Discover Financial Services
NLU migration and disambiguation design for messaging channels
Capabilities leveraged
Experience
Content strategy
Flow design
Transcript analysis
Language
Model tuning
Node architecture
Prompt engineering
Quality assurance
Product
User research
Roadmap planning
Performance management
Vendor procurement
Organization
Team enablement
Challenges addressed
01
Inadequate intent recognition: Suboptimal intent detection accuracy resulting in lower self-service containment and heightened operational burdens.
02
Unrefined prompt architecture: Redundant menu options that disregard users’ in-depth inputs, forcing them to repeat information and inadvertently increasing friction throughout the conversation journey.
03
Overuse of technical terminology: Over-indexed reliance on industry-specific financial nomenclature, undermining user confidence and detracting from overall sentiment.
04
Skewed and synthetic training data: Imbalanced training sets and synthetic modeling techniques yielding unpredictable NLU performance and hampering experience efficiency.
Timeline
$4.1M
In new cost savings, for a total of $25.4M.
9%
In additional messaging containment, for a total of 56%.
15.1M
Yearly messaging volume.