Intercom's AI Product Strategy

Lenny Rachitsky interview with Paul Adams recap

Welcome to the 3,039 smart, curious AI folks. Not with the group yet? Subscribe here

​

Hey crew! I heard Lenny interview Paul Adams, Head Of Product @ Intercom on an awesome podcast and I had to tell you about it.
​

I got up early, grabbed my cup of coffee, a notepad, and as the morning sun illuminated the steam rising from my cup, I listened to the sweet words of AI adoption.
​

You see, this closely aligns with a topic that is near and dear, how a company approaches building with AI. I’ve chatted with 100s of companies doing the same thing, even interviewed Mike Knoop (Head of AI @ Zapier) on this topic.
​

I pulled out all the best parts from Lenny's chat with Paul and added a bit of my own color commentary. If I had your phone number, these are the things I would text you about.

​

TLDR:

• Intercom completely changed their product strategy after ChatGPT came out. They went back to first principles, “Why do people use Intercom?”
• Their biggest challenge hasn’t been the tech, or coming up with cool ideas to build…it’s been helping their customers think about organizational change. (I know I shouldn’t be surprised at that, but I still am.)
• Paul claims that Intercom’s product response to AI, Fin, can handle 50%-70% of a customer's inbound support requests. Wild if it’s true.

Lenny <> Paul

Paul says his “Before & After” moment with AI started with ChatGPT, symbolized by Fergal’s (VP Of AI @ Intercom) tweet.

When Paul talks to people about AI, he noticed that they generally fall into 2 camps:

1. Either they know a “meteor” coming towards everyone
2. Or they are the “I’ve heard this hype before” crowd (crypto/metaverse)

​

Can you guess which camp Paul falls in?

​

However it wasn’t always like that, Paul said that if he and Lenny had chatted 18 months ago Paul would have said that ‘AI’ isn’t real. “Machine Learning is real, but AI isn’t.”
​

Paul compares the AI wave to mobile throughout the pod. Lenny recalled that when mobile was first taking off, Facebook would say “everything you show me now needs to be in mobile mockups.” Paul liked that sentiment but said AI is harder to show. It’s hard to show intelligence on a whiteboard.

​

Paul goes on to say that AI will be a big deal for any product that has a workflow (which is most) as well as any product that deals with media.
​

But his biggest wake up call same from a Sam Altman quote, “customer service is going to be the #1 impacted area.” He realized that Intercom’s product needed a foundational change.

So how should you be thinking about building products with AI? Paul says (and I agree) that you shouldn’t start with what the tech can do. You should start w/ the value prop you’re delivering for your customer. What problem are you solving?

​

Then ask yourself, can AI assist with that? Paul says your answer will likely fall within 3 buckets: 1) Yes, 2) partially, or 3) not yet.

​

For Intercom’s product thinking, they ripped up their existing strategy and went back to first principles, “why do people use Intercom?” They took a "bet the farm" mindset and totally changed how they think and work.
​

But something caught my ear, “Our biggest challenge is helping customers think about organizational change. The tech is way ahead, it’s actually like people wrapping their heads around what this means for the role, the teams.”

​

Intercom’s flagship AI product is “Fin,” which Paul claims helps customers answer 50%-70% of their inbound questions with it. Pretty wild if true.

​

But the more interesting part is that Paul says their AI products haven’t transformed their business financially yet. “You’d think it would be exponential, but it’s still so early that we're in the flat part of that graph.”

​

Paul recommends not thinking of AI as just “bolted on.” Meaning, don’t have AI specialists in a separate function, integrate them with the rest of the team. He is a big believer in generalists. He said they’ve even hired conversation designers to help with chat flow. Are these the new wave of social media managers?

How should people get started with AI? Paul responses with:

• You need to make it a priority and put in the hours
  • Greg’s note: This is similar to Rachel Woods tweet about tinkering with AI for at least 100 hours before you can start to get comfortable with it
• You need to stay up to date with the AI news (at least at a cursory level)
  • Greg’s note: Hopefully you get a slice of that pie from here ;)

​

What do you think? Agree/disagree with Paul? Anything surprise you?

​

Till next week 👋

In case you missed it

• Semantic Deduplicator - We all have lists of things: groceries, user feedback, surveys. I wanted a python package that would deduplicate them for me. But not by their keywords, by they semantic meaning. Check out my how-to video and the package.
• Didn't get into DevDay? Me neither, come hang out and watch it with me watch party + happy hour​

Greg Kamradt

​Twitter / LinkedIn / Youtube / Work With Me​

​Unsubscribe​