News & Events Humanizing financial services with next-gen AI chat and voicebots

The automation potential of AI-powered bots is too great for financial institutions to ignore—most don’t. But customers want more than simple call, retrieve, and regurgitate functionality. They want guidance as they choose between complex financial products, personalized tax advice, and detailed account analysis as quickly and efficiently as possible.

Just the time, then, for decision-makers to prioritize user experience (UX) during procurement and rollout of conversational AI.

Because ChatGPT, Gemini, and other generative AI tools are designed to follow the classic rules of conversational UX design, they’re trained to lead more human-like conversations. Something that used to be a “nice to have” in the fields of banking and finance? Now: essential.

The opportunity for better AI-led service is there. So, how do you tap into it?

It wasn’t until recently that large language models (LLMs) so drastically changed the game. But even with generative AI in play, it takes an intentional focus on conversational design to deliver better experiences for customers.

Incorporating generative AI adds another layer to deployment. It’s no longer just about understanding language; it’s about generating responses that are contextually and emotionally appropriate while also being accurate. No easy task.

First and foremost, human language is complex. You must account for more than just vocabulary:

Traditional challenges

• Conversationalizing complex financial jargon
• Robotic, impersonal chatbot interactions
• Lack of personalized guidance and support
• Frustrating navigation and limited self-service options

Then, there is the matter of prioritization. It’s somewhat easier—possibly less expensive—to deploy a bot capable of automating routine tasks. The upside can be significant: an MIT Technology Review survey found that 90 percent of respondents saw “measurable improvements in the speed of complaint resolution”;  80 percent saw AI improve their call volume processing.

Understandably, this tends to be the focus during procurement and implementation: the ROI for upfront automation can be relatively quick and significant. But it’s an incomplete picture at best.

The Conversational Design Institute calls them human-centric conversational experiences. They’re the result of close attention to a number of best practices during design: welcome messages and onboarding dialogue; personality content and confirmations; turn-taking and discourse markers. The list goes on.

Having built human-centric AI solutions for our own customers, we find that nearly all deployments require the following:

Leverage LLM capabilities for improved UX

The pool of available LLMs has never been stronger. Many deployments use proprietary LLMs with great success, OpenAI, Anthropic, and Google the foremost among them. Yet, as Philipp Heltewig, CEO of Bucher + Suter partner Cognigy, notes, open-source solutions like Mistral’s LLM, Meta’s Llama 2, and Microsoft’s Phi2 are all perfectly viable.

What these LLMs can do better than ever before is:

• Understand natural language and context
• Generate engaging human-like responses
• Continuously learn and improve from vast datasets
• Streamline complex financial concepts into clear language
• Offer conversational guidance for navigating platforms
• Enable proactive assistance and personalized recommendations
• Help deliver human-like empathy and support

Integrate AI with human expertise

Assuming you want to go beyond 1-1 automation, consider the ways you can integrate actual human touch into AI experiences. Use AI for initial screening and data gathering to enable seamless handoff to live agents for complex issues.

In this scenario, the bot handles the nuts and bolts up front while empowering human agents to focus on personalized solutions where needed. With Salesforce Einstein, for example, an agent working with Salesforce Service Cloud Voice could be served a Next Best Action or Knowledge Article Recommendation.

Personalize interactions with voicebot personas

You might have read that Meta has announced the launch of personality-driven chatbots. They’re part of a trendline toward chatbots with distinct personas. An attentive bank teller, for example; or a business-like financial advisor.

Chatbot personas provide another avenue for deeper humanization and personalization. With personas, you can automate the generation of persona-specific prompts. You can also give the interaction a more realistic feel, which helps to build trust and rapport through natural, relatable dialogue. Other examples include Sendbird, a comms platform, and SearchBot, a generative AI tool for search marketers. Both have recently unveiled their plans to deploy custom personas.

Embrace continuous feedback and improvement

AI interactions inherently create data points, many of which come directly from your customers. Position your AI solution to continuously gather and learn from this data. This includes tracking customer sentiment and interaction data, refining chatbot responses based on user feedback, and measuring and optimizing chatbot performance for efficiency.

To understand the power of next-gen AI chatbots—of deploying in adherence to the abovementioned best practices—it helps to see a real-world application.

As you may or may not know, Bucher + Suter is a close partner with Cognigy, a leader in the conversational AI space with 120 conversational chatbot specialists. Together, we worked on one of the largest conversational AI projects in the insurance and finance industry for a global insurer.

The results? More than 15 million automated customer dialogues and a 27% decrease in Average Handle Time (AHT). To achieve these results, Cognigy and Bucher + Suter focused on four core building blocks:

1. Centralized and highly scalable central platform: A central CAI platform for adapting and rolling out voice and chat applications on a business and country-specific basis. The insurer can thus drive and optimize developments internally with its teams at its own speed.
2. Concierge bot capable of relieving agents: An AI agent that fields customer inquiries 24/7 in over 17 countries and clarifies essential inquiries through natural conversation. The concierge bot is capable of intent-based routing to the correct team or employee.
3. Targeted increases in added value: A customized KPI system that clearly displays the key figures in real-time via Cognigy Analytics and various other data systems. As part of its commitment to continuous improvement, this global insurer can monitor success metrics like interaction volume, AHT, error messages, technical performance, and support statuses.
4. Insistence on consistently high-quality interactions: The agent sees customer data, concerns, and the appropriate passages in the contract or a proposed solution (a next best action) on their screen and can help the customer individually, for example, by offering a rental car or an extension to the insurance policy.

The common thread here is customers. To what extent is your team willing to prioritize customers in your chatbot deployment from the very beginning? The idea, of course, is to empower customers to self-serve confidently. Clearly, the means for doing so have never been more numerous.

That said, the ability to do so will be a critical differentiator, especially in a sector in which building strong customer relationships is essential. This begs the question: How humanized should your AI be?

If you’d like to discuss best AI practices or options for your contact center, use the form below to reach out.