Aaron White Dives Deep on Navigating AI's Horizon: Future-Proofing Products in the Age of AI

Navigating AI's Horizon: Future-Proofing Products in the Age of AI

Navigating AI's Horizon: Future-Proofing Products in the Age of AI

We're continuing our chat with AI enthusiast, Aaron White in part 2 of our series. This time, we're diving into how different AI products measure up against the ever-evolving S-Curve.

Before we jump into it, following our podcast summary, we've got a BONUS (edited for time) chat where Aaron, Fareed, and Brian weigh in on the advantages of the top players in the AI race.

This is exclusively posted to our YouTube page, so after you finish listening or watching this episode, make sure to check it out. 📺

🚀 How far ahead should we plan for?

It's a delicate balance. Build too far into the future and you risk creating a product that doesn't yet have a market. Don't build far enough and you risk being left behind by rapidly advancing technology.

🤖 Today, there are three dominant types of AI products:

1. Co-Pilot: The human is the driver and the AI assists.

2. Human as Guider: The AI does the bulk of the work, but humans guide it.

3. AI as Agent: You give the AI a goal and it plans and executes.

Today, each of these options comes with its own set of pros and cons. Co-Pilots and Human as Guider products may seem tailored to the current state of AI. However, many companies opt for building Agents. They understand that while the output may not be flawless now, advancements in technology will eventually surpass Co-Pilots and Human as Guider products, which only take the task partway to the finish line. 🏁

🎯 Aaron suggests beginning with the end in mind.

“Here's my take. I think that this technology is advancing so rapidly that a lot of classic application builders are not taking a forward enough view of how to build the value they provide their users such that they're going to find themselves left in the dust before the current cycle has even finished.” - Aaron White

Taking a leaf from Stephen Covey's "Seven Habits of Highly Effective People," AI products should be designed with a clear end goal. This allows for a more focused approach to problem-solving and can lead to more efficient and effective solutions.

Aaron defines the end goal as accomplishing the task for the human.

⚖️ Intercom’s Fin solves the end state Job to be Done, while Jasper doesn’t today.

In Aaron’s view, popular start-up Jasper, while it made a big splash, was quite innovative, and does the job it sets out to do well, is unfortunately not built with the end in mind. What marketers really want is a tool that can end-to-end improve its SEO and conversion. Jasper is multiple steps away from that, and Aaron feels someone will make this happen.

“Let's say you're a marketer using Jasper to create blog posts for the purpose of SEO, for the purpose of attracting customers. The job to be done isn't to assist me writing blog posts… the job to be done is ‘I want articles that convert meaningful SEO traffic from my ICP…The real use case for marketers is just to build out a long tail SEO funnel for me automatically and plausibly.” - Aaron White

Conversely, Aaron feels Fin, Intercom’s AI offering, can largely replace human customer support teams. It really only needs humans to handle exceptions, which is the end goal of the Job To Be Done.

🔮 The Innovator's Dilemma: How do we know if the alpha will last long enough to be worth it?

Brian pushed back on Aaron’s view of Jasper, asking “how far ahead should we plan?” For instance, imagine if Jasper or someone actually created a product that generated high-converting content tailored for ICP and SEO. Once that happens, the SEO space will quickly fill up with such content, diminishing its value rapidly, right?

🏃David’s typical advantage of speed appears to favor Goliath this time.

Typically in a startup vs incumbent scenario, the key advantage a startup has is speed. What has Fareed spooked here is that in this battle, it seems to be the opposite.

"I've seen two things that seem scary to start a company or build something new in this space. One is that the tech is advancing so quickly. It can very quickly be subsumed by new stuff from the biggest players, from the foundational model creators, from the largest companies, etc. There's a whole slew of stuff that seemed good because GPT-3 could only do X, Y, and Z. And then four comes out, and it's like, whoops - my whole value proposition doesn't matter because it's been subsumed by a thing. Then there's the second, which is that the incumbents have been very quick to add productivity-type, copilot-type stuff to their products, which has meant being the same as Microsoft Word, but with copilot it's not good enough.” - Fareed Mosavat

🤔 Exceptions are the exception here.

