An Agnostix business
It knows the network, not just the conversation.
SaidaAI is the engine inside SaidaLink and LitSpan Africa. It works from the same subscriber record as billing and the NOC, and does the first pass of the work an operator's team would otherwise do by hand.
Where it works
Support
A subscriber writes on WhatsApp in Swahili: "no internet". SaidaAI recognises the language and finds the account. It checks the balance and due date and runs a connection diagnostic, then answers with the actual cause. If a person needs to step in, the ticket arrives with the full thread on it. It works the same way over SMS, web chat and voice, in more than 15 languages.
Billing
When the subscription has lapsed, SaidaAI sends the mobile-money payment prompt inside the same conversation and confirms the payment there. The subscriber is reconnected automatically when it clears.
A payment that arrives on the paybill with a mistyped account number does not sit in a queue. SaidaAI suggests the likely account and puts it in front of finance, and nothing is applied until a person confirms.
NOC
An OLT port goes dark. SaidaAI correlates the alarms so that one fault does not turn into fifty tickets. It finds the customers on that port and the route they share, and drafts the outage notice for the NOC to approve.
Provisioning
A technician registers an ONU. SaidaAI checks the OLT port, the VLAN and the light level, and says what is wrong before the technician leaves the site.
How it is plugged in
SaidaAI runs inside each operator's platform, not beside it. It reads the same billing, RADIUS and device records the operator's staff see, and it opens tickets in the same queue. When a conversation needs judgement, it hands over to a person.
Each operator's data stays with that operator
Each operator's subscriber data stays on that operator's own server. SaidaAI requests carry only the context a task needs, and one operator's records are never used to answer another operator's customer.
Metered per tenant
Every request is logged with its tenant, its purpose and the tokens it used, so an operator can see what SaidaAI did and what it cost.
Under the hood
SaidaAI runs on Azure AI Foundry. Requests are classified by intent and routed to one of three model deployments sized for the task. Every call is recorded for audit and quality review.