Editor's Note: This article by Mike Olsen previously appeared in Bridgeforce insights article and is re-published here with permission.

AI-powered chatbots are poised to bring efficiency to banks and streamlined service to customers. Here’s what you should keep in mind when it comes to a new technology like AI chatbots, and five tips to help you monitor and maximize their effectiveness.

Every year, we find that one of the leading priorities for senior executives of the country’s top credit and consumer lending institutions is to enhance technology for an improved customer experience. Technology advancements have propelled the industry into the digital age, offering customers a very competitive selection of banking services. This year, all eyes will be on artificial intelligence (AI) and its potential to disrupt traditional customer service channels in banking. One practical application of AI gaining ground in customer service is the use of AI chatbots.

CFPB Indicates Changes to Chatbot Compliance: Part of New Tech “Growing Pains”

The competitive edge that AI could bring looks promising. However, keep in mind that as chatbots have developed from a rudimentary tool handling limited requests to an advanced conversational assistant, we can expect increased regulatory scrutiny to follow. In June of 2023, the CFPB published an “Issue Spotlight” indicating concerns on how chatbots could be impeding customers from resolving their problems.

Specifically, customers felt like they were in an endless loop without having their problem resolved. We can expect continued pressure on consumer protection to ensure chatbots provide accurate information and are equipped to escalate issues to human agents when necessary.

It’s not out of the ordinary for new technology to have growing pains. Think back to IVRs – an older technology designed help customers get answers without human interaction. There were similar adjustments and lessons learned that organizations endured before realizing the true intended benefits.

5 Helpful Tips to Monitor Your AI Chatbot’s Effectiveness

So, if you have chatbots or want to implement one, what can you do to monitor their effectiveness? How can you ensure they can ride out any wave of regulatory or consumer behavior change? Here are five tips that Bridgeforce recommends for helping you monitor the performance and success of chatbot usage and prepare you for future regulatory oversight.

Tip 1: Use Your Complaint Tracking System to Identify Customer Concerns From Chatbot Interactions

Within your complaint tracking system, start by creating categories specific to the chatbot or other similar AI customer interaction tools.

Next, determine if regulatory risks are surfacing from these complaints about chatbot interactions. For example, AI systems could potentially make errors or provide inaccurate results causing customers to file complaints about the quality of service or outcomes. Regulatory bodies could begin to enforce standards for the accuracy and reliability of AI systems.

Finally, implement a feedback mechanism outside of complaints to understand what your customer base likes and does not like about their chatbot experiences.

Here are a few examples of reliable feedback mechanisms:

  • Star rating system
  • In-chat or post chat surveys
  • Follow-up emails

Tip 2: Create a Formal Reporting System for Your Chatbot

Like anything else, tracking performance to assess the success of initiatives like chatbot technology is essential. Key Performance Indicators (KPIs) should include but are not limited to:

User engagement metrics:

  1. Average response time – measures the time taken by an artificial intelligence system to generate a response or output to a given input or query.
  2. Average session duration – measures the duration of user interactions with a chatbot from the initiation of a conversation to its conclusion during a single session
  3. Average number of interactions per session – would include the count of interactions or exchanges (user query and AI system responses) between a user and an artificial intelligence system during a single session.

Conversion rate: The number of chatbot conversations that end unresolved and get routed to a human agent.

After you calculate KPIs, compare your chatbot’s performance against industry benchmarks. Search for industry reports and studies published by research firms, consulting agencies, or industry associations. These often contain valuable insights into trends, benchmarks, and performance metrics for chatbots. You can also explore reports from reputable vendors in the chatbot space to gain insights into common benchmarks.

Tip 3: Review Chatbot Dialogue Examples

Create a monthly routine to review a specific sample size of all chatbot logs with your compliance and legal partners. As part of this routine, first identify what type of inquiries are falling short in their response completion. Refer to your complaints to help dig into inquiry types causing the lion’s share of customer concerns.

Then check for any risk to compliance with regulations, such as potential UDAAP violations.

Go back to the text logs and confirm accurate information is being displayed by the chatbot. Ask yourself, “would the customer get the same response if they spoke to a human?”

Finally, modify chatbot responses based on learnings to increase completion rate and accuracy rate.

Tip 4: Integrate Your Chatbot with Analytics Tools

Oversight here is key and you should integrate your chatbot with analytics tools to gather data on user interactions. This can provide you with insights into user behavior, preferences, and patterns. All of which you can use to tweak the chatbot and improve the consumer experience.

Regularly take steps to enhance the chatbot’s knowledge base over time and as more data comes in.

Once you’re receiving actionable data on chatbot performance from your analytics tools, try A/B testing. Experiment with variations of your chatbot’s responses or flows. This allows you to identify which approaches are more effective to help increase your conversion rate.

Tip 5: Invest in Your Human Staff

While chatbots are paving the way to automating simple requests, it is imperative that you simultaneously upskill your human staff to efficiently handle more complex inquiries. Be sure to track results resolution proficiency and complaint reduction.