In a world where Agents can largely replace Humans, the question arises: How do Humans contribute? Aaron's insight - Exceptions.

Humans shine in handling exceptions, where they bring value even to the most efficient Agents. However, Aaron believes that once an exception is resolved, the agent should learn from it, empowering them to tackle similar challenges next time.

🛡️ Some functions are more vulnerable than others.

Brad Gerstner and Bill Gurley's new podcast, BG Squared, featured Aaron Levie, Box's CEO, shared an insightful observation. He noted that certain software engineering sectors, including customer support, share a unique combination of traits:

1. Work performed in text editors

2. Large online training datasets

3. A tendency toward structured output

While this applies less to support than coding, Levie argued these factors create an ideal setting for technologies like co-pilot products or Fin, which automate tasks. However, as Brian notes, most knowledge work lacks these elements, diverging into areas requiring more nuanced or strategic thinking, where the vast online resources aren't as applicable and exceptions are the norm.

🚲 Knowledge work relies on tacit knowledge.

Fareed pulls in a framework he stumbled upon over a year ago from Cedric Chin. Cedric introduces the concept of tacit knowledge, emphasizing learning through experience rather than mere instruction. For example, truly understanding bike riding requires actual practice, beyond verbal explanations. Chin highlights that business expertise often revolves around tacit knowledge.

At Reforge, we frequently encounter individuals seeking concrete guidance, yet realizing solutions depend on direct experience and context. With many forms of knowledge work, every situation feels like an exception.

🎨 The future favors the creatives OR those in roles where exceptions are the norm

AI is going to make it harder for professionals to hide behind structured documents or presentations. If the core idea isn't unique or creative, AI will expose that. This could lead to a shift in the workforce, where creativity and the ability to handle exceptions become even more valuable. Jobs that allow for creativity and handling constant flows of exceptions will be more defensible against automation.

🎬 Could we see the rise of niche businesses and full-stack services?

As AI continues to advance, we're likely to see an explosion of niche businesses, each doing more creative things on average per person. Perhaps we’ll also see companies that won’t just provide software tools, but actually offer complete solutions that solve specific problems. From AI-powered law firms to animation studios, the possibilities for the future are endless.

YouTube Exclusive: Who Will Win the Incumbent AI Battle? 🥊

The AI battle is heating up with OpenAI, Microsoft, Google, and potentially Apple, all vying for the top spot. But who will come out on top? 🤔

🎯 OpenAI's Creative Advantage 🎯

OpenAI, despite the chaos, is still attracting top talent in the market. In a world where creativity could be the winning factor, OpenAI's bold and sensational approach might just give them the edge. After all, in the modern world of content, divisive, polarizing, and sensational content wins eyeballs, right? 🧐

Their weakness is pretty straightforward - they lack a tangible connection to the end customer like the other three. This is likely why Sam Altman is delving into hardware…

🍏 Apple's Hardware Advantage 🍏

Apple, if they enter the race, could do incredibly well. They own the terminal relationship with the consumer on the most fundamental level - the hardware. If they can move AI onto the device, they could continue winning the aggregation of digitally mediated value. 📱 The question is how quickly can their former car team catch up here?

🌐 Google's Search Threat 🌐

Google does have hardware and immense data, but we wonder if Google's advantages might be overstated. Search is under threat, and unlike the other three companies, this will force Google to draw a line in the sand at some point. Google's core business might be deeply constrained by their cash cow, making it hard for them to go all in and risk further exposure on Search. 🕸️

💻 Microsoft is looking pretty… pretty… pretty good. 💻

Microsoft, with its install base around Copilot and ownership of 75% of the desktop operating system market, is in a good position. They have an advantage that is less disrupted by this technology and more complemented. Plus they are hedging their bets from a tech perspective with a strong internal team, recent investment in Mistral, and… of course, their OpenAI position. 🖥️

So, who will ultimately win? Or will there be multiple winners? It's a tough call, but the AI landscape is about to get a whole lot more interesting. We have our popcorn and our microphones at the ready…🍿 🎙️