Are you a marketer, customer service manager, or business owner? Do you use CallRail (or any other call tracking platform), as well as ChatGPT? Then get ready for what you’re about to learn, because AI is about to change the quality of your customer service. All you need is the right prompts.
CallRail is powerful. It’s a great tool for tracking phone calls from your customers and leads and will track a variety of metrics and dimensions. It can track recorded calls right down to the Google Ads keyword. Not only that, but it can track your average call length, filter calls based on keywords found in the transcript, and so much more.
But what if it could track things like the number of times customers were put on hold? What if it could provide concise call summaries so that you don’t even have to listen to the phone call or read the transcript? What if it were so smart that it could rate the quality of your customer service…and then track that over time?
If you’re familiar with ChatGPT, then you already know where I’m going with this. And there’s no turning back. Well, I guess you could, but I don’t recommend it. This is 2023. The future is now. Ditch the past.
To get this to work, export the phone call transcript, or simply copy it directly out of the CallRail platform. If you’ve enabled PPI redaction (which you absolutely need to do), then all personal identifying information should be stripped out of the transcript. However, there may be other elements of the transcript that you may want to remove, such as the company name.
Next, paste the transcript into ChatGPT. No need to format it, just paste the raw transcript (I’m not kidding…it’s that smart). Now, copy/paste one of the prompts below directly before the transcript. Example of the full prompt you need to use: “Did the customer purchase something on this phone call? Transcript: <<Paste Call Transcript Here>>.” After the transcript has been submitted to ChatGPT, your subsequent prompts don’t need to include the transcript again.
ChatGPT CallRail Prompts:
(Copy/Paste – Each One is a Different Prompt)
1. Provide a summary of this customer or lead phone call.
2. On a scale from 1-10, how do you think the customer would’ve rated the satisfaction of this call?
3. What could customer service have done better to assist the customer here?
4. Did the customer purchase something on this phone call?
5. Did the customer complain about anything on this call? Answer with either “Yes” or “No”. After providing a yes or a no, explain how you came to that answer.
6. Was this a phone call from a current customer? Answer with either “Yes” or “No”. If you are unsure, then the answer is “No”. No other answer is considered acceptable. Just “Yes” or “No”.
7. How many times was the customer put on hold during this phone call?
8. Did the customer use any profanity during this phone call? Answer with either a “Yes” or “No”. If you are unsure, then the answer is “No”. No other answer is considered acceptable. Just “Yes” or “No”.
9. How many words did the customer say on this phone call? How many words did the agent say on this phone call? Count each word that the person says as one word. For example, if someone said “Hello hello hello,” then could that as them saying three words.
AI Answer:
I used prompt #2 for this example. By the way…ouch…a two out of 10! Not surprising, though, if you read the script (you did read it, right?).
If you’d like ChatGPT to only answer with a number, all you have to do is add this to the prompt: “Only respond with a number. I don’t need any explanations. Just a number. Nothing else is considered acceptable.” This way, if you wanted to store the number in a database field, you could. Check it out.
Cool, right?
Let’s do another example because I like you.
Spot on! I mean, seriously, this is exactly what needs to happen with Overpriced Furniture’s customer service.
Important note: how would you like to run multiple transcripts in the same chat window and have ChatGPT analyze them all at the SAME TIME? That’s possible, and I recommend it. It could, for example, look at five transcripts and provide a high level list of recommendations based on all of them.
Now it’s time to really enter the Matrix, or at least step into it for a few seconds…
Taking This CallRail Method To Another Level
You’re probably thinking, “Well, this is awesome. But can I scale it? I don’t have time to run EVERY transcript through ChatGPT manually.”
The good news is you don’t have to. You could take a random sample of 10 transcripts every week (for example), run them through ChatGPT, and easily get ideas of how customer service could be improved. It would likely take around 15 minutes.
Or, if you have a subset of customers or marketing campaigns that are having issues, you can run a report in CallRail to produce transcripts for ONLY those calls, and then just run those through ChatGPT.
But what if you have a burning desire to scale it? More good news. If you want to actually scale this thing, it IS apparently possible. But I haven’t done it yet. I will be updating this post AS SOON AS I set up an automated solution for scaling it.
In the meantime, while you’re waiting patiently, I’ll expose my plan. Feel free to tinker with it.
My plan is to use this Zapier integration (screenshot below).
There is a field in CallRail’s API called “Transcript”, which quite unsurprisingly contains phone call transcripts (imagine that!). When a call is completed, it would send one of my prompts featured on this page to ChatGPT, along with the phone call transcript right beside it.
Then, another Zap in Zapier for Google Sheets to ChatGPT would send the output from ChatGPT directly to Google Sheets.
The end goal would then be to pipe the data from Google Sheets over to Google Looker Studio and visualize it.
Just imagine…
• Being able to track your customer service rating over time.
• Trending your number of customer complaints. This would be HUGE!
• Getting highly intelligent AI to filter customer calls out of your conversion data for good without having to rely on keyword spotting.
• Seeing if your customers are getting put on hold, and trending that over time.
It opens up a whole new world of possibilities.
Update: CallRail must’ve been reading my mind (or reading my site). As I type this, I found that they will soon be adding a new feature called “Call Summaries”. It’s promised to do the exact same thing as my ChatGPT prompt for call summaries. But what are they missing? Every other prompt I mentioned. They have a long way to go to catch this, but they’ve clearly put on their running shoes. I’m excited to see what they add in the future.
Disclaimer: If you work in an agency or have clients, be sure to check with them to ensure that they are okay with running their phone call transcripts through ChatGPT. Enable CallRail’s PII Redaction feature, which strips out all unique identifiers. Also, update all privacy policies accordingly (if needed). Be safe. Be smart. We’re in the “wild west” phase of AI. Run The Prompts is NOT responsible for anything you do. Don’t be stupid. Don’t be a fool. With that said…prompt the planet!
If you like ChatGPT, you'll love Venice. Venice is private and uncensored! Try Venice today for free or get 20% off Venice Pro for a limited time with promo code "RUN20".
