How Carmoola's AI enhances customer experience and drives operational efficiency

Whilst I’m not a fan of buzzwords or acronyms… I am going to share some learnings today about Carmoola’s journey into the world of AI. Carmoola won’t be the only company excited about AI and the incredible opportunity it presents for driving efficiencies and an enhanced customer experience, but I am passionate about us starting to exercise this muscle early, so we can be at the forefront of those benefits. 

Adoption of AI at Carmoola 🦾

Carmoola's journey with AI in customer support began (as most of our projects do) with an exploration of which channels in particular could drive the biggest impact, both across the business and for the customer experience. 

As you may have seen from our TrustPilot reviews, the service we provide is praised daily, and this means that we’ve got a reputation to protect - introducing something new here isn’t more about enhancement than the need to fix something. It was clear to me that AI needed to deliver something better than our team is already doing - otherwise, what’s the point? 

Aside from support, we’re also trialling elements of AI in other departments such as marketing. More on this later…

AI in customer support: Our approach 🙌

Our first priority was to ensure we created a system that enabled a fast feedback loop to train, test and learn, and re-train the model ourselves. We wanted to make sure that the Support Team could handle this without being blocked by engineering requirements.

The next step was to understand the use cases where AI could “get work done” for us. It’s unrealistic to expect AI to be able to handle absolutely everything, and I wanted to set realistic goals. It went a bit like this:

Another important step was to set Carmoola’s principles of how we work with AI. The critical principles to call out for us are:

1. We will make it clear to customers that the response was generated by AI, and give customers the ability to talk to a human if they need more support
2. We will train our AI model as if it were a new Customer Support Team Member
3. We will configure our model to be sensitive to the communication channel. For example, WhatsApp responses should be short, whereas emails should have more of a structure and offer a complete response
4. We will take a continuous improvement and continuous delivery approach, evaluating responses and re-training every wee
5. We will not pretend that AI responses were written by a human on the team
6. We will not use AI for phone call conversations
   
   

AI in customer support: How it's going 🏃‍➡️

Our AI-driven customer support system has significantly transformed how we interact with customers. By automating 44% of WhatsApp tickets and managing 62% of email responses, AI is freeing up the team to focus on more complex cases, whilst delivering lightning-fast first responses for simpler customer queries.

We’re delighted to see comments about our AI Support in customer reviews. David seemed to appreciate and experience both our AI responses, and a human touch when he needed it. He said: "They [Carmoola] have a really intuitive 'AI' that provided information that the usual ‘bots’ just don’t get. But when you need a human, they’re available every day, including weekends."

Expanding AI's role in supporting customers 🌱

The next exciting milestone for our AI Support is to start enabling the bot to work with the customer to make changes to their account. Things like changing a repayment date, sharing copies of agreements, updating personal details, or retrying an application step are all possibilities that we’re starting to explore in 2024. 

We also see a huge opportunity for AI to assist in some of the less-sexy tasks, like reviewing bank statements or verifying order forms provided by the dealership, to deliver an even faster service to our customers.

Another exciting step for us is to utilise AI to monitor the team’s performance. One example of this is Call Quality Scoring. Today, we have a framework for this, but the opportunity is to automate this framework and score every single customer call utilising AI to detect intent, tone and customer satisfaction. I’m excited to see how this process will help us deliver “wow” more consistently to customers during phone calls.

Overcoming preconceptions and ethical AI ✅

It’s been really interesting to learn how customers’ preconceptions about AI and chatbots are shaping their behaviour today. It’s clear to see that so many customers have had poor experiences in the past (I’ve definitely experienced a few endless loops, or chats where they just cut you off or share irrelevant answers and articles), and we’ve noticed this can drive a sooner-than-necessary request to escalate conversations to a human.

I hope that as AI technologies continue to advance at the astonishing pace we’ve seen in the last couple of years, customers will have more positive experiences, and this will drive more openness to interact with AI-driven solutions, and in turn, increase automation opportunities across businesses in all industries.

AI-driven marketing content creation ✏️

Similar to the operational gains seen in the customer support team, AI has been a powerful tool that the Marketing team have embraced over the past 12 months.

The use cases have been really broad, ranging from creating and editing images across a wide range of marketing touch-points, generating high-quality audio, including new music tracks and speech, as well as text-based tool like ChatGPT for writing first-draft copy, generating new ideas, automating Google Sheets formulae and fixing HTML code.

In the last six months, we’ve sped up the end-to-end turnaround of new video creation by approximately 30% and increased our number of new paid ads by 51%, whilst not sacrificing quality.  In fact, our cost per direct plan (an key efficiency metric for us) has reduced by 45% over this time period, albeit not all of these gains are down to AI alone.

Carmoola Marketing Director, Rob Gill, added: “What I’ve been most amazed by is the pace of change that continues to take place and the new opportunities AI is creating on an almost daily basis.  It was only a few months ago that ‘weird hands’ were something we were all talking about with AI-generated imagery, but the quality is now almost on a par with a number of the licence-based image tools.” 

We’ve also been having some fantastic debates in the office around balancing the ethical considerations of using AI for content creation, vs the obvious scalability and cost benefits. 

For example, we wanted to make our process of sharing customer stories more scalable, but it’s challenging today to get real customers to take a great selfie with their car, and permit us to use it across our marketing channels. I’m not sure I’d want my actual face on Carmoola ads shown to millions - as much as I love Carmoola!

I think the main thing is that, as long as you are transparent with your use of AI, there need not be a trust issue.

Conclusion: A future powered buy AI? 🔮

Carmoola's strategic use of AI is more than just a technological advancement and a buzzword of the moment—it's a commitment to customer-centric innovation and operational excellence. As we continue to flex our AI muscles, I’m excited to see how much easier and faster we can make Carmoola, and discover even more ways to delight our customers!