Skip to content
PORTAL LABS
ENES

Utilities

Clara

First-line customer service for a utility, on WhatsApp, where its customers already are.

Sector
Utilities
Stack
Next.jsMongoDBWhatsApp Business APIn8nCloudflare
Overview

Receives a utility's inbound contact on WhatsApp, classifies it, answers what it can, and routes the rest with the context already collected.

Utility support is enormous volume concentrated into a few recurring intents, plus a handful that are genuinely urgent. The difficulty is telling them apart fast.

Clara sits on WhatsApp, so there is no app and no portal password. It resolves routine requests and hands the rest to a person with the conversation attached.

Technical challenges

  1. Classification with asymmetric consequences

    Misrouting a billing question costs time. Misrouting a hazard report costs something else, so urgent classes accept more false positives on purpose.

  2. Escalation on signals, not sentiment alone

    Frustration is a weak signal; people are terse when busy. Repeat contact and explicit hazard language drive escalation instead.

  3. Serving people with no account and no app

    A caller may not know their account number or may be reporting for a neighbour. It has to work when the best identifier is a street address in free text.

  4. The unanswered questions are the roadmap

    What the assistant could not resolve is logged, grouped and reviewed. It drives the next round of answers and changes to the process.

Tell us where you are starting from. We take it from there.

Start a project