With the use of generative AI technology in customer service enterprise products still in its relative infancy, B2B vendors and customers witnessed progress with the technology this year as well as new problems created by challenges that must still be overcome.
Like with any technology, GenAI in the contact center has benefits and drawbacks. Generative AI is making processes swifter and products easier to use. On the other hand, generative AI-produced responses can be unclear or inaccurate, and commercial-grade large language models (LLMs) might be too insecure for some enterprises.
Tracking agent and customer behavior
Amid the largescale entry of generative AI systems into many vertical industries in 2023, home security company Brinks Home started using Cresta's generative AI-supported contact center technology in March. The vendor's tools summarize calls, offer agents suggestions in real time, and present statistics on customer sentiment and queries. Cresta also integrates with Brinks Home's proprietary CRM system.
Philip Kolterman, the company's vice president of digital transformation, worked with Cresta before at another company. He said he thought the vendor's generative AI software could address some of the problems that Brinks Home contact center workers were facing.
For example, Cresta supports Brinks Home agents with hints and process flows in real-time based on the call conversations. Cresta also enhances the coaching and QA experience. For example, Cresta can quantify and visualize statistics relating to frequently mentioned queries or identify when callers are having a negative experience.
This data helps Brinks Home focus on how to better support customer service teams, according to Kolterman.
"You can be much more targeted about where you QA, and then that leads to better coaching for the agent," he said.
Cresta also provides real-time coaching interventions that notify an agent's coach when a certain issue arises or if an agent has been on a call for a long time. These automations notify managers when agents need assistance, which lets them provide support faster.
In addition, Brinks Home uses Cresta to transcribe and analyze entire conversations between agents and customers. Before, the company could only analyze the agent's post-call notes, Kolterman said.
"Our ability to identify problems, issues and concerns in the process is greatly enhanced because we're doing that based on a much broader set of data," Kolterman said.
GenAI needs to be fine-tuned
But while generative AI in the contact center supports some agents, it hinders others, according to Christophe Martel, CEO of FOUNT, an employee experience software vendor.
FOUNT is a SaaS platform that gathers data from employees to understand what creates friction in their day-to-day work processes. FOUNT's clients include large financial, insurance and healthcare organizations, all of which rely on call centers.
Analyzing clients' use of generative AI in the contact center, FOUNT found that generative AI is quite helpful in the hands of more seasoned agents, Martel said.
But newly hired agents often have difficulty interpreting generative AI's suggestions for customers' less-common, more-complicated problems or situations that require more intricate processes to solve.
Philip KoltermanVice president of digital transformation, Brinks Home
"If that agent has to rely on the GenAI response to serve the customer, the agent is actually unable to know whether the answer is right or wrong," Martel said.
These agents need more specific, contextualized answers, rather than multiple suggestions on how to solve a problem, according to Martel. Many new hires would prefer more human support in these circumstances, he added.
If generative AI tools are in place and the new hires lack the skills to adequately understand the systems' responses, the agents can become discouraged and more likely to leave the job, Martel said.
"Whatever tool is implemented, whether it's a software tool, or generative AI or any other, you actually have to check that employees are able to make the best use of it in the stage of their work journey that they're at," Martel said.
But if the generative AI tools aren't providing adequate context and giving the best answer, then they're not doing their job, according to Futurum Group research analyst Keith Kirkpatrick.
"If it's done correctly, it should also have the context around it, like, 'Recommend Product A, and here's why,'" he said. "Because hopefully it's reaching back into that corpus of internal knowledge to say, 'This is why this is appropriate, and that's why you're recommending it.'"
The verticalized, end-to-end strategy
Meanwhile, businesses' customer service teams are best served by language models that have been specifically trained for their industries, according to Kirkpatrick. That's because different industries have different products, services and commonly used phrases, which factor into the language model's intelligence and usefulness, he said.
ServiceNow, which provides a suite of IT service management tools, offers models that are trained for different verticals and, therefore, are more attuned to each one, according to Michael Ramsey, ServiceNow vice president of product management for customer workflows. These verticals include banking, insurance, media and telecommunications.
