How we are using AI and ChatGPT to transform our internal support

A few months ago, Klarna reported significant cost savings thanks to their AI-powered customer service assistant. This assistant, built with OpenAI’s technology, handled a whopping two-thirds of Klarna’s customer service inquiries, which translates to 2.3 million conversations across 35 languages with a high satisfaction rate.

The efficiency gains are impressive. The AI assistant works like 700 customer service reps combined, bringing the average handling time for issues down from 11 minutes to just 2 minutes (an 80% improvement!). Additionally, the AI’s accuracy on first attempts has reduced repeat inquiries by 25%. These improvements are estimated to drive $40 million in profit for Klarna by 2024.

Klarna’s success story brings to light a challenge we’re facing at Lendsqr. We have incredibly high standards for recruitment and placement, and somehow, we’ve applied those same standards to our product support team.

This can lead some support staff feeling like they’re overqualified and undervalued. Retaining staff in these positions becomes difficult, and onboarding new hires adds another layer of complexity because every time we bring in new people, they need to be trained from scratch.

So, inspired by Klarna’s example, we’re rethinking our approach to product support.

We started thinking differently. What if we could transform product support from a purely reactive role? Instead of lowering our standards or just hoping people stay for a second year, why not elevate the role itself?

When life gives you lemons, make the best lemonade money can buy

We wondered if AI could address the challenges in our product support team. Building a complex AI support system from scratch seemed expensive and time-consuming, but Lendsqr’s culture of experimentation encouraged us to try a different approach.

We decided to develop a basic version of the system, called a minimum viable product (MVP), using a custom GPT built on ChatGPT. We loaded all our existing knowledge base into the ChatGPT support system, focusing on the information that our support team used most often, and the results were impressive!

The AI system wasn’t just able to handle basic support tasks; it could also connect to some benign APIs to access even more information. This was a major breakthrough.

Featured read: Why we built our lenders’ web app and how it has changed lending forever

Our love for content and experimentation paid off

Investing in creating high-quality written content throughout Lendsqr paid off in a big way with our AI experiment, thanks in part to our founder, Adedeji Olowe, who loves writing and has been blogging for 23 years.

Our focus on content marketing, along with the fact that many of our employees work remotely, means that we have a lot of well-documented information. This stockpile of content turned out to be the perfect resource to train our AI system. Our exploration of ChatGPT support highlighted the value of our emphasis on written content at Lendsqr.

The positive impact on efficiency and team morale has been significant. We’re still in the early stages, but we’re planning to invest more heavily in AI in the coming quarters.

Featured read: Lendsqr vs Loandisk: Which loan management software is right for you?

We went on to create Career GPT to support our HR team

The success with product support coincided with a surge in job applications, which overwhelmed our HR team. Seizing this opportunity, we developed a separate AI support system, Lendsqr Careers, specifically for interacting with job candidates in the early stages of the application process.

Lendsqr Careers GPT benefited both applicants and HR by streamlining the initial communication.

We also experimented with Bland AI to make phone calls. Our initial experiments at using Bland AI for debt collection calls weren’t successful (the accents were a bit unusual, and let’s be honest, people don’t enjoy getting collection calls!).

However, we found success using AI to reach out to job candidates. Automated reminder calls significantly increased our application completion rates.

Overall, our venture into ChatGPT support has been a very positive experience. It has reduced pressure on our team and allowed us to provide better service to both our customers and potential hires. We’re excited to see what the future holds as we continue to explore the possibilities of AI at Lendsqr.

Our early exploration of AI is already transforming the way we do business

Our experience with AI hasn’t been all sunshine and roses. We’ve found that voice AI, in particular, still struggles to understand natural language and respond to prompts accurately. We’re working on it, but for now, it’s a challenge.

On the other hand, generative AI has been a real boon for us. Our initial experiments with ChatGPT support were very successful, and we’re planning to develop a comprehensive AI architecture that will integrate smoothly into our ecosystem.

To accomplish this, we would need expertise on both the user-facing side (front-end) and the data processing side (back-end), but the potential benefits are immense.

Featured read: How AI can transform lending

Our experience with AI has us excited about the future of customer support

We strongly suspect large language models like ChatGPT would completely change the way businesses interact with customers with embedded AI chats. This could put the likes of Zendesk out of business.

Imagine having AI chat features built right in that can have normal, interesting conversations, instead of just giving the same pre-written answers over and over. This would not only make customer support faster, but more reliable, and ultimately much better for everyone.

We also expect a future where customer service would be multimodal, with people easily switching between texting and talking to AI. They might even be able to do video calls. We’re not there yet, but this is what we’ve been up to and it’s already transforming the way we do business.