ChatGPT has taken the world by storm. Everyone's LinkedIn feed is overflowing with tips on leveraging AI to work smarter and faster, write books for Amazon, create newsletters and marketing campaigns, do homework, and pretty much outsource all work and thinking.
So how about customer support? Can ChatGPT do a better job than automotive call centre agents?
Let's find out by asking each the same question and comparing the answers.
The question's pretty generic and simulates a car buyer in the early stages of their shopping journey who is considering an EV and likes the looks of Peugeot 2008:
"What are the considerations for choosing between Peugeot 2008-e and Peugeot 2008?"
Agent
"No agent is available at the moment".
It's Sunday afternoon, which must be the time when people are not looking for cars.
Agent attempt 2
It's Monday mid-morning, and the agents are back. This is the answer:
"Main considerations? It's very simple, do you need to drive a car with no emissions? If you do an EV will be your choice if not a petrol car might be for you as it is cheaper to buy or finance.
If you drive EV in central London it will save you on the "C" charges. If you drive on a motorway, there are plenty of fast chargers available, a quick to-up would take a few minutes.
Would you like a PCP finance quote? "
The agent's response isn't very thoughtful or helpful. The final question about the PCP quote also comes way too early and shows a lack of empathy or understanding of the buying process stage I am in.
Let's see how ChatGPT handles it.
ChatGPT
The same question (What are the primary considerations for choosing between Peugeot 2008-e and Peugeot 2008?) produced this answer:
"The Peugeot e-2008 and the Peugeot 2008 are both compact SUVs, but the e-2008 is an electric vehicle while the 2008 is powered by internal combustion engines (gasoline or diesel). Here are some main considerations for choosing between the two:
Powertrain: One of the main differences between the e-2008 and the 2008 is their powertrain. The e-2008 is powered by an electric motor and has zero emissions, while the 2008 is powered by a gasoline or diesel engine. Choosing between the two will depend on your preferences for fuel type, environmental impact, and driving experience.
Range: The e-2008 has a range of up to 193 miles (311 km) under the WLTP standard, while the range of the 2008 will depend on the fuel type and engine size. If you are looking for a vehicle with longer range and lower emissions, the e-2008 may be the better choice.
Performance: The e-2008 has quick acceleration and a smooth, quiet driving experience due to its electric powertrain. The 2008 has a variety of gasoline and diesel engine options with varying levels of power and performance. If you prioritise performance and a more traditional driving experience, the 2008 may be the better choice.
Features: The e-2008 and the 2008 have similar features and options, including touchscreen infotainment systems, advanced safety features, and driver assistance features. However, the e-2008 may have some additional features related to its electric powertrain, such as regenerative braking and charging timers.
Ultimately, the choice between the Peugeot e-2008 and the Peugeot 2008 will depend on your personal preferences and priorities for fuel type, range, performance, price, and features. It's important to research and compare the specific features and options that are important to you, and to test drive vehicles that you are considering purchasing to get a sense of how they drive and how they meet your needs."
And the winner is ...
ChatGPT.
That's because:
It responded on Sunday when the agents were not available (and it would do that consistently 24 hours a day, 365 days a year)
It came closer to answering my question, and although there were a few details wrong (2008 no longer comes with a diesel engine, and EV doesn't have a longer range than combustion engine), the response was much more detailed and "thoughtful."
The problem is that the victory is quite hollow because even this answer wouldn't help me continue my shopping journey.
Why were the answers wrong?
The agent lacked knowledge, time, or both.
ChatGPT lacked knowledge because it crowdsourced the response from the public Internet.
This is the biggest problem with ChatGPT today. It is very easy to get impressed by it because it far exceeds what was publicly available before. Yet on a deeper look, it lacks originality, creativity and thought because it only presents generic knowledge from the Internet in the safest possible manner.
ChatGPT doesn't provide insightful advice or expertise, and it doesn't represent your brand. This means it can't replace call centre agents quite yet.
How to fix it?
To make ChatGPT, or any similar AI technology truly useful, car manufacturers will need to use their data to train it.
Generally, the amount of data and time required to train AI depends on the size and complexity of data and the language, the specific tasks you want the chatbot to perform, the computing resources available for training, and, above all, the size and quality of the dataset.
In the automotive context, you would need millions of examples of quality text data to train a basic ChatGPT model. That's the first problem.
The second is that training a ChatGPT model requires significant expertise in machine learning, natural language processing, and computing infrastructure and a team of experienced data scientists and machine learning engineers to ensure that the model is adequately trained and produces the desired results.
What's the solution, then?
AI chatbots are here to stay. In many cases, they already provide better answers than call centre agents. When AI technology becomes more robust and affordable, it will likely take over many customer support functions altogether.
All that might take a while, but start getting your data ready now. Questions, answers and their outcomes. From forms, emails, surveys, live chat, phone calls, your website search box, social media, dealers, car configurator, analytics, e-learning platform, and knowledgebase. The more quality data you collect, the better your AI chatbots will become.
In the meantime, supplement your call centre agents with "smart" chatbots. These rely on predefined paths and rules and only use AI to recognise and process unstructured queries or speech. This makes them much easier to configure and launch, they don't need much data or training to be useful, and you will have their responses entirely under control.
They could do the Sunday shifts, support buyers in the early buying stages, and cheerfully respond 24 hours per day, 365 days per year.
As your learning and knowledge base grow, you could gradually reduce the number of call centre agents while reinvesting the savings to train and pay the remaining ones better. These superstars could then focus more on the latter buying stages and convert more shoppers to loyal customers.
And as you would feed all information captured from your super agents' and chatbots' engagements to your ChatGPT model, it will become increasingly smarter until it's ready to replace your preconfigured chatbots altogether. And one day - in a very distant future - also your call centres.
We can help you to accelerate the journey. Contact us.
