Luma Automation tool promises to take ‘AI powered virtual agent technology to a whole new level of automated response’. What is the single special attribute that is responsible for this?
Luma Automation, combined with Luma Virtual Agent, includes several individually unique points, but the attribute that creates the most value for customers is the ability to combine them architecturally, functionally and contextually in a pipeline approach. With Luma, you can design a complete end-to-end interaction consisting of inbound and/or outbound conversations to invoke a series of skills that can trigger actions and collect information from other systems. This is inclusive of branching based on 1. The conversation 2. External environment 3. Events.
How do you differentiate this product? Why should anyone prefer it to others in your competition?
In addition to what I said above which is a fundamental differentiation, Luma is a low/no-code GUI driven product. Primarily built for ITSM and ESM, Luma Virtual Agent is also designed as an open architecture (the solution can connect to Serviceaide’s service management application as well as other leading applications like ServiceNow, Cherwell, CA, BMC, etc.). Other products require large service projects and support only their own service management applications.
Do you foresee any challenges in having a completely automated service response product? Have you considered how the lack of EQ or human connect will impact it?
Not in the near team as Luma primarily addresses “level-1” transactional services. Domain skills are more important. The parties looking for service know what they want and know they are talking to a machine. Their expectation is to get the job done as required.
While AI is the technology that will drive almost all processes in the future, do you see any challenges in the NLP or the ML aspect, in terms of racial biases, systems drawbacks or even issues in deep learning algorithms? If there are it will directly impact response processes. What will your solution be?
B2B service and support is an implementation where accommodating flexible domain skills is the pre-eminent requirement. There is little opportunity or requirement to handle evocative topics. We believe it will be several years before non-transactional conversations are commonplace in business-oriented Virtual Agents. It is too easy to go down a path that is controversial, and most businesses stay away from social controversy, politics, etc. as much as they can. Furthermore, although the system learns from user responses, which could be biased, the administrators control what is used for supervised and unsupervised learning. In a transactional mode, the primary opportunity for biases and other “bad behavior” to creep in is when users respond to the system not working properly or not supporting their issue – and these are known response types that are typically not used in an unsupervised mode. In the end, it’s our clients’ decision about how they want Luma to be trained. Supervised training could still pick up biases but a certain level of supervision should clean it up. Ultimately, it’s more about the process within the company and not the virtual agent.
A good automated response should have fantastic data collating abilities. Does this solution have the power of data for strategic insights into customer behavior?
Yes – although much of the data within an enterprise is still mastered in other systems, be it CRM, ITSM, billing, field force management, and often separate offline big data-based customer analytics – so there are still back-office operations, customer contacts and analytics that are not visible to Virtual Agents without integrations into back-office systems. These systems often “own” the customer relationship and is the system of record for that data. They drive the end to end customer experience, campaigns, and who gets what level of treatment etc. The virtual agent can add interaction Meta data to the current “system of record” but we believe that is still in the stage of “becoming” and a few years away.
“In the end, it’s our clients’ decision about how they want Luma to be trained. Supervised training could still pick up biases but a certain level of supervision should clean it up. Ultimately, it’s more about the process within the company and not the virtual agent.”
Wai Wong, CEO at Serviceaide
Wai is Founder, President and Chief Executive Officer of Serviceaide – a global provider of enterprise service management solutions. A serial entrepreneur, he has been in the high-tech industry for 36 years. Wai held executive leadership roles at BEA where he was the EVP and Global Chief Product Officer and at CA Technologies where he was the SVP and GM of the Unicenter brand, SVP, and GM of WW Services, and SVP and GM of interBiz. Wai received his Master’s and Bachelor’s degrees in computer science focusing on AI and Systems Architecture from Columbia University in NY.