There is a very strong case building for the adoption of Conversational AI Agents within every contact centre. This paper outlines the key arguments for greater adoption, as well as highlighting challenges being experienced and ways to overcome them.
The shift to digital self-service
The customer experience (CX) landscape has undergone a significant transformation with the shift towards digital self-service. This evolution is driven by the increasing demand for convenience, speed, and personalized interactions.
Digital channels have become the primary touchpoints for customer interactions, offering 24/7 accessibility and instant responses. This same expectation of immediacy is being transferred to the voice channels, where customers no longer tolerate waiting for the next agent to become available.
Powering digital CX with human capacity
Traditional contact centres are designed to service customer demand using human agents. Recruitment, selection, onboarding, scheduling, call channeling and quality assurance processes and systems aim to maximise efficiencies while reducing cost. Most work well within a consistent, predictable environment but struggle to adapt to surging volumes, diverse customer needs and rapid changes in rules and context.
Maintaining consistent sales and service levels across channels is an ongoing workforce management challenge. So much ‘fat’ needs to be factored in just to ensure enough capacity is available, should demand suddenly spike. This increases the costs – something all contact centres are under pressure to reduce.
The value of Conversational AI Agents automating voice calls
Having Conversational AI Agents automate more voice calls offers a number of business benefits.
- Reducing the average cost per call: The total cost per minute of an automated conversation is typically less than 50% the cost of calls handled by a human agents. This is largely due to the fact that you can train a single AI Agent to automate hundreds of concurrent calls in a consistent, compliant way without adding to infrastructure overheads (no desk, headset, terminal etc).
- Reducing call waiting times: Conversational AI Agents can scale instantly to meet surging call demand, removing waiting times and offering immediate access to customers looking for support (we have seen over 60% reduction in call waiting times)
- Improving First Call Resolution: Conversational AI Agents, if trained effectively, don’t need to transfer calls to other agents to get them resolved. They can resolve them first time without a human in the loop (we have seen between 2%-7% improvements in FCR).
- Improving call consistency and compliance: Conversational AI Agents can be designed to stay on specific guardrails, ensuring consistency and compliance across channels. In some of our clients, they have seen between 1% and 5% improvements to already high call quality scores.
- Improving call insights. By tracking every request made, question asked, answer given and system triggered, Conversational AI Agents can offer rich call insights that can allow for more proactive servicing.
- Reducing employee training: By reducing call volumes to human agents, and offering them real time call support, Conversational AI Agents can enable human agents to perform better with less training dependency. In one of our clients, they saved over 40% on training simply by offering their human agents access to a Virtual SME.
- Improving customer satisfaction: Having calls answered immediately, and getting requests and queries resolve first touch improves the overall experience (we have seen between 5%-10% improvements in CSAT scores).
- Improve employee satisfaction: With less pressure and stress, human agents have the space to thrive (we have seen between 3%-8% improvements in employee satisfaction scores).
The capability required to deliver this value
Not all Conversational AI Agents are equal. Some are designed to have effective unstructured conversations that answer Frequently Asked Questions, yet struggle with the more rule-bound structured conversations that are required to get things done e.g. resolve a billing query or claim request.
For a Conversational AI Agent to really make an impact on your key performance measures, it needs to be capable of the following:
- Understanding what it is you are asking. Whatever the language spoken, the Conversational AI Agent must be capable of understanding what it is you are asking for. Its not good enough to ‘kind of’ understand – it needs to be accurate to avoid frustration.
- Ask relevant questions to gain context. Before simply responding, a Conversational AI Agent must be capable of asking further clarification questions to ensure the full context is understood before an answer is given or a process is applied.
- Apply the right process logic in context. For some queries or requests, a certain process must be applied to get resolution. This may start with an authentication process, followed by the relevant sub-process linked to the request e.g. a claims process. Being able to jump across processes and not get tied up in knots is a key skill that a Conversational AI Agent must perfect if it is to work in a contact centre.
- Adjust to any change in direction or sentiment. Customers are unpredictable and don’t follow the script. The Conversational AI Agent must be capable of handling unstructured and structured conversations at once, while never going off the guardrails.
- Provide a transparent record of every conversation. To ensure you can trust your Conversational AI Agent in production, it must provide rich detailed reporting of every conversation, plus the assurance that it won’t go off script – ever.
Why many contact centres still don’t have Voice Agents in production
Many Conversational AI Agents really struggle to follow rules in context while adjusting to a customer who changes their mind or suddenly shifts sentiment. This is most often because their conversational guardrails are built using decision tree or dialog flow scripts that may work in a testing environment but get all tangled up in production, when operating at scale.
As a result, many contact centres using these generic tools find that context really matters and customers hate having rigid menu-driven conversations. Yet as soon as they try to give the Conversational AI Agent more leeway to adjust to the customer, they lose control over how it applies their rules and processes.
As a result, most contact centres only use Conversational AI Agents for FAQs or to automate basic service requests. The rest still gets channeled to human agents.
Ways to overcome these adoption challenges
To overcome these challenges, the following should be considered:
- Choose the right technology. You need Conversational AI Agent technology that allows you to build multi-dimensional, data-driven logic guardrails. Avoid any authoring tool with decision tree or process flow logic – it won’t hold in a dynamic voice conversation.
- Start small. Focus on automating the high volume, simpler conversations before increasing the complexity. Get some runs on the board before expanding.
- Get your integration sorted. A real challenge is if your systems lack APIs and your Conversational AI Agent can’t access them. This forces humans into the loop, and undermines the true potential of unassisted conversation automation.
- Learn off the data. Use analytics to track performance and continuously optimize your Conversational AI Agent’s training for better results.
In Summary
Conversational AI Agents are getting better every day. Their ability to converse in multiple languages via both text and voice means you can offer improved self-service across all your channels.
By deflecting high volumes of routine, rule-bound engagements away from your human agents, you can ensure that the calls handled by staff are of high value and impact.
The question is not longer should you employ Conversational AI Agents. It is ‘why haven’t you?’