Chat Bots

The Latest

18 August 2021: While natural language processing AIs are becoming increasingly accurate in how they respond to questions, their ability to explain how they arrived at their answers has been limited. As The Doctor reveals, confronting a rogue AI in the Green Death, ‘Why?’ remains, perhaps, the hardest question for machine intelligence. IBM’s AI Horizons Network is developing a method to enable AIs to explain their reasoning with a common sense data set.1 

Why it’s Important.

Today, virtual service agents, both customer facing and internal IT held-desks, are effective and very efficient FAQs. They can identify a context from natural language and then provide answers to questions, as well as provide follow up answers based on the original context. However, they cannot provide details as to how they arrived at any given answer, which generally leads to a request for human manual intervention.

Specialists who develop conversation virtual service agents, work around these limitations by programmatically refining the answers AIs have available (i.e. curating the FAQ) to include reasons. E.g. “Your transaction has been declined because of XYZ.” 

IBMs work to allow AIs to report back on their reasons, may not only minimise the programming effort needed to develop virtual agents, but allow them to report decision-making in ways that organisations have not considered. 

While AI development will remain a niche activity for most Australian organisations, AI will increasingly find its way into enterprise SaaS products. Natural language AIs coupled with machine learning over knowledge assets held in core enterprise systems will see a rapid increase in the use of virtual agents, both for internal and external services. 

Who’s impacted

  • AI specialists
  • Service automation / customer experience teams
  • ICT strategy leads

What’s Next?

The rapid improvements in AI quality, coupled with their integration into most enterprise SaaS products, will make them ubiquitous for customer service delivery within the next 2-5 years.

Organisations need to start exploring the AI service agent capabilities already available in their SaaS products, and develop plans for how to leverage such capabilities. The goal should not be to deliver an ‘all-singing and dancing’ virtual agent experience, but rather to incrementally introduce capabilities over time, learning how clients and staff wish to interact, and continually leveraging advances in technology as they become available. 

Related IBRS Advisory

  1. Chatbots Part 1: Start creating capabilities with a super-low-cost experiment
  2. Preparing for the shift from digital to AI-enabled transformation
  3. BMC Adds AI to IT Operations
  4. Trends for 2021-2026: No new normal and preparing for the fourth-wave of ICT
  5. Software Agents Maturity Model
  6. Artificial intelligence Part 2: Deriving business principles

 

Footnotes

1. COMMONSENSEQA: A Question Answering Challenge Targeting Commonsense Knowledge, 2019 Association for Computational Linguistics

The Latest

2 February 2021: Google has announced general availability of Dialogflow CX, it’s virtual agent (chatbot) technology for call centres.  The service is a platform to create and deploy virtual agents for public-facing customer services. Google has embraced low-code concepts to allow for rapid development of such virtual agents with a visual builder. The platform also allows for switching between conversational ‘contexts’, which allows for greater flexibility in how the agents can converse with people that have multiple, simultaneous customer service issues.

Why it’s Important

While virtual agents are relatively easy to develop over time, two key challenges have remained: 

  1. the ability to allow non-technical, customer service specialists to be directly involved in the creation and continual evolution of the virtual agents
  2. the capability of virtual agents to correctly react to humans’ non-linier conversational patterns.

Google’s Dialogflow CX has adopted aspects of low-code development to address the first challenge. The platform offers a visual builder and the way conversations are developed (contexts) can be described as ‘program by example’. While there are third-party virtual agent platforms that further simplify the development of agent workflows (many of which build on top of Dialogflow), the Google approach is proving sufficient for non-technical specialists to get heavily involved in the development and fine-tuning of virtual agents

Who’s impacted

  • CIO
  • Development team leads
  • Business analysts

What’s Next?

If not already in place, organisations should establish a group of technical and non-technical staff to explore where and how virtual agents can be used. Do not attempt a big bang approach: keep expectations small, be experimental and iterative. Leverage low-code ‘chatbot builder’ tools to simplify the creation of virtual agent workflows, while leveraging available hyperscale cloud platforms for the back end of the agents. 

Related IBRS Advisory

  1. Chatbots Part 1: Start creating capabilities with a super-low-cost experiment
  2. Virtual Service Desk Agent Critical Success Factors
  3. SNAPSHOT: The Chatbot Mantra: Experimental, experiential and iterative
  4. New generation IT service management tools Part 1
  5. Artificial intelligence Part 3: Preparing IT organisations for artificial intelligence deployment
  6. VENDORiQ: Tribal Sage chatbot