The national telecommunications carrier of a small island nation.
About a thousand chats a day, handled the same day.
Support staff open each chat with a summary and draft reply already prepared.
Tell us which project relates to your team and what you want to improve.
Our team will reply within one business day.
At a glance
Volume
~1,000 chats/day
Across the carrier’s primary channel
First-response time
Minutes → under 30 sec
Through the day, including peak
Routine chat resolution
60–70%
Without agent typing; repeat contact unchanged
Staff handling time
15–25% faster
On the cases that reach a human
A thousand customer chats a day across a small, dense market. Most of them were the same handful of questions: bill enquiries, prepaid top-ups, plan changes, roaming, SIM activations, outage check-ins. Each one looked routine until you’d answered it for the eighth time that morning.
Agents handled what they could during the day and the evening peak built into a queue that the overnight shift inherited. By the weekend, genuine service incidents were mixed in with sixty repeat questions about the same evening’s mobile data top-up.
A chat support assistant.
Handles the first stage of incoming chats only. No billing changes, no service credits, no plan adjustments, no enterprise accounts.
Tools and information
- Customer account database (read-only)
- The chat platform
- Knowledge base and standard reply library
- Outage and incident status feed
- Rules for passing cases to the right person
- Read each incoming chat and sort it into one of the six main categories
- Check the account details before drafting: plan, recent usage, last bill, and latest service activity
- For routine questions with a clear answer, draft a reply in the team’s usual tone
- Pass service credits, complaints, enterprise accounts, and outage compensation to a person with a one-line explanation
- Give the support agent the details and a draft to review
Once the system was handling real chats, the categories needed refining. It identified prepaid top-ups and bill enquiries well, but sometimes confused roaming and plan questions and drafted the wrong answer. The sorting rules were tightened using those real examples, and fewer drafts answered the wrong question.
Support agents were also trimming routine replies that were correct but too wordy. Drafts became shorter, clearer, and closer to the team’s usual tone. Cases needing a person came with a one-line explanation of why they needed help and what was already known.
- Billing adjustments, credits, refunds
- Plan and service changes
- Enterprise and business-to-business accounts
- Complaint resolution and any escalation
- All replies during an active outage, when an incorrect reassurance could make things worse
- Reading the chat and classifying it
- Pulling account context before drafting
- Drafting the routine reply in the right voice
- Detecting outage patterns and pausing drafts
- Passing unusual cases to a person with the reason attached
For context, published reports of comparable telecom AI systems describe these ranges: first-response under thirty seconds during peak, 60–70% of triaged chats resolved without an agent typing, and 15–25% faster handle time on cases that reach a human, with no increase in customers contacting support again about the same issue.
- Touch billing. Adjustments, credits, refunds: all human.
- Change a plan or service. The employee can describe options; only an agent makes the change.
- Handle enterprise accounts. Business-to-business enquiries go straight to a person, with no exceptions.
- Speak during an active outage. Outage-pattern detection holds drafts and routes to a human.
- Send without review. Every customer-facing draft passes through an agent.
Agents now open each chat to a pre-built brief: customer context, recent activity, draft reply already prepared for the routine categories, edge cases pulled into the right queue with a summary attached. The evening peak still happens (it always will), but it clears inside the evening instead of waiting for the next morning. The work that needed a person (the complaints, the outage conversations, the enterprise escalations) is the work agents are actually doing.
Chat #2814: prepaid data top-up question, account in good standing, drafted from top-up template. Held for Aishath’s approval.
Chat #2901: roaming rates for travel next week, plan and destination matched to knowledge-base entry. Draft held for agent review.
Other teams using the same approach.
Energy · legal & regulatory drafting
A tier-1 Australian energy retailer.
First drafts that used to take days are now ready in under an hour. The in-house legal and regulatory team reviews and revises instead of retyping.
Read the caseHigher education · course advice & customisation
A leading Australian provider of online higher education.
Clearer course explanations and study options tailored to prospective students. Advisers who used to repeat the same unit explanations now spend more time on conversations that need their judgement.
Read the caseHave a similar task? Tell us what needs doing.
Tell us which project relates to your team and what you want to improve.
Our team will reply within one business day.