Implementing voice AI technology in contact centers has challenges. However, it’s not different from other IT projects. Three steps can increase the probability of success.
1. Determine priorities and use cases:
Every business is unique, and so are priorities and needs. Enterprises use voice AI to manage agent capacity, understand customer sentiments, reduce fraud, or improve NPS based on strategic goals. Determining the use cases linked to strategy and priorities brings clarity to projects and helps with internal buy-in. A Mckinsey study shows that an internal help desk provider could save 30% on costs and improve customer satisfaction by 10% by determining just ten use cases.
Below are a few use cases in different verticals:
Contact Centres Voice AI Applications for Banking & Insurance
- IVR for FAQs and customer support
- Compliance and fraud detection
- Accounts, statements and other product information
- Speech analytics and sentiment analysis
- Biometric authentication
- Proactive outbound calls for timely cross- and up-sell opportunities
Contact Centres Voice AI Applications for Retail
- Customer inquiries and complaints
- Store inquiries and requests
- Product and inventory inquiries
- Order tracking and returns
- Speech analytics, such as automated surveys
Contact Centres Voice AI Applications for Utilities
- Automate outage or fault reporting
- Work order automation
- Customer inquiries and complaints
- Statement, billing and payment information
- Agent coaching for troubleshooting
Voice AI applications improve productivity and bottom line through various applications, such as industrial voice assistants, in-store applications in retail or HIPAA-compliant dictation in healthcare.
Low-latency Voice AI agents built with Picovoice can automate repetitive call center tasks:
2. Encourage collaboration:
Contact centres carry valuable information for every business unit, yet every team has different priorities and business requirements. Legal teams can monitor the collection of personally identifiable information (PII), or product teams can track customer feedback. Yet, enterprises have contact centres as separate business units or under marketing with limited integration into other departments.
Working in silos will have a limited impact. Contact centres are not the only ones that can benefit from conversational intelligence.
First-call resolution (FCR) is a metric to track operational efficiency in call centres. However, HR can also use that information for training program effectiveness. Contact centres may build an IVR after analyzing repetitive tasks to increase agent productivity. Yet, the product team can use it to improve the onboarding process to avoid such calls, improving agent productivity and experience.
3. Select the right voice vendor:
Enterprises focus on their core competence and procure most IT hardware and software externally. Only a few enterprises can meet all their data and analytics needs internally. Thus asking the right questions to the vendors is critical. Vendors should offer a solution because it is what the enterprise needs, not what they have. Especially legacy players, which do not keep up with the recent advances, may offer outdated technology*.
Here’s a quick list of questions to ask voice vendors:
- Platform: What’s the variety of voice AI technologies the vendor offers? Is it an end-to-end platform or a single-product company?
- Accuracy: What’s the accuracy of software? Is there a self-service platform to train models? Do you (or the vendor) need to gather data for model training?
- Real-time: Does the software process conversations with zero latency? Is the delay predictable?
- Privacy: Does the data leave the device and go to a 3rd party cloud for processing? What are the security measures during the transfer, processing and afterwards?
- Analytics capabilities: Does this solution understand emotions, provide custom insights or detect keywords and phrases?
*You may find the below articles helpful for asking the right questions:
Different technical approaches in voice AI:
- End-to-End vs Hybrid Speech-to-Text
- End-to-end SLU vs Conventional SLU
- Direct Speech Indexing vs Speech-to-Text for Voice Search
Selecting the “best” technology:
