Capstone: design an AI-driven NPS recovery program for your circle
اختتامی منصوبہ: اپنے سرکل کے لیے AI سے NPS بحالی پروگرام
40 min read
Three ways to see it
NPS, Net Promoter Score, asks subscribers a single question: 'How likely are you to recommend us, on a scale of zero to ten.' Promoters are nines and tens. Detractors are zero through six. NPS is promoters minus detractors as a percentage. A negative score means more people would warn their friends off you than would recommend you. In a market with frictionless MNP, this is how you bleed. Recovery is not magic; it is the disciplined application of every previous lesson in this track to the actual root causes that made promoters into detractors.
Way one to design it: start with the diagnosis, not the prescription. Pull every NPS comment from the past six months. Tag each by topic using lesson three's Urdu NLU intent classifier. The likely top three categories in any Pakistani circle are voice quality on specific routes, data slowdown in evening hours, and care experience after a complaint. The fourth and fifth might be unexpected, like a confusing JazzCash bundle or a fraud experience. Until you know what is actually killing the score, every intervention is theatre.
Way two: triage interventions by lesson. The voice-quality complaints map to lesson one's per-sector anomaly detection: identify the worst sectors, raise them with the network team, and visibly fix the top ten. The data-slowdown complaints map to lesson six's planning model: identify the over-loaded sites and add a sector or upgrade backhaul. The care-experience complaints map to lesson four's care copilot: deploy agent-assist on the queues that handle that circle, with personal AHT scorecards. The unexpected fraud complaints map to lesson five: tighten Wangiri detection and proactively credit affected subscribers. The unexpected JazzCash confusion maps to lesson seven: redesign the in-app bundle screen with bilingual clarity.
Quick check
Quick check: what makes modern AI different from a rule-based program?
The why-tree
Why-tree level one: why NPS as the headline metric and not churn? Because NPS leads churn by three to five months in a Pakistani context. By the time churn moves, the damage is done. NPS gives you a window to act before subscribers actually port.
Try this with Claude
AI-edge prompt: 'I am the new CX head at a Pakistani operator. Draft a 90-day NPS recovery program for one circle that integrates per-sector anomaly detection, churn prediction, Urdu IVR, care copilot, fraud reduction, AI-driven planning, and Customer 360 under PDP Act 2025. Output: one-page program, dashboard schema, weekly cadence, day-30 CEO email, and the three things you would explicitly NOT do because they are vanity. Then critique your own program and tell me where it is most likely to fail.'
Sources
Sources and further reading. Reichheld and Markey, The Ultimate Question 2.0 on NPS. Bain and Company NPS prism methodology. Pakistan Telecommunication Authority Quality of Service annual reports. Zong China Mobile Pakistan, DeepSeek partnership announcements February 2026. Jazz parent VEON AI strategy investor materials. Telenor Group customer experience reports. McKinsey, Telco of the Future. BCG, AI in Telco Customer Operations. The previous seven lessons in this track for the underlying techniques.