Another of ServiceNow's missions is to dissect service workflows from end to end to better understand what departments within a business must be consulted or accessed to give customers the appropriate answers or solve their problems.
For example, it's possible that an interaction or transaction with a caller would not only require input from the front office but also the back office. Generative AI can be useful in improving these workflows and processes.
For some calls, generative AI can quickly provide the information that a caller wants through automated processes. At other times, generative AI can assist agents by providing the information to the agent or providing the necessary action steps to help the caller, whether they are self-service or human-to-human encounters, Ramsey said.
"In some cases, it's automated steps. But in many cases, there are people doing stuff and performing tasks," Ramsey said.
This year marked some major product advances for ServiceNow in the generative AI realm.
In May, the vendor added generative AI capabilities to its Now platform along with connections to Microsoft Azure's OpenAI Service. In September, ServiceNow expanded its Now Platform with the Now Assist family of generative AI assistants. The new capabilities include Now Assist in Virtual Agent, for businesses to create their own generative AI chat experiences, and Flow generation, for quicker workflow development.
A GenAI chatbot that is easier to configure
Another enterprise that started using generative AI this year for chat experiences and to create flows is State Collection Service, a debt collection agency.
The company needed a new interactive virtual assistant (IVA) tool for inbound calls in its contact center. In February, it decided on U-Self Serve, a generative AI-supported virtual assistant offered by conversational automation technology vendor Uniphore on its X Platform.
U-Self Serve provides generative AI-supported capabilities for interacting with end users over text and voice in a more natural and approachable way.
"We want to make sure that every enterprise can impact the customer in a human way," said Tushar Shah, Uniphore chief product officer.
Contact center teams at State appreciate how U-Self Serve enables them to make quick alterations to IVA scripts or flows without having to first consult the vendor, according to business intelligence manager Shannon Strahan.
"We have the power to do whatever we need to do on our own," Strahan said.
One example of a flow change would be if a customer receiving the inbound call or text swore after interacting with the IVA. Then the contact center team could program the IVA to automatically reroute the call to a human agent to mitigate the caller's dissatisfaction or in instances when a caller uses profanity, according to Strahan.
With U-Self Serve, the team at State can also map intents, write their own phrases and train bots themselves -- no coding required.
"It's a very easy tool," Strahan said.
U-Self Serve helps agents send debt notifications to State customers via text or phone call, and then customers can make their payments directly in a two-way interaction with the IVA. For text messages, State uses rich communications services (RCS) messaging, which includes either State's logo, colors and branding or a clients' logo, colors and branding, to build better trust with end users when they receive the text messages.
Many customers prefer paying over text because it takes less time to type out a credit card number and other information compared to speaking it over the phone, according to Strahan.
Apple finally says yes to RCS
In another development with a significant effect on contact center technology in 2023, Apple finally revealed in November that it would adopt the RCS messaging standard for iPhones, starting in 2024. For a long time, RCS messaging was only possible on Android devices. This is a breakthrough that lets RCS flood the market, according to Inderpal Singh Mumick, CEO of Dotgo, an RCS business messaging hub vendor.
"This is one of the most significant developments for RCS" since Google in 2015 acquired Jibe, a company that advanced RCS technology, Mumick said.
RCS is an upgrade from short message service (SMS) because it is an open standard that enables the transmission of richer content like pictures, videos and files via data connectivity like Wi-Fi and LTE or 5G rather than on cellular networks. RCS has also become a popular messaging method to compete with platforms like Facebook Messenger and Whatsapp, which rely on data plans for transmission.
With Apple adopting RCS, Mumick predicts that SMS will eventually become obsolete.
"All of the communication between smartphones now is going to be rich databased, either iMessage to iMessage, or RCS to RCS, or iMessage to RCS. So the value of SMS is going to start declining," Mumick said.
Mumick said that he also thinks RCS will pervade B2C and B2B text messaging interactions in a way that builds trust and improves CX, because its text messages can include the company's name, branding, and sender verification, unlike SMS. It also allows two-way messaging between the sender and receiver, so RCS interactions can include chatbots and journeys, Mumick said.
"[With RCS,] every business that sends you a message gets a name, so you see their name in there. You see a verification mark, a green tick mark, saying that they actually have been verified to their business," he said. "The whole ecosystem -- whether it is the carriers, whether it is the brands -- has been holding back because of Apple's reluctance to adopt RCS. This decision by Apple will unblock that."
"For us it means more growth," he added.
GenAI for lead routing in RevOps
The year also saw some generative AI startups make inroads into enterprise sales teams.
Scalestack, founded in 2020, developed a revenue operations (RevOps) workflow tool for lead routing. It uses generative AI to conduct research internally and on websites such as CrunchBase and LinkedIn Sales Navigator to find ideal customer profiles (ICP).
By using Scalestack, sales teams can spend less time researching and more time communicating with customers, Scalestack CEO Elio Narciso said.
Scalestack helps sales teams find ICPs that they might not have seen based on their traditional methods of research, according to James Underhill, senior director of sales operations and strategy at database vendor MongoDB, a Scalestack customer and partner. Most importantly, the AI layer knows which qualities are important when seeking ICPs.
"What's crucial is it's able to judge the relevancy," Underhill said.
Meanwhile, Scalestack uses MongoDB's Atlas Vector Search to enable retrieval-augmented generation operations in its AI layer, according to Underhill.
"It's an interesting partnership where the AI is powered by MongoDB to help MongoDB sell more MongoDB," he said.
GenAI for CX players big and small
Scalestack is one of number of smaller companies that are using generative AI to enter the RevOps market, according to Craig Rosenberg, chief platform officer at Scale Venture Partners, a venture capital firm.
"The idea of RevOps is still emerging," he said. "AI and the ability to build apps like that is actually creating a new set of RevOps tools."
One stalwart RevOps vendor, LeanData, also specializes in lead routing. But generative AI, in the hands of smaller players, might rattle bigger and longer established vendors.
"All these new guys are coming in with AI and they're doing it. … A Scalestack being able to do that is really intriguing," Rosenberg said.
Other vendors that offer similar functions include Relevvo, which also has a tool that conducts external ICP research, and GetRev.ai, which also uses AI in its tools for pipeline growth.
Clari, one of the leading RevOps vendors, in January closed on a $225 million Series F funding round, which put the company's total valuation at more than $2.6 billion.
Clari is one of the largest RevOps vendors. It offers tools for sales forecasting, pipeline management, deal execution, analytics and reporting, and integration.
Regarding AI, Clari's strategy is to infuse different types -- whether it's generative, descriptive or predictive -- into different personas or company departments, according to Julien Sauvage, Clari vice president of brand, content and product marketing. That's because of the array of departments within an enterprise that affect revenue operations.
"The vision is to be able to now absorb more data that even sits outside of traditional revenue data sources," Sauvage said. "What we want is to ultimately enable those revenue-critical employees and bring them value and address some of their use cases."
Looking to next year and beyond, contact center tech vendors such as Cresta are aiming to use generative AI to make contact center tools more integrated with companies' various systems and knowledge bases.
Cresta CTO and co-founder Tim Shi envisions that it will be more common for generative AI in contact centers to have access to various channels across a platform as well as across systems to give contact center workers and customers a more unified experience.
"It's hard for an AI to do well without knowing all the information from all the sources," Shi said.
In addition, generative AI should be able to integrate with an organization's private data and knowledge sources to better serve agents in real time while they are handling inquiries, he said. Finally, generative AI should be able to manage simpler problems and tasks so humans can handle the more complicated ones.
"Humans are going to focus on more complex workflows because the simple ones are going to be automated by AI," he said.
Overall, Cresta's goal is to make contact center operations more automated.
"We're essentially imagining that the future of the contact center can be run like autonomous cars. If one car makes a mistake, the other cars can learn," Shi said. "We think that AI can help humans to achieve that kind of goal where contact centers run highly autonomously and intelligently."
Mary Reines is a news writer covering customer experience and unified communications for TechTarget Editorial. Before TechTarget, Reines was arts editor at the Marblehead Reporter.