Fortifying Your AI Business Innovation
by Clay Turner, Co-Founder
Last month, OpenAI announced and immediately released a preview of its GPT4-Turbo model, along with a torrent of new tools. And in an instant, POOF! The misplaced hopes of countless startups went up in smoke like a dumpster fire.
First and foremost, a vanilla implementation that's based on a Youtube video that you watched is unlikely to have much staying power as a viable model. Last week's OpenAI announcements prove it. I'm also reminded of something that I witnessed while attending a recent AI presentation... I watched someone stand up and interrupt the speaker to demonstrate a chatbot that he conjured based on the speaker's published presentation notes - right in the middle of the presentation. Some people in the audience were impressed. I was not. I knew exactly what he had done. Anyone who has watched a Youtube video, read an article on Retrieval Augmented Generation (RAG), or can even marginally code using some of the simple wrappers that LangChain offers could have pulled it off. Yet, it was clear that his goal was to garner business support - to capitalize on "the hype." I commend his boldness, but you have to consider the audience... That I was unimpressed with his very public display was far less significant than the fact that VCs in attendance were unimpressed. You see, even though AI and all of the buzz seems so new, many applications of AI are quickly becoming saturated and really rather passé to educated VCs - particularly models that center on knowledgebase chat (chat for your data), content creation, etc. At least in its general application... If there's a Youtube video, then someone else is doing it. Lots of people are doing it. Remember too that these introductory videos usually present simple machine-bound examples that require a lot of additional engineering and experience to actually scale affordably in production. So use these examples for "general" AI application to build an understanding of what "can" be done, but consider where and when it "should" be applied to meet a "specific" business need. VCs are not interested in vitamins. They are interested in pain killers. You should be too.
Something else that I've observed rather consistently in recent startup pitches is an obvious need to say something - anything - about how they're AI-enabling their product. For some, AI was an obvious afterthought, which is totally understandable given its popularity rose astronomically while many of these startups were in the middle of product development. Don't get me wrong here... Any new product innovation must ask a myriad of AI-related questions now. These are healthy questions. Where can an AI-enabled feature add real value for customers and, conversely, should we include one that really doesn't make our customer's life better? Where can we use AI to get to market faster, better, and more efficiently? VCs are keen on seeing you extend your runway without diminishing your acceleration, and AI makes this possible. If after careful consideration you find that AI shouldn't necessarily transform your "product" innovation, perhaps (absolutely) it can transform your operational capacity to deliver your product or service. Assume that your competitors - known and unknown alike - are considering the same.
AI innovation is likely to be more fortified within a specific vertical application. The real opportunity for most of us who can't afford to create and pretrain a Large Language Model natively exists in the countless business workflows that can be overhauled with targeted, specific application of AI and the business workloads that can be automated by AI agents. Here, the landscape is wide open. Every mundane business problem out there - in even the least sexy industries and sectors across the planet - is an opportunity for someone to carve out a niche and apply the technology in very rewarding ways. Taking a bit of time to excavate these gems instead of running towards the apparent "city of gold" on the horizon can be worthwhile. The "city of gold" looks great until you're alone and hungry in the jungle. Operating within a given vertical also tends to deal with or create unique data. Unique data can absolutely help to fortify your position and make a stronger argument to investors.
Where it makes sense to reconsider chasing the horizon, there are highly sought after applications of AI buzzing right now that you should be considering regardless. Responsible AI is a hot topic for all. Bias is real, and where there are some obvious extreme examples, there is a clear danger in the more subtle examples as well. Corporations are quickly beginning to adopt policies that demonstrate a commitment to using responsible AI in their decision support. Some industries will feel this urgency more than others and seek to get ahead of legislation that is coming without a doubt. Likewise, any interesting application of AI that speaks to addressing the global climate crisis is getting VC attention. There are some really big problems that we must tackle together as innovators, and no single juggernaut is likely to capitalize on the effort alone.
If you are one of those very fortunate people who can afford the millions of dollars required to create, pretrain, and/or fine-tune a unique model, then you may be in a well-fortified position already. If your IP is difficult and expensive to reproduce, speaks to a real pain, and has a large market opportunity primed for rapid growth, then you may be in a stronger position than I could even begin to speak to from my own experience. That's just not most of us.
Where there is so much more to say about this topic, I'll stop here and simply say this: Be cautious, but get on with it. All sound business principles continue to apply here, but have been intensified like nothing I've seen in my nearly 30 years.