← Selected Work

AI & CX

Vodafone AI Chatbot

I was promoted early from Vodafone's graduate programme to own a €1.2m AI chatbot integration for the customer care department — from product vision to launch.
Role
Product Owner
Timeline
Oct 2022 – Jan 2024
Team
Cross-functional design, engineering & test team
Recognition
Innovation of the Year, Irish Loyalty & CX Awards 2023

Outcomes

€0k
Annual savings in year one
+0%
NPS uplift
0 → 0%
Retention, doubled
0%
Query containment

The context

Vodafone's customer care teams handled enormous volumes of repetitive queries. The business invested €1.2m to leverage emerging AI capability — and needed someone to turn that investment into a product customers would actually choose to use.

The problem

Customers abandoned self-service because it felt like a dead end. Containment without satisfaction just hides failure — the goal was resolution customers preferred.

What I did

  • Modernised a traditional team to agile ways of working.
  • Built the product vision and roadmap bridging business objectives with AI capability.
  • Wrote user stories and specifications through cross-functional collaboration.
  • Established a voice-of-the-customer programme pre- and post-launch whose insights were adopted by multiple departments.
  • Coordinated a dedicated design/development test team to ensure seamless API integration.
  • Presented monthly to C-suite executives.

Key decisions & trade-offs

  • Prioritised the highest-volume, highest-frustration query types first rather than the technically easiest.
  • Held deployment dates by cutting scope, never quality.
  • Measured success on customer outcomes (NPS, retention) rather than deflection alone.

Results

€800k annual savings in year one, NPS up 20%, retention doubled from 40% to 80%, and 70% query containment — recognised with the 2023 Innovation of the Year Award at the Irish Loyalty & CX Awards.

What I'd do differently

2026 postscript: built today, I'd design this on LLM foundations rather than intent-tree architecture — richer conversational repair, retrieval over knowledge bases, and evaluation loops instead of hand-tuned flows. The product discipline — voice of customer, containment with satisfaction, C-suite transparency — would stay exactly the same.