BANKING
Client
Riyad Bank is one of the largest banking groups in KSA (Kingdom of Saudi Arabia), with 5,600 staff.
Project partner was iST Networks (Egypt).Program/Project
Replace part of existing DTMF menu structure with Natural Language Call Steering, with an eye to eventual speech-enablement of all functions.
Supplement existing PIN-based security with passphrase-based Voice Biometric solution.
Support Arabic and English speakers.My Role
User Interface Design Lead for project Kick-off in Riyad, KSA, then Requirements phase. My primary focus was on Natural Language Call Steering (NLCS): I was accompanied by a Voice Biometric (VB) specialist from Nuance Canada.
User Interface Consultant/Mentor to two Lead UI Designers, handling design of NLCS and VB modules respectively.My Responsibilities
Supported UI Design Leads with preparation/presentation of NLCS/VB designs.
Mentored NLCS UI Design Lead with semantic tagging of c. 60,000 pilot and post-deployment caller utterances in Arabic. I tagged English utterances.
Acted as this designer’s wingman during multi-day Tagging Review workshop in Riyad, KSA with client staff, including Product Manager, IVR/Operations Managers, and Business Analysts.
Coordinated alignment of VB UI design with Canadian team handling VB technical aspects.Challenges & Opportunities
Saudi culture is quite different to say the UK or US. In banking, the environment is very formal. Respect is exceptionally important. Very direct language can offend, so indirect suggestions are preferred.
The sequencing of NLCS and VB elements - which the caller hears first - was an important discussion, but complicated by legacy design, Marketing and partner integration factors.
The great majority of customers speak Arabic, so NLCS semantic tagging was challenging. I worked very closely with a Jordanian UI Designer, including an intensive 1-week joint working session in Amman, Jordan.Because we needed to support both Arabic and English speakers for NLCS, we had to find an approach that would allow consistency between languages. This required separate NL grammars, but we established a common set of semantic concepts.
VB is typically considered a security feature. For the client, this was a medium-term goal, but a higher priority for initial deployment was to reinforce the Bank's position as a market leader, without upsetting a conservative customer base.
For post-deployment tagging, I worked with a team of 4 Jordanian native-speakers, who had no prior tagging experience. This required me to adapt the standard tagging process significantly.
During analysis of log data, I identified a frequent-caller class of user that had not previously been apparent to the client. Having confirmed these were not simply auto-dialers (robots), we proposed some further research to the client, to help inform next actions.
Repeat Caller Analysis identified an unusually
high proportion of frequent callers