Comprehensive software reviews to make better IT decisions
Generative AI in the Enterprise Contact Center – Use Cases, ROI, and Preparation
Contact center as a service (CCaaS) enterprise providers are steamrollering ahead with embedding generative AI functionality into their platforms. To give a sense of speed here, the last two weeks of September saw:
- Zoom launching a workforce management suite, enabling contact center managers to automatically generate four-week staffing forecasts via AI models.
- Amazon is to invest $4B into Anthropic, giving Amazon a minority ownership stake. Amazon Web Services (AWS) will also be Anthropic’s primary cloud provider, including foundational model development. Amazon Connect users will have straightforward access to leverage Anthropic’s technology via Amazon Bedrock.
- IntelePeer launching SmartOffice. SmartOffice aims to empower local-level customer interactions via an automated customer experience process. This is achieved by preserving customer context and data intelligence across the entire enterprise, centralizing customer information records.
The future of generative AI within the contact center is upon us whether organizations are prepared or not. In this note, I explore a positive outlook for how generative AI can be used to enhance organizations' customer experience capabilities while generating ROI. This includes:
- Listing the major use cases for generative AI in the contact center.
- Discussing how we might calculate ROI from utilizing generative AI in the contact center.
- Considering what organizations can do to prepare for CCaaS vendors’ release of generative AI functionality.
Of course, the above implies there is a negative look at how generative AI might impact an enterprise contact center. I reserve that viewpoint for another post.
What Applications Are There for Generative AI In the Contact Center?
While the list of use cases for generative AI in the contact center will continue to broaden, below is a list of five commonly cited applications:
- Personalized Service via Chatbots. A clear use case that organizations are already starting to leverage are automated chatbots. Generative AI plays a major role enhancing the quality of response to customers via these chatbots. While general AI-powered chatbots are structured to provide exact responses to exact prompts (e.g. that the customer chooses from in the chat window), generative AI captures and responds to unstructured conversations. Anthropic’s Claude is illustrative of this technology: by being trained on an organization’s customer information record (such as analyzing past interactions and resolutions), Claude can understand specific customers’ preferences and tailor interactions and solutions.
- Enhancing the Agent Experience. Providing fast resolutions to customer queries is an important selling point of providing an omnichannel experience. However, this experience is only as good as the resources an agent has to support customers. Agent assist functionality is designed to dramatically speed up an agent’s ability to provide resolutions in this context. As a form of conversational AI, agent assist listens to a call and provides text assistance to agents in a screen widget. For example, Talkdesk’s Agent Assist application supports agents in real-time by offering “next best action recommendations” based on trigger words, sentiment analysis, drawing from that customer’s interaction and information record, and the organization’s knowledgebase articles. As an additional help, Agent Assist also transcribes and summarizes calls; this functionality includes automated post-call wrap-up notes that the agent only has to review and sign-off on.
- Workforce Management. An important part of streamlining contact center operations is schedule control. Generative AI functionality can quickly forecast and optimize contact center scheduling by cross-analyzing a range of factors. This saves contact center managers hours of time, given the amount of inputted data can include anything from anticipating call volume over a set period, distributing agents across channels based on skillsets or attributes (e.g. geography), ensuring specialized agents are not preoccupied with generalist queries, and evaluating and predicting employee performance based on metric scores. As an example, ujet.cx’s WFM suite (in partnership with Google) offers the above while also empowering agents to select shifts based on availability and customer demand.
- Generating Useful Data to Optimize Processes. A premise of generative AI is that it provides useful outputs based on its analysis of very large amounts of data. We’ve seen above that the kind of trends, reports, and data visualization generative AI could provide in the contact center covers workforce management and call analytics. Other important applications include analyzing which complaints require more attention, auto-generating FAQs and knowledgebase articles, suggesting insights into how the contact center’s performance ties into business goals, and pinpointing where quality management is needed most. The latter, for instance, is offered through Five9’s integration with OpenAI; by clustering recorded customer conversations by different traits, Five9's solution can identify areas for process improvement.
