping, not prototyping
The gap between a demo and a production system is engineering design: how knowledge is retrieved and grounded, where the permission boundary sits, and how the system fails safely. Those calls are learned by making them on real systems, not by reading about them.
02
Track Record
What survives first contact
We have been at this long enough to know what actually moves a business number, what people will use twice, what governance a risk committee will sign, and what quietly falls over in month three.
A consultancy sells the expertise again on every project. A platform company sells software and leaves the hard part to you. Our engineers arrive standing on everything the last ten years taught us.
Our experience does not reset when a project ends. It compounds.
How We Work
Five steps, and none of them end with a slide deck.
The work is judged by one thing: whether the system is still running a year later.
We learn how your business actually runs
Before any design, time with the people who do the work and the numbers that work moves. What we are looking for is the judgement that makes you good at this, which is rarely written down anywhere.
We design for the outcome, not the demo
The system is shaped around the result you need: the number that has to move, the people who have to adopt it, and the controls it has to clear. Every technical decision follows from those three.
We ship a working system in six weeks
Not a prototype and not a slide. A working system on your data, inside your environment, narrow enough to be honest about. Six weeks is possible because Platform 5.1 is already there.
We stay in hypercare for three months
The first weeks of real use tell you more than any pilot ever will. We keep engineering against what your people actually do with it, so what you end up with is better than what we shipped.
We leave you an AI that helps you win
Not a pilot that impressed a steering committee. A system your people use, your risk team has approved, and your own engineers can inspect.
Founder’s Note
PAND.AI started here.
I left banking in 2013 to start my first fintech company. It failed. By the end, I had about $2,000 left in the bank and a lesson I didn’t want to learn twice:
Start with a real problem, not the technology.
So when I started PAND.AI here in 2016, we did it the other way round. We had a signed deal for our first chatbot before we had written a line of code.
It ran on technology that was cutting-edge then and looks primitive today. Over the next decade, the models changed beyond recognition. But better models alone never guaranteed better outcomes.
The last two years made that plainest. A&GEL took all of them, inside Allen & Gledhill, before it was good enough to put our name to. The model was never the hard part. What took the time was everything around it: the permissions, the routes back to source, and learning how their lawyers actually build an argument.
Every deployment since has taught us something we didn’t want to relearn. Our Platform is where we keep those lessons.
The technology still changes every few months. The way we work hasn’t. Our clients bring deep expertise in their business. We bring deep expertise in AI engineering. Neither half is sufficient on its own.
Ten years in, we are still learning. We are running the same AI transformation inside PAND.AI that we build for our clients, because understanding it from the outside was never going to be enough. Staying in the trenches is the only way I know to stay useful to them. I write up what I’m learning as we go at StartupUncle, rather than waiting until I have the answers.
The first conversation is not a pitch.
Tell us what you are trying to change. If we are not the right answer, we would rather say so early than bill you to find out.
Based in Singapore · Serving Southeast Asia