Your suppliers already reply: by email, by WhatsApp, with a scanned delivery ticket or with a spreadsheet. The problem was never a lack of information, but that nobody could read all of it in time. That is what the AI does in Activa. What it does not do is decide on its own.
From an email or a WhatsApp message it identifies which order and which lines it is about, proposes the status from your company's catalog, the date and the reason. Every proposed data point comes with the supplier's literal sentence that backs it.
From a PDF or a photo it takes the number, the date, the carrier, the pickup date and the quantity of each row, and says which order line each one belongs to. What it reads only adds: it never overwrites what the supplier wrote in the text.
When the supplier sends an Excel file with their twenty lines, each row is located by order and line and is born as a separate proposal. What cannot be located is said, not guessed.
"Mid-November", "Friday at the latest", "15 more days", a handwritten stamp that says "16 10 26". Everything is converted to a date, with your operation's fixed rules above the model.
"As of today Juan replaces me", "I no longer work at the company", a signature with a new cell number. The AI proposes the addition or removal and the reason. A person approves it.
Replies to the supplier by WhatsApp or email, in your operation's tone and rules, for a person to review and send. If something needs escalating, it picks up the previous thread instead of opening another.
On a line is approved in a review queue, with the message, the attachment and the proposal side by side.
And every addition whose identity is inferred.
An escalation to the buyer, a reply to a supplier who asked something, an extra reminder.
Switching off pilot mode, enabling automatic upload, changing an activation rule.
The queue is built for decisions. What the master already knows and did not change does not show up. What the supplier already said is not asked again.
Your operation's rules apply on top of any analysis, whatever the model says. If a data point contradicts them, the rule wins.
Every approval records who decided, when and with what evidence. Every closure records why it was closed.
Conversations are kept for 7 days, and up to 14 when a pending decision cites them. We do not hoard your suppliers' email.
About the models. Activa uses Anthropic language models through their commercial API. Your conversations are not used to train models.
In 20 minutes we look at which channels your suppliers reply through, how much of that is processed by hand today and how much would change if every reply arrived already read, with the proposed status and the evidence in view.
Activa Proveedores