The Future of AI is for Subject Matter Experts, Not ML Engineers

Here's my take on where AI is headed and why subject matter experts, not ML engineers, will be the ones building the next generation of AI applications. Most of this comes from the podcast above.

The Big Picture

One of the main predictions I base a lot of my work on is this: you're not going to win by building AI products with machine learning engineers anymore. The real AI builders that are going to exist in five years and beyond are non-technical subject matter experts. These are the people who are going to be interacting with LLMs and building the most effective AI systems.

Why Domain Expertise Matters More Than Technical Skills

Think about it - if you're building a legal contract AI tool, it's not going to help if you're a software engineer (like me) who doesn't know what these contracts are supposed to look like or understand the terms. You really want the problem solver to be in the driver's seat.

I love talking about Microsoft Excel because there are so many parallels here. Excel is just a database - they took a database and said "hey, let's let people who don't know SQL access all the things SQL can do" through a UI. That created whole fields of new types of knowledge work. Nobody's going to school specifically for Excel, but it's a tool in their toolbox that they're learning to use.

The Future of Work

The technology to automate 90% of jobs is already here. The really interesting question is: what's next? What is it going to create? My mental model is that AI-assisted roles and agents are going to be 10x, maybe 100x, maybe 1000x more productive. Instead of needing a hundred workers to do something, maybe you just need one.

Think about how many roles and types of work have been bottlenecked by language. Nothing's bottlenecked anymore because we've basically solved the problem of understanding and outputting language. If you think about a sales cold calling job, that's bottlenecked by how many people you can call in a day. That bottleneck no longer exists - now you can blast out conversations because we've solved language.

Building in Uncertainty

At PromptLayer, we're building the airplane while we're in the air while the laws of nature are changing. There are new models now, multimodal support, streaming capabilities, developer messages - it's crazy. You could spend your whole life predicting what's going to happen in the future, but anchoring to real life is always the best antidote. We stay heavily anchored to what our customers tell us, with a grain of salt - sometimes we need to give it a step function change, but we also need to be honest with ourselves that nobody might use it if it's coming from our brain and not from the customer.

The Rise of the Domain Expert Prompt Engineer

I think we're going to see therapists learning more of a computational way of thinking rather than coders learning more about mental health. You're going to have legal experts, doctors, and other professionals learning how to build AI systems and adding that skill to their resume. It's about teaching domain experts how to prompt rather than teaching engineers about specific domains.

We've seen this work firsthand. Teams come to us where the main prompt engineer is not technical, and when we dive deeper to understand how that happened, it's usually because the engineers on the team set up Prompt Layer and then handed it off to this new prompt engineer, saying "hey, click on this and you can change this."

Learn more about prompt engineering in our comprehensive guide.

The winners in the AI space are going to be the ones with the subject matter experts. If you're building an AI kindergarten teacher, you're going to lose if you're only hiring machine learning engineers. You're going to win by hiring kindergarten teachers and teaching them how to prompt. That's how you'll get more cycles, test more things, and ultimately build the best product.

Want to learn more about prompt engineering or try out PromptLayