Customer support · Production

Flutter UK & Ireland automates over 70% of customer contacts with UiPath NLP/NLU, saving £4M+ and raising NPS to +40

The problem

Flutter was struggling to connect with its customers, suffering a low net promoter score and a challenging contact center journey. Initial integration of NLP and NLU tools yielded only 5% automation in year one and an NPS of -10, with no guarantee of customer acceptance.

First attempt

Flutter's first-year AI chatbot deployment achieved only 5% automation and produced an NPS of -10. Deploying an AI-powered chatbot proved technically complex and initial customer acceptance failed.

Workflow diagram · grounded in source
1
Customer chat contact
trigger
“trying to enhance its online chat facility through automation”
2
NLP/NLU language interpretation
ai_action
“UiPath to automate backend processes with natural language understanding bridging the language barrier”
3
Concierge bot resolution
ai_action
“That journey would begin with concierge bots, before adding more predictive capabilities to automate complex processes”
4
Action Center escalation handling
human_review
“Using UiPath Action Center and Flutter's new Escalation Virtual Assistant has done more than just enhance the customer experience—they're delivering commercial value, too”
5
Next Best Action cross-sell
ai_action
“Flutter's AI-powered Next Best Action tool makes it easy for agents to cross-sell at the end of each suitable contact”
6
KYC process automation
ai_action
“AI-powered automation has also helped Flutter solve problems with customer experience around its Know Your Customer (KYC) process. Frustration has turned into satisfaction, with response times slashed from 24 hours to just 15 minutes on …”
Reported outcome

Flutter achieved over 70% contact automation in a single year (75% for Paddy Power), with resourcing requirements falling by around 33%, saving more than £4 million and avoiding £12 million in cost increases.
Online activity and gameplay rose 140% without a corresponding increase in contact center load, NPS jumped to +40, customer transfers fell from 20% to 6%, and KYC response times dropped from 24 hours to 15 minutes.

Reported metrics
Contacts fully automatedmore than 70%
contacts fully automated (Paddy Power)75%
Online activity and gameplay140%
Contact center conversation volume with activity increasewithout a corresponding increase in conversations into the contact center
Show all 13 reported metrics
contacts fully automatedmore than 70%
contacts fully automated (Paddy Power)75%
online activity and gameplay140%
contact center conversation volume with activity increasewithout a corresponding increase in conversations into the contact center
resourcing requirementsaround 33%
cost savingsmore than £4 million
cost increases avoided£12 million
net promoter score+40
customer transfers to other teamsfrom 20% to just 6%
KYC response timefrom 24 hours to just 15 minutes
new automations deployment ratethree new automations every six weeks
year one automation rate (baseline)5%
year one NPS (baseline)-10
Reported stack
UiPathNLPNLUUiPath Action CenterEscalation Virtual AssistantNext Best Action
Source
https://www.uipath.com/resources/automation-case-studies/paddy-power-betfair-adopting-rpa-for-gambling
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Flutter achieved over 70% contact automation in a single year (75% for Paddy Power), with resourcing requirements falling by around 33%, saving more than £4 million and avoiding £12 million in cost increases.

What tools did this team use?

UiPath, NLP, NLU, UiPath Action Center, Escalation Virtual Assistant, Next Best Action.

What results were reported?

Contacts fully automated: more than 70%; contacts fully automated (Paddy Power): 75%; Online activity and gameplay: 140%; Contact center conversation volume with activity increase: without a corresponding increase in conversations into the contact center (source-reported, not independently verified).

What failed first in this deployment?

Flutter's first-year AI chatbot deployment achieved only 5% automation and produced an NPS of -10.

How is this customer support AI workflow structured?

Customer chat contact → NLP/NLU language interpretation → Concierge bot resolution → Action Center escalation handling → Next Best Action cross-sell → KYC process automation.

Related customer support cases