AI chat and memory
The centre-column conversation is where the work actually happens. Describe what you need, drop in a document, and the AI does the drafting — while its memory quietly learns your client so it stays consistent over time.
The conversation is where you work
Open a task and the centre column becomes a conversation with the AI. Type what you need — "Appoint Jane Tan as a director from 1 August", "generate document", "what’s this company’s next deadline?" — and the AI replies, drafts documents, and moves the work forward. The composer’s prompt changes to match what you have selected, for example "Ask about this task, or say ‘generate document’…".
This is the heart of CorpSec AI’s design principle: the AI does the work and you confirm it. Wherever there is a decision to make, you get a button to press rather than a blank field to fill in — approve, sign, view the document, file it. You can also just chat generally, or ask about a specific company, without opening a formal task.
Documents appear right in the chat
When the AI drafts or updates a document, it shows up as a card in the conversation — with the title, its status, and buttons to open it or download the Word file. Opening it reveals the full formatted document in the right-hand Document tab, where you can review it and make pinpoint edits by selecting text and commenting.
Reviewing, editing and sharing documents is covered in depth in Documents.
Drop a document straight into the conversationLive
A corporate secretary’s raw material arrives as attachments — a passport, an ACRA BizFile, a signed resolution. Instead of re-typing what is in them, hand them to the AI directly. Drag a file (or a whole folder) anywhere onto the conversation and a “Drop documents to file them” overlay appears; release to attach. Or click the paperclip in the composer to pick one or several files.
- Drag or click the paperclipDrag files/folders onto the chat, or use the paperclip to pick them. Each file can be up to 15MB.
- The file is stored and anchored to the taskIt appears as a chip in the conversation so the file — and the fact you provided it — is on the record.
- The AI reads what it canText-based files are read straight away; scanned PDFs and images are recognised through the metered vision pipeline. Recognition results show on the right-panel Insights tab.
On a phone the paperclip opens a small menu — Take photo or Choose files — so you can capture a passport or BizFile with the camera. See Mobile app.
For bulk document intake (a folder or a zip of many files at once, with a cost estimate and auto-filing to companies), use the dedicated Document Inbox.
AI Memory — the AI gets to know your clientLive
Open the Memory tab (under More in the right panel) for any task or company. This is CorpSec AI’s per-client memory: as tasks complete, the AI distils what happened into short, structured notes — so the next time you work on that client, it already knows the context and stays consistent. The header sums it up honestly, e.g. "AI has learned 12 things about this client · 3 new this month · distilled from completed tasks."
Memory is grouped into kinds so it reads at a glance:
- KYC risk — the client’s current risk rating and why.
- Officers — director and secretary changes.
- Shareholding — share allotments and transfers.
- Registered address and Company name changes.
- Filings and Events — what was lodged and when.
- Notes — free-form things worth remembering.
| Source badge | Where the memory came from |
|---|---|
| Manual | A note you added yourself. |
| AI | Distilled by the AI from a completed task. |
| Auto | Written by a workflow rule. |
| ACRA | Derived from registry data. |
Add your own memory with the Add button — e.g. "Client prefers email delivery; signatory is Jane Lim." Each AI-distilled entry links back to the task it was learned from, so you can always trace it, and any entry can be deleted with a write-enabled role.
Memory is stored per company and used to keep the AI’s drafting consistent. Personal identifiers are masked in the AI’s context — see Security & your data.
Frequently asked questions
Do I have to phrase things in a special way?
No — plain English (or Chinese) is fine. Say what you want the way you would tell a colleague; the AI works out which document or action you mean and gives you a button to confirm.
What happens to a file I drop into the chat?
It is uploaded, stored securely and anchored to the task, and shown as a chip so it is on the record. Text files are read immediately; scanned PDFs and images are recognised via the metered vision pipeline, with results on the Insights tab.
Where does AI Memory come from?
Mostly from completed tasks — when a job finishes, the AI distils what happened into a short structured note. You can also add notes by hand, and every entry shows its source and (for AI entries) links back to the task it came from.
Can I edit or remove something the AI remembered?
Yes. With a write-enabled role you can delete any memory entry, and add your own. Viewers can read the memory but not change it.
Is my client data used to train AI models?
No. You accept the data policy during setup, and your data is not used to train models. See Security & your data.
This is a product guide for CorpSec AI. Where a feature runs on demo data or is not yet released, it is labelled as such. Compliance references are general information for Singapore corporate service providers, not legal advice.