“It does not confirm or cancel an order, set or change a price, create a product or a customer, send a message, close a deal or delete a record. It cannot read a record the person asking is not permitted to see, and it will not invent a fact to fill a gap.”
6 things.
And what each one is for.
Reading orders, not deciding them
It is built to read a WhatsApp message or a purchase order and propose the items, quantities, units and dates, each shown beside the words it came from. Matching to your products and pricing each line are done by fixed rules, not by the AI. Order reading arrives with the Orders module, which is taking pilot customers.
Benefit It is built to take the typing. The judgment stays with your team.
Covers: Proposed lines · source shown · rule-based matching and pricing · person confirms
Customer and lead summaries
It is built to summarise a lead or an opportunity on request, with the specific fields and activities it read to produce it. It runs inside the asker's permissions, so it cannot summarise a record they are not allowed to open.
Benefit A seller enters the next conversation informed without rereading every activity.
Covers: Summary · fields and activities read · permission filtered
Priority and next-best-action
It is built to rank which work looks most worth doing next, so the queue is ordered rather than alphabetical, and to propose the next action on a record.
Benefit The day starts from a considered list instead of the loudest item in the inbox.
Covers: Ranked queue · proposed next action · human decision
Follow-up drafting
Follow-up emails, campaign copy, the message you have been putting off since Tuesday. Every draft is marked as AI-generated. Approving one records the decision. KeelSG does not send it: the person sends it themselves. There is no setting that changes that, because it is not a setting.
Benefit Writing time goes down without handing the customer relationship to a machine.
Covers: Draft · marked AI-generated · approval recorded · never sent by the system
It says when it does not know
When the data is thin or out of date it reports that the evidence is insufficient. It will not invent a company size, a revenue figure or an intention the customer never expressed. Text arriving from outside the system, in a customer message, a purchase order, an imported email or a note, is treated as data and never as an instruction, so a message written to manipulate it does not get to.
Benefit That rule is the difference between an assistant and a liability.
Covers: Insufficient evidence response · outside text treated as data
AI work queue and audit trail
Recommendations land in a decision queue with accept, reject and failure states, and every AI action writes an append-only audit event carrying who ran it, what it touched and whether it succeeded.
Benefit AI work is managed like accountable work rather than hidden magic.
Covers: Queue · accept/reject · terminal failure · append-only audit
Four steps, in this order, every time.
- 01
Ground
Load only the records this user may already see.
- 02
Propose
Return a draft order, a summary, a ranking or a reply.
- 03
Review
A person edits, approves or rejects it.
- 04
Record
Write the audit event, or fail without changing anything.
Thirty minutes, and you leave with three answers.
Whether a pilot fits the way your orders arrive, what we would set up, and what it would cost. Written down, so you can put it in front of a partner or a board.
Bring three things
- A few real orders, as they arrive today.
- Your product and price list, in whatever form it is in.
- Who takes and confirms orders now.
Leave with three answers
- Whether the pilot fits.
- What we would set up, and how.
- What it would cost, written down.