AI Development in Singapore

No GST is charged on our invoice, and the invoice says so. Whether you account for it under reverse charge depends on your input tax position: IRAS's guidance is that a GST-registered business making only taxable supplies is not impacted, a partially exempt business or one with non-business receipts accounts for GST on imported services in its own return, and a business that is not registered should ask its accountant, because an exempt business importing more than S$1 million of services in a year can be required to register. We would rather your accountant confirmed your position before the first invoice than reconciled it at year end.

We build to it rather than claiming a certificate. The consent, purpose limitation and notification obligations shape the forms and the data model; retention limitation means records are deleted on a schedule rather than kept indefinitely; access and correction requests are features with a queue behind them rather than a manual database job. Processing happens in India, so the transfer limitation obligation applies, and the processor agreement we sign sets out the comparable standard of protection the Act asks for. The obligation to design for early is breach notification: an organisation has three calendar days from determining a breach is notifiable, which means one likely to cause significant harm or one affecting 500 or more people, to tell the PDPC, and answering what was in the breach is only possible if the logging existed beforehand. Appointing a Data Protection Officer is your obligation rather than ours, and we will say so if a design choice would make that person's job harder.

Not PSG. The Productivity Solutions Grant funds pre-scoped packages from pre-approved vendors for Singapore-registered companies, and a custom build from an overseas team is not that shape, so we will not pretend it is. The Enterprise Development Grant is assessed project by project against its own criteria, and whether a build like yours qualifies is a question for Enterprise Singapore or your grant consultant to answer before the scope is written rather than after a budget has assumed it.

We can supply what a due-diligence pack asks for, accept audit and access rights in the contract, and name every sub-processor that touches your data. What we will also tell you in the first conversation is what we do not hold: we are not SOC 2 or ISO 27001 certified, and our security page says so in those words. Whether an engagement with us is a material outsourcing arrangement under the MAS guidelines, and what that requires of you, is your compliance team's determination, and it changes the contract before it changes the code, so it belongs in week one.

Often not, and we will tell you before you spend anything. Roughly half the briefs that reach us are better solved by a database query, a rule, or a clearer form. AI earns its place when the input is genuinely unstructured and the output is genuinely open-ended. If your problem is deterministic, a model just makes it slower, costlier and less predictable.

Grounding and checking. The model answers only from context we retrieved from your data, we validate the structure of what comes back, and we make it cite the source so a human can verify it. For anything where a wrong answer is expensive, we design a human approval step rather than pretending the problem is solved. No prompt makes a model incapable of being wrong.

That depends on model choice, context size and volume, and it is one of the first things we size rather than one of the last. We build per-request cost tracking and hard budget ceilings in from the start, and we route cheap requests to cheap models. A typical grounded support assistant runs a few thousand rupees a month at moderate volume; a heavy document pipeline can run considerably more, which is exactly why it gets measured before launch.

We use enterprise API tiers where prompts are not used for training, and we scope precisely what leaves your systems. Where data cannot leave at all, we build on open models you host yourself. This gets settled in scoping, not discovered during a security review.

Model Context Protocol is a standard way to give an assistant controlled access to real tools and data, so it can query your systems instead of guessing. It is useful when you want an assistant to actually do things rather than only talk about them. If you just need a grounded chatbot over a set of documents, you do not need it.