Q&A: Greg Stack on Blazing a Trail to Successful AI Migration

Article Featured Image

Q: What is the essence of your presentation at the SpeechTEK Conference, April 27-29, in Washington, D.C.?

A: The presentation will cover successful AI migration roadmaps, options, and deployments. Specifically:

  • Architectural AI migrations that leverage existing IVR business logic, flow, and backend database interactions to create a lower cost omnichannel environment,
  • Detailed migration steps and IVR customer interaction targets for hybrid AI Cloud architectures,
  • Options for standardizing AI digital omnichannel approaches across voice, chat, SMS, mobile and virtual assistant channels, 
  • Privacy strategies for protecting PCI, HIPPA, customer and corporate data

Q: What are the biggest problems faced during AI migrations?

A: The first and biggest problem is usually how and when to start! The organization has to decide on its appetite for change and comfort with project scope. What is the risk/ROI reward ratio? What are the methods for training the AI model and tuning it to reduce intent conflicts from similar responses? 

My presentation will address these questions. I will also cover how to get started and how to avoid pitfalls while still achieving ROI results. 

Q: Sometimes there is a bias in the training data that gets passed on to the ML application.  Have you found this to be a problem, and if so, how do you resolve it?

A: Training data is not perfect, and neither is AI. All models need to be tuned and updated after initial testing and on an ongoing basis. If mistakes and biases are detected, then scenarios and rules need to be created to address these issues.  This is not a one-time exercise but rather an ongoing part of maintaining the AI system.

Q: How do you manage customer expectations during deployment?

A: Many believe that AI is a magic bullet, that it will miraculously provide the right answer and correct itself when needed.  NOT! This is often management’s expectation from publication promises. Expectations need to be set at the onset of the project.   A step by step approach with measurable targets for each phase showing positive ROI results is critical to success.

Q: How do you manage the ongoing maintenance of the system?

A:  AI and ML need human checks and balances.  Think of this as AI tuning or optimization.  This should be done daily or weekly at the start of a deployment and evolve to monthly and quarterly once the system has matured.  Systems with frequent change could permanently require monthly tuning. New applications or added functions should be target tuned, which means tuning becomes an integral part of the operation.  Conversations can be tested against the AI model and then modified where needed to provide the proper answer or intent. Similarly, the AI model can be used to identify requests for which there is no resolution, and these can be positioned as opportunities for future releases. 

Q: How do you maintain a consistent user experience across channels?

A: An Omni-channel architecture is the key.  While this is easier said than done, there are several vendor suites, hosted solutions and products that facilitate this.  The presentation will review various approaches to omnichannel architecture and the top vendors and solutions that star in this arena.

To see presentations by Stack and other speech technology experts, register to attend the SpeechTEK Conference.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

Transformation in the Contact Center

Customer service has moved from call centers to a multichannel and now omnichannel customer experience approach, with cloud-based communications, well-designed IVR, and powerful reporting capabilities thanks to artificial intelligence (AI), chatbots, natural language processing (NLP), and more.

Q&A: Dr. Nava Shaked on Evaluation, Testing Methodology & Best Practices for Speech-Based Interaction Systems

Get a sneak-peak into Dr. Nava Shaked's SpeechTEK workshop in this Q&A. Learn everything you need to know about evaluation, testing, and best practices for speech-based interactions.

Video: Detecting Deception in Speech, Pt. 1: Methods

Interactions LLC NLP Scientist Yocheved Levitan discusses a recent research study on identifying cues of deception in speech, and machine learning-based approaches to classifying deceptive speech in this clip from her presentation at SpeechTEK 2019.

Q&A: Bruce Balentine on the Basics of Conversational Chatbots

Conversational chatbots may be the future (or even the present) but Bruce Balentine is taking developers back to basics in his workshop at SpeechTEK in April. Learn how to build chatbots that get results.