Loading
Because the risk is asymmetric. A confidently wrong figure quoted to a customer or a colleague can be acted on, and a model with broad access is a data-exposure question nobody wants to answer in a review.
So it gets restricted to public content, where it is safe and largely useless.
It resolves permissions the same way a person does — role, team, record and field — so it cannot surface a figure the person asking could not open themselves.
Every answer cites the records it read, and every conversation is logged with those citations, so an answer given in March is reviewable in September.
It will not post a journal, raise an invoice or release a payment. It reads and drafts; anything that moves money is a person's action.
And it is not an advisor. It reports what the ledger holds and does not interpret treatment.
Which invoices are overdue for this customer, what was this month's spend against that cost centre, what did we agree with this supplier — questions that are answerable from the ledger and currently cost somebody twenty minutes.
It drafts chasing messages from real ageing rather than from a template with blanks.
No. It resolves the same permissions the person asking has.
No. It reads and drafts; postings and payments are a person's action.
Yes, logged with the records each answer cited.
Half an hour on your own numbers is usually enough to say whether Pulse is the right place to start.