- Training and Simulation. We’ve seen how agent assist can help onboard agents faster, offering them real-time recommendations in trickier situations. However, generative AI can also review employee information (such as position and performance) to identify gaps and training opportunities. From there, generative AI can produce a tailored catalogue of training materials for each agent. Moreover, for agents training to use text-based channels, generative AI offers an excellent way to roleplay difficult customer questions and provide improvement feedback. NICE’s Enlighten AI is a good example here, with NICE further demonstrating how this solution could extend across an organization to enhance sales effectiveness, too.
How Can an Organization Calculate ROI for Generative AI In the Contact Center?
Each use case described in the previous section has the potential to improve contact center operations across a range of processes. However, it is easy to get distracted solutioning with generative AI and lose sight of how all of this ties back to business goals. Contact centers have historically been considered a necessary cost to handle customer complaints; generative AI functionality in the contact center must be positioned as a value-driver for the organization as a whole.
Let’s take each above use case and explore how ROI could be derived. For purposes of the below calculations, I assume a year is 250 working days. I also emphasize the calculations are illustrative only; alter the calculations to meet your organization’s functional operations accordingly.
- Personalized Service via Chatbots. By automating customer service with meaningful chatbot resolutions, operating processes can be made more efficient. There are at least two ways we might think of calculating ROI here. First, costs of hiring or outsourcing agents for general interactions can be reduced; here, we simply calculate savings from not needing to pay these costs against the cost of chatbot implementation and maintenance. Second, we can look to get a higher return on the salary of specialized agents handling more complex queries. Take a scenario where a senior call center agent makes a $49,000 salary. A contact center may have 50 of these agents, working eight hours a day, five days a week. On average in this contact center, these agents spend 30% of their time handling general inquiries that could be managed by a chatbot. Given their salary (or $0.42/minute), the weekly cost for these agents handling those general inquiries would be $15,120/week in total. If a chatbot can handle these inquiries at a cost of $0.10 per interaction (and assuming each interaction lasts five minutes) the weekly cost for the chatbot would be $3,600/week. As such, our ROI with the chatbot now implemented would lead to $11,520/week in savings (or nearly $415,000/year). Notably, this analysis doesn't account for the fact that such chatbots will also be online near 24/7 – the benefits will likely be greater when taking that into account.
- Enhancing the Agent Experience. If agents can provide a faster call to resolution via agent assist, they can get back into the queue quicker to help a new customer. This has large potential for ROI, even if we consider just five minutes difference of resolution time between two agents. Let’s say Agent A has access to automated wrap-up notes and Agent B doesn’t; consequently, it takes Agent A approximately ten minutes to resolve a call against Agent B’s 15 minutes. Both agents make $0.32 per minute ($40,000 average salary) and resolve 20 interactions per day. The cost for Agent A would be $3.20/inquiry against Agent B’s $4.80/inquiry. Over a year, the savings from Agent A against Agent B per inquiry (again, only for this five-minute difference) is roughly $40,000/year – enough to afford another agent. If this saving is expanded across all agents with greater time differences, then contact centers can be looking at substantial savings from improving their agent experience.
- Workforce Management. Using generative AI to streamline contact center operations is an attractive feature, especially for forecasting and optimizing scheduling processes. At a macroeconomic level, work automation via generative AI could add 0.2 to 3.3 percentage points annually to productivity growth. At a micro level within the contact center, consider the savings if generative AI could cut time spent on workforce management scheduling by 50%. Let's say each agent spends 20 minutes on workforce management tasks per day (e.g. planning/swapping shifts), at a cost of $0.32 per minute (given average $40,000 salary). If the organization has 100 agents, the daily cost would be $640/day. Cutting workforce management time by half for agents would result in savings of over $80,000/year.
- Generating Useful Data to Optimize Processes. We’ve seen already how generative AI could impact ROI across a range of processes – from agent experiences to chatbot utilization. Another aspect of process optimization discussed above was quality management. Quality management in the contact center regards evaluating customer interactions and agent performances, ensuring key performance indicators (such as first-call resolution) are met. What is often time consuming for quality analysts or supervisors is sampling enough calls or interactions to achieve baseline performance scores. If generative AI can produce accurate outputs for quality analysts for certain data, their time is better utilized. For instance, let’s say a quality analyst reviews 25 interactions per day. Each review takes 15 minutes at an average of $0.5/minute ($62,000 average salary). The cost would be $187.50/day. If generative AI reduces that time by 30% (not even five minutes per review), the cost would now be $131.25/day instead, resulting in savings of over $14,000/year.
