May 16, 2023
I’ve been playing around with ChatGPT for some time now. I've been experimenting with things like summarising the emails I receive, finding ways to improve my own email writing, and trialling cool things like building out my personal running and strength training programs.
*And no, it doesn’t actually run for you, you still need to do that part!
Outside of this, honestly, I’ve been too 'head down' to think of use cases that will accelerate my day-to-day tasks and my team’s outputs. That was until last week, when the penny dropped.
I stumbled across something, and which now feels like the obvious thing to be doing already. Ramp uses Gong.io to record all internal and customer calls (unless they ask us not to of course) for sales training and onboarding purposes. Gong has some pretty cool built-in AI, but if you're reading this Eilon (Gong's Co-Founder and CPO) you need to integrate what I'm about to discuss next.
I had just finished a call with a prospect, a call which consisted of some really engaged conversation topics that I wanted to unpack with my team. So, I dumped a section of the Gong call transcript into ChatGPT-4. It was a small section (totally random one too) since the prompt box is currently limited to 15k characters. I then asked it to summarise 5 key content ideas based on the learnings in the call, and I got the below magic:
The above is incredible for two reasons:
Now that I had a taste of this power, I wondered what I could do for sales. So, piece-by -piece, I dumped in every call I had with current prospects (those who we are pitching to and preparing a business case for).
I asked ChatGPT two key questions:
It's important to get your prompts right by the way. First I told ChatGPT which individuals belonged to the prospect, and then specified that the request was only for that particular prospect, and not that of the Ramp team. I also asked ChatGPT to write it in a way that I could use the answer and present it back to the client.
The results of the challenges uncovered were astonishing to me:
This outcome was truly mind-blowing — I can’t share this unfortunately, as it has some of our IP in it. 🤐
Off the back of my Slack celebration after this discovery, Greig (our CTO) quickly jumped to the challenge and gave himself a weekend to hack. By Monday, we had a prototype of all our data from Gong, Slack and Notion all piped into a web-application and where we can now query any conversation.
There are some challenges with this though, mainly being that the current setup doesn’t allow for ChatGPT-4 so the responses (using previous versions) are way less impactful. There are several companies with integrations in place that we’re investigating to solve for this though. Once Greig is done, he’ll write up the technical implementation he went through to get us here.
What gets me excited is the benefit across multiple teams at Ramp, including:
By integrating ChatGPT usage across our various departments, we can now tap into a wealth of insights, improve our processes, and ultimately, enhance the overall performance of our business.
Additionally, I’m currently testing https://bard.google.com/ as seems like a neat tool for using GPT with live web data. A good use case here is “List 10 companies in X industry in Y country and their CEO’s and revenue”
See below sample response — pretty awesome for your SDRs!