Features & Roadmap

Extract exactly what you need from any document.

Define a schema describing the fields you want- names, dates, amounts, line items, or any custom data, and AISAY returns structured results in a single API call. We combine OCR with  LLMs to understand context, so it finds the right information even when documents use different labels, formats, or layouts. Supports simple fields, lists, and nested objects for complex documents like invoices, timesheets, and annual reports.

Go beyond extraction to evaluation and assessment.

Ask the AI to evaluate, summarise, rate, and provide commentary on document contents. Analyse sustainability reports against ESG criteria, evaluate vendor submissions against specifications, or generate structured assessments of lengthy tenders- all returned as structured JSON or Excel.

Detect anomalies, inconsistencies, and signs of tampering.

Define your verification criteria in the prompt and AISAY checks documents against them at scale. Flag missing information in bank statements, detect suspicious transactions, identify misaligned text or logos that suggest tampering, and run logical consistency checks- with results and reasoning returned for each check.

Automatically sort documents into categories you define.

Send a document along with your list of categories and descriptions, and AISAY returns the best match. Useful for triaging incoming submissions- sort a mix of passports, bank statements, payslips, and NRICs before processing each type differently.

Extract all visible text from any document in clean markdown.

When you need the complete text content rather than specific fields, the OCR API returns everything in structured markdown- preserving headings, tables, lists, and paragraphs. Handles typed, scanned, and handwritten content across PDFs and images.

Handle repeating, nested data structures without writing code.

Documents like timesheets, invoices, and phone bills contain rows of related data. AISAY's Object system lets you define grouped fields (e.g., each line item has a description, quantity, and price) and extract all instances automatically- no matter how many rows the document contains.

Choose the right model for your document type.

Flagship models handle text-heavy documents up to 150 pages with the highest accuracy. Local models are hosted in Singapore, process faster, and perform better on image-heavy and visual content. Try both on your use case and pick the best fit.

Process documents without writing a single line of code.

Any officer with a gov.sg or edu.sg email can log in to the AISAY Web Portal from a COMET device at aisay.ai.tech.gov.sg, write prompts in plain English, upload up to 100 files, and download results as Excel- no registration, no API keys, no technical skills required. Save and reuse prompt templates across sessions.

Purpose-built capabilities for detecting document fraud- going beyond general verification to identify sophisticated tampering, forgery patterns, and document authenticity issues across common government document types.

For agencies processing large or complex documents where synchronous calls may time out, we are building async API support with webhook callbacks- submit a document, receive a job ID, and get notified when results are ready.

API key provisioning will be available directly through GovTech's PlatformAI portal, allowing developers to generate and manage their own staging and production keys without manual requests.