Eoghan McCabe
speaker
119 appearances
1 recordings
1 series
first heard Oct 2024
last heard Oct 2024
Eoghan McCabe’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
Most VC money is going to support a lot of excitement around AI and actually not do better than the S&P 500. I think that building your own AI agent platforms is going to be as smart as in the previous generation, building your own SaaS workflow. A terrible fucking idea. I'm a big believer in individual liberty and peace. or in a time where both of those are being threatened.
It's crystal clear that Trump and fans are the people who are promoting the most personal liberties and freedoms. I think it's deeply evil to decide what can or cannot be spoken about. It's really dangerous. Governments become insanely and dangerously powerful when their citizens cannot say what they believe is true or not.
Thank you for having me back.
Said with love. I appreciate that. I mean, everyone's fucked by AI in the fullness of time. I mean...
That's the intro, that's the segment. Unfortunately, I regret that already. Yeah, in the fullness of time, increasingly powerful and abstract AI will do away with all of the work we need to do today to deliver software experiences. That has always been the case with technology. There's no new technology that won't be obsolete in some period of time.
And I wouldn't be working my ass off in Intercom if I thought that it was about to die and be crushed by AI in a year or two, so... I think we've got a decade, two decades.
You know, the reality is that adoption of new technology takes far longer than we always imagine. We were the first to come out with an AI agent that would do customer support for you, like actual conversations with customers and closing tickets. We now have, forgive the fucking pitch, we now have the highest performing AI customer service bot.
You look at everything that's on the market, they can't resolve tickets at the rate we can. The average resolution rate of our bot is now approaching 50%. How is that, Owen?
Yeah, it's all the existing data sources. With intercom, it's your historical conversations that your human reps are also having. But what's really interesting and what's key to understanding how AI is going to develop, and we can go back to your questions if I manage to remember, is that building domain-specific AI systems requires a phenomenal amount of work on top of these LLMs.
Of course, yes, over a decade or two, the LLMs will get powerful enough that they will disrupt away all of these detailed intricacies that we've built. But Finn, our AI agent, for example, is the product of over 100 unique experiments to figure out how exactly to understand customer questions and resolve tickets. There's patented components to it.
I think there are somewhere around seven different LLMs intricately connected together. There's homegrown AI, and just a lot of very detailed prompt engineering and massaging to make it highly effective at closing tickets. When we first released our AI agent, the average resolution rate was in the high 20s.
So that's the rate at which it successfully closes tickets when it gets involved in the conversation. And now, like I said, it's approaching 50%. I think Zendesk is still high 20s. And so it just takes a long period of time and a lot of work to build all of the functionality on top of the LLMs to make them really, really great at these domain-specific things.
That's why these AI application companies are going to win. That's why they're going to become so big, because I don't think, at least in the next five, 10 years, that any of the AI labs are going to build all the domain-specific stuff.
Yeah, I think that's right in the fullness of time. And again, if you held on to your stock in fax companies, you'd be crushed. But for the first 10 years, you would have made a fucking fortune. So in technology, it's all about timing and it's all about waving success of growth cycles and riding the trends in that moment in time. And then you must reinvent yourself and kill yourself.
You asked an interesting question at the start. We've already forgotten what it is. Well, I asked the question, which is like, why are you not fucked by AI? And that's the answer. It's because sure, AGI crushes all of software in the fullness of time, but it takes a really long period of time to pick up these technologies.
There's various aspects of it. The rate at which AI is going to change the whole world is a little overhyped. You know, the raw technologies will be built long before they have real impact in the world. That said, even the technology community at large has not yet made contact with technology. how valuable AI already is.
There are 100, 200, maybe there are 500 companies globally that have built AI-first products that are blowing up. I saw some statistic that showed that the average time it took for a SaaS company to reach 30 million an hour or from a specific investor who was picking great SaaS companies was something like four or five years.
And the average time for these AI companies to reach 30 million ARR was less than two years. AI application companies are currently blowing up in a way that people don't realize. They're currently delivering a lot of value.
Of course, when anything new happens, any new trend, there's going to be so much noise. And it's your job to be one of the few investors who can spot the signal and the noise and beat the market. Most VC money is going to support a lot of excitement around AI and actually not do better than the S&P 500.
Yeah. Now, they were the very early days. And still, this has to be the very early days of AI. And there will be a lot of AI companies that totally flame out. I stand by the assertion that there are AI companies delivering really tangible value. And the things you pointed to are exactly right. Usage and retention is how you spot the difference.
Showing 1–20 of 119 · page 1 of 6
Next →