Xect AI combines network, billing, care and digital events into a governed operational profile built from your existing systems, not a replacement for them. It updates as each event arrives, and rep, routing and chatbot systems all read from it, with lineage to the source event for every tracked attribute.
Customer attributes are refreshed by overnight jobs, network telemetry stays with engineering, and each channel keeps its own copy of the customer. Xect AI registers every customer-touching event once, scores it as it lands, and publishes one inference to every surface that needs it.
Fifteen customer activities — from a lost ONT to a compared bill — register as governed events the moment they happen.
Location intelligence joins in: hub troubles per 100, neighbor status, outage state, install NPS, RFS date.
Churn, propensity and health models score the record; multi-shot prompts turn scores into plain-language context.
Customer APIs, the chatbot, the voice router and notifications read the same record and the same score.
Each agent owns one stage in the event-to-inference pipeline above and hands off cleanly to the next.
Builds the customer from events, not exports.
Registers every customer and location event as it happens, resolves the customer to one identity across account, device and session so duplicate records merge into one profile, and versions every change so conflicting values stay traceable to their source event.
Re-scores the customer on every signal.
Reads the fused record on every new event and recomputes churn, propensity and Location Score, explains which signals moved the number, and retrains on a fixed schedule.
Delivers the governed profile to every channel.
Reads the current record and pushes the same context to rep desktop, IVR and chatbot, exposes it through Customer APIs, and logs every read for audit.
Figures below reflect the schema used in this example deployment, not a fixed limit on what Xect AI tracks.
Lost connectivity, light-level changes and bandwidth spikes land on the customer timeline, next to the payment that failed.
Rep assist, IVR routing and the chatbot read the same predicted score, updated as new events arrive.
Customer and location attributes update on the event, not on the batch calendar.
Any value on the record links back to the event, the model and the version that produced it.
Agent Assist, the chatbot and predictive routing are three readers of one record.
When an ONT loses light and the customer opens the app, the record already says "outage, hub H3002, ETA 40 minutes."
See the industry pageWhen a smart meter reports loss of supply and a bill spikes, one record shows both before the customer calls.
Talk to us about utilitiesWhen a policyholder files a claim and searches "cancel" the same week, the record scores the risk and routes a retention call.
Talk to us about insuranceWhen a patient misses a follow-up and messages the portal, the profile flags the gap for the care coordinator.
Talk to us about healthcareWhen a guest reports a broken unit twice in a stay, the loyalty record and the maintenance history meet in one place.
Talk to us about hospitalityOwns the customer experience across every channel, and needs rep, IVR and chatbot to act on the same context instead of three separate guesses.
Owns the scoring models, and needs their outputs to update the customer record directly, not just be summarized in a quarterly report.
Owns routing and deflection, and needs the queue to route on what the customer is about to do, not just which button they pressed.
Owns retention and offers, and needs the driver behind each account's churn risk, not a segment-wide guess.
Book a 30-minute demo and see one event move through the record, the score and the rep's screen.