At Enterprise Connect 2024, Genesys was the winner in the Best Innovation in Customer Experience category for AI Experience, described as “integrating generative AI for agent productivity.” However, at the time, there was no press release announcing new AI capabilities for the company’s flagship Genesys Cloud CX platform and there was nothing new on the website that fit the description. Enterprise Connect stated eligibility requirements are that products/services need to be announced by August 2024.
Well, it’s mid-May, and at Xperience 2024 the AI news was announced. As I suspected, the news expanded on an AI Experience bundle of services revealed in September 2022. At that time, four AI capabilities were packaged together: predictive routing, predictive engagement, digital bots, and knowledge management. AI Experience now includes all of the original capabilities with the addition of automated assistants, all powered by large language models (LLMs).
The graphic below provides an overview of how Genesys will deliver its AI Experience solutions. The Genesys Cloud platform has a set of native AI services. All AI features (e.g. Copilots and Virtual Agent) and applications that use AI (e.g., workforce management and journey management) rely on this set of common services.
Powered by LLMs, Genesys Copilots will provide real-time support to contact center employees, agents, supervisors, and administrators; each job role will have its own copilot.
The initial copilot, Agent Copilot, monitors what’s happening during service interactions and provides agents with a head start on task completion, helping to improve average handle time, interaction wrap-up time, and customer satisfaction. Agent Copilot can also surface information from knowledge management systems in real-time.
General availability for two more role-specific copilots, Supervisor Copilot and Admin Copilot, is expected later this year.
Copilots are tools for the people who work in a contact center. Virtual agents are designed to enhance self-service, i.e., automation of interactions. In a Digital/AI roadmap session at the conference, Genesys Virtual Agents were described as, “evolving bots into something more.” In the first release, Virtual Agents will provide an interaction summary and wrap up codes for interactions.
In the near-term future, Genesys will incorporate LLMs “in a safe, controlled, and secure way” into the building of Virtual Agents. This will allow the creation of natural, conversational Virtual Agents . Organizations will be able to build Virtual Agents to address specific customer intents and tasks as well as generate workflows based on top agent interactions.
In an interview with Rahul Garg, vice president, product management, AI and Self Service, Genesys, I asked what ‘generating workflows based on top interactions’ meant. Like any good software engineer, he opened his laptop and showed me!
Assuming that the customer has Genesys Quality Management (QM) – and almost 90% do – the Virtual Agent building tool would identify agent interactions that had high QM scores, e.g., in the top 10 percent. The scores could be based on supervisor rating or automated QM.
Garg explained using an airline contact center example. A number of live agent interactions with high QM scores that were tagged with the customer intent “lost luggage” are selected. A flow of the steps those agents took would be automatically created and stored in a flow library. That library could then be used to build a voice and/or chat bot to replicate the steps taken by the live agent for automated completion of such requests in the future. Or the flow could be used to build guided steps for other agents who receive calls that are identified as having a “lost luggage” intent.
The general availability of Genesys Virtual Agents is expected during the third quarter of 2024.
The importance of agent empathy has been a Genesys theme for over two years. In a blog in February 2021, Tony Bates, CEO, Genesys, defined the term. “Empathy is the human ability to listen and understand one’s situation and treat that person with respect,” he wrote. The theory is agents can drive better results when they’re more in tune with how their actions and responses influence the effectiveness of customer experiences.
Now, organizations can use Empathy Detection to measure the level of empathy displayed by agents during interactions so that plans can be developed to sharpen their emotional intelligence through customized training and coaching plans.
“I asked Garg how empathy was being measured. Has Genesys built an AI model to measure it? “It (Empathy Detection) is built on an open-source language model,” Garg explained. The model uses open source AI model-based training with examples of empathy as well as phrases that display a lack of empathy (i.e., unhelpful).”
It is not clear from the available documentation how examples of empathetic and unhelpful phrases are created, but they are likely created by an LLM prompt and then fine-tuned by each customer.
Examples offered for empathetic phrases include:
Unhelpful agent phrases include:
I then asked how identifying the phrases leads to a quantified score for each interaction. Garg pointed me to a Genesys reference document online that provides the equation that is used.
Empathy score = 100 * [(Number of empathetic phrases – Number of unhelpful phrases)/
(Total number of empathetic and unhelpful phrases)]
For example, if an interaction has 2 empathetic phrases from an agent and 3 unhelpful phrases, the score is calculated as follows: 100 * [(2-3)/(2+3)] -> 100 * [-1/5] -> 100 * (-0.2) -> -20. The scoring scale would vary from +100 (of all the empathy-related phrases, all were positive) to -100 (where all the related phrases were not helpful).
AI pricing is always a topic of interest and for its updated AI Experience portfolio, Genesys is introducing a new pricing strategy based on tokens. Organizations consume Genesys AI Experience tokens as they use Genesys AI products at the rate shown in the table below. For example, 51 digital bot sessions cost 1 token. The quantity of tokens consumed appears on the customer’s invoice each month. This table describes token allocation.
Each Genesys Cloud organization gets a free allocation of tokens every month, 250 for each named organization. Free tokens are designed to make it easy for customers to trial any of the AI products without having to go through a sales process. After that, tokens are priced at $1.00 each and the usual customer discounts apply. Customers can purchase tokens through the Genesys marketplace, AppFoundry, or through their sales executive.
Most of the AI products listed are available to all customers and license types with the exception of the Genesys Cloud CX 1 license, which is a voice-only license and can only consume voice-related AI products (predictive routing and voice bots).
Now we all know why this year’s Enterprise Connect Best of Show judges chose Genesys AI Experience as this year’s winner. The combination of some unique capabilities (e.g., Empathy Detection) and a creative, flexible pricing strategy impressed the judges.
For more on Genesys’s AI strategy, we have an in-depth conversation with Genesys SVM and VP Brett Weigl on “How Generative AI Can Assist Contact Centers.” (August 22, 2023)
