Sign In

Mohamed Abedelmalik

Founder & CEO, Brevy (YC W21)

Aspiring #healthcare entrepreneur | Alum @columbia

Highlights

Mar 25 · linkedin

In the words of Sam Altman "GPT4 sucks". Each iteration of this technology is going to make the one before it look like a toy.

I had a great time chatting with Deon Nicholas from Forethought about the history of customer support AI and the step changes of improvement over time.

Deon has been working on customer support AI for almost a decade! Forethought won TechCrunch Disrupt back in 2018. Here's how things have evolved since.

Originally customer support AI centered around decision tree automations. These rigid systems really gave support AI a bad name. Press 1 for Billing. Press 2 for Account Management. All you wanted to do was figure out how to talk to an agent.

Before LLMs, the main innovations in the space were around understanding support tickets and pulling the right policies and information to answer a given ticket. Generating human-like text for a response was a difficult and unsolved problem. AI could hear but not speak.

Text generation was the main unlock that OpenAI provided with GPT3. With that in place, AI could now understand the ticket and also provide a response.

OpenAI brought both the technology and consumer interest for the technology. They've really done three main things.

1️⃣ Built Amazing Technology: The ability for GPT3 to generate human like responses was absolutely critical to providing great support. 2️⃣ Exposed the Chat Interface: ChatGPT allowed anyone to experience this amazing technology. All at once, the entire world was experiencing the latest AI advancements. There was now a huge demand for AI. 3️⃣ Made Developing Simple: The OpenAI APIs are so simple that a developer can access the most advanced AI with several lines of code.

More great stuff to come later this week from my chat with Deon.

Mar 1 · linkedin

Commerce brands are throwing away some of the most valuable data they own.

That data is buried inside customer support tickets.

Every support ticket is a customer reaching out to tell you more about themselves. Currently, brands simply send a generic response. Sometimes it helps, sometimes not. They move on.

But in these tickets, customers are telling you so much about themselves.

Who are they purchasing for? Why are they churning? What do they love about your product? Why did your product miss their expectations?

This is all data that can be used to retain, win-back, and upsell customers.

But I don’t blame support teams. Until recent advancements in AI, they were just trying to tread water and stay on top of the flood of incoming tickets. Agents didn’t have time to tag and annotate tickets, nor were they measured on that.

Customer support was seen as a cost center. Reduce cost per ticket and improve resolution times. The end.

But in a world of customer support automation, this data about the customers is the exhaust that comes out of an AI model that’s answering the customer. It just needs to be properly captured and used.

This data is valuable. It can be used to retain, win-back, and upsell customers. In a world where CAC is on the rise, retention matters more than ever.

  1. Retain By knowing that someone isn’t happy with their experience, you have a chance to make things right. This is your chance to surprise and delight - to turn things around before the customer churns.

Maybe you send them free product, an AI personalized direct mail, a discount, a checkup call, etc

  1. Win-Back If you know that someone is returning your product or churning from a subscription, you can capture information that can help you determine when and how to contact them again for a win-back.

For someone that’s cancelling because of price, offer them a discount. For someone who’s temporarily pausing because they have too much product, reach out to them in several months. For someone who wasn’t happy with a flavor, let them know about your new flavors.

Customers aren’t all the same, so your win-back shouldn’t be one size fits all.

  1. Upsell By learning about the customer’s needs, you have the opportunity to upsell them.

Maybe they’re looking for a discount so you can offer them a subscription plan. Maybe they’re loving the results of your supplement and you can offer another that you expect to have good results.

Your individual techniques may vary, but what holds true is that personalization is the new name of the game.

To provide a personalized experience don’t ignore your most obvious source of data. Customer support tickets.

If you're running a eCommerce brand and would like to chat about this, my DMs are always open.

Related Questions

What is Mohamed Abedelmalik's current role and company?
Where did he study computer science?
Which accelerator program is mentioned in his profile?
What previous technical roles and internships has he held at Microsoft and other organizations?
Mohamed Abedelmalik
Add to my network

Experience

CEO at Brevy
Head of Seattle Chapter at Columbia Venture Community

Education

High School, High School, The Indiana Academy for Science, Mathematics, and Humanities
Bachelor’s Degree, Computer Science, Columbia University in the City of New York
S20, Y Combinator

Location

United States