- Training and Simulation. Training new contact center agents is time-consuming and expensive, though it varies by industry. One way to calculate this cost is: (Cost of Trainers + Trainees)Time + Cost of Training Systems. In 2022, the average cost of a new agent was $7,500, including 30 classroom training days. However, this does not account for “time to proficiency,” which is four to six months. This is compounded by the high rate of agent turnover: the average turnover rate was 35% in 2021 and 38% in 2022. Consequently, there is strong room for ROI if training processes can be sped up, alongside utilization of agent assist. If we double the average training cost to $15,000 (perhaps to account for time to proficiency) and a contact center hires 100 new agents per year, the total training costs would be $1.5 million. If training time is conservatively reduced by 10% using AI role simulation and agent assist, yearly savings would be $150,000. These savings could tie into business goals of cost containment while maintaining operating levels of staffing.
If we take all the above savings and hold that against the cost of implementing generative AI functionality, the ROI will likely be positive. This is especially true if generative AI implementation is done in accordance with other initiatives across the organization, exploiting opportunities created by other departments and spreading costs of technology use.
Of course, the above ROI calculations are in the context of explicit business goals. There are other indirect benefits that flow from providing greater customer experiences. These can include an increase in customer acquisition from strong service reviews, enhanced customer retention rates, better informed strategic decision making for customer markets based on service analytics, and lower agent turnover rates from being a destination to work.
How Should Organizations Prepare for Generative AI Functionality in the Contact Center?
I believe that most organizations will encounter generative AI functionality via the SaaS applications they already subscribe to. CCaaS software is a clear example of this occurring, as we saw in the introduction to this note. Banning generative AI for fear of negative risks is not going to be sustainable nor competitive in the long run. Organizations should look to review their business processes and strategically examine where generative AI will benefit them. Four elements will be important to best prepare contact centers to implement generative AI:
- Do Not Implement Generative AI in a Vacuum. All moves toward utilizing generative AI functionality must be done in lockstep with the organization’s defined data and AI governance documentation. If this documentation does not yet exist, IT as a whole has an opportunity to become leaders within their business for how this new technology can fast-track competitive growth. Contact center leaders should work cross-functionally to evangelize generative AI opportunities, spreading costs across departments to maximize ROI; most immediately, this could be with sales and marketing teams.
- Explicitly Document the Organization’s Goals and Outcomes. This should be done in tandem with understanding the context of the organization’s data and AI governance. All reference to realizing the gains of generative AI must be tied to the organization’s goals, easing creation of a persuasive business case. Contact center technology optimization is not typically top of an organization’s priority list; as such, the organization may not understand why investing in Agent Assist is important compared to traditional agent UI experiences that might be cheaper. Of course, this is unless a clearly outlined ROI of generative AI functionality in the contact center is presented.
- Evaluate the Art of the Possible. Engage with your incumbent vendor or conduct an RFI to educate key stakeholders on what generative AI functionality makes sense for your business processes in the contact center. This is especially recommended if you have an upcoming vendor renewal – is your incumbent provider a stable market leader for enterprise CCaaS software? Do they have a roadmap that includes generative AI functionality? The answer to both needs to be yes.
- Alleviate Risk Concerns. As part of your business case and in conversation with vendors, perform a risk assessment that at minimum addresses the major queries of your organization on generative AI. While this risk assessment should be conducted at the organizational level, there may be specific data concerns with customer information use by CCaaS providers – especially given that data will be fed into a foundational model for fine-tuning.
Generative AI has applicable contact center use cases that potentially offer strong ROI. Without a strategic focus toward leveraging this technology, organizations risk losing competitive advantage for customer service, alongside losing agents due to poor user experiences. However, there is still time to engage CCaaS vendors and evaluate what functionality best meets your business processes. Given the speed of technology change and typical vendor procurement process for enterprise contact centers, I believe a two-year timeframe is the window of opportunity for change. Beyond 2025, the negative impacts of not leveraging generative AI will be increasingly felt.