Our first time using OpenAI’s ChatGPT
Like everybody else, I remember being fascinated with what OpenAI had done with ChatGPT. From having it explain complex topics in a rappers tone to creating frameworks for customer conversations, it was definitely something I never imagined I would be interacting with in my lifetime.
However, I didn’t give much thought beyond that, and continued on my day. At that point we were in the midst having customer conversations with small businesses in the US and Europe. We were trying to understand problems service-based small business owners had in managing their business. I was on that path, since we had pivoted from our first product, realising that we didn’t have a large enough total addressable market (TAM), and we didn’t spend enough time understanding what the customer problem was.
Building a quick OpenAI prototype
We had finished around 45-50 conversations, when we started seeing that a lot of small business owners spent time post-work, going through voicemails from clients about appointments and cancellations. Since they couldn’t answer calls during work hours, they ended up spending 30-45 minutes after work. We decided to build a quick proof of concept to see if we could build something to save them time.
For our POC, we quickly hacked away, used APIs to see what was possible
- We used Twilio’s API for a virtual number to record voicemails
- We used Assembly AI to transcribe it and identify speakers.
- We used OpenAI to extract data points like – appointment time, appointment type, preferred service provider and categorisation
- We used Sendgrid to mail this to a client.
This was when I started getting my hands dirty with OpenAI and realized while chatGPTs generative capability was amazing, its ability to interpret and understand instructions could really be a game-changer.
For business owners, instead of having to listen through a number of voicemails, they would just need to skim a list of tagged emails, significantly saving time.
This got us wondering if we could do something in the customer support space and we decided to take a harder look e-commerce customer support.
Why we chose Shopify support
When we started using Shopify
Before we worked on our first product, my two co-founders and I had started a digital advertising agency 9 years ago. We’d grown the team to around 60 people at the time. However, we always had this desire to build a product and see if could make some sort of an impact there.
In 2020, we decided that we’d give building a product a shot. The agency at that time was doing extremely well – we’d built a great team, we were working with some amazing brands and had won awards along the way. We decided that one of the founders would continue to lead the agency, while the two of us would try and work on building a product.
During my time at the agency, I was heading creative design and development. Around then was when I started messing with Shopify. Building an e-commerce store for scale in 2014 wasn’t as easy task. But then came Shopify, and it was one of the easiest things to work with. It allowed us to focus on launching e-commerce brands, while they took care of the heavy lifting with the functional side of things.
Reasons behind choosing to build for Shopify
Fast forward to 2023, when it came to deciding if we wanted to approach customer support in general or pick a vertical, the decision was easy – Shopify stores.
- We understood Shopify deeply
- Solid businesses are built on it
- The ecosystem is large
- If we’re solving for e-commerce, what a great place to start
We proceeded to have 20-25 conversations with Shopify business owners to try and understand problems in customer support from their lens. We spoke to businesses averaging 300 tickets a month, to behemoths, handling 20,000 tickets a month. While each business had their own way of going about their things, the pattern seemed clear as day – 60-70% of their ticket volumes were redundant and had their support teams using canned responses to answer.
We were convinced that there was a problem which could be solved here. We then decided to finally start building towards this.
What building for Shopify has taught us, so far
Building in a crowded space
I’ll be the first one to admit, that we when looked at the landscape of CS tools, everyone and their dogs were building something for customer support with AI. We obviously had thoughts around – are we another me-too? is this a bubble? is this just another web3, crypto fad?
I’d like to believe that we had answers, but we didn’t. We decided not to over analyze since we believe AI is that next big wave (if it isn’t already), which can be compared to what the iPhone did. It changed behaviours, creating an entire eco-system of apps doing incredible things.
We understood that focusing on AI wasn’t important. Solving the actual problem was. AI is a means to solving that problem, albeit, a super-powered one. Hearing Microsoft CEO, Satya Nadella’s, take on AI being more of a co-pilot also helped mould our perspective on how to use AI as a solution.
“We are moving from the auto-pilot era to the co-pilot era of AI. This co-pilot pattern has made me a better editor, right we sort of put a lot of emphasis on creating, not as much on fact-finding, and maybe that’s at a premium”
Satya Nadella
Building for a specific vertical
Building for Shopify has us convinced that if you go really focus on one vertical, you discover nuances, challenges and solutions specific to it. I know it sounds basic, but when you have to make a decision about the problem to solve, we found ourselves questioning if we should go horizontal and cater to a larger audience, or focus on a particular niches problem.
I read somewhere that focusing on a niche, doesn’t mean the problem isn’t big enough, or there aren’t enough people facing this problem. It just means that there’s a certain set of people out there, for whom this problem is unique and big enough for a number of reasons.
This has impacted us in the way we’ve been building the product as well. All our workflows, guardrails and thinking is around what customer support means in Shopify’s ecosystem.
Closing thoughts
If you’ve made it this far, thank you 😀
We’ve seen the amazing work that Intercom, Zendesk and a host of other companies are doing in the customer support space. Does it scare us being a startup? Absolutely not. In fact, quite the opposite. It made us believe that AI is helping completely change customer support inside out and we all have access to the same tools!
While the larger players have access to years of building CS tools, research and capital; being small allows us a few advantages which we hope to leverage as well.
All I know is that I’m excited to be in this space and, more importantly, excited to actually solve a problem which customers have identified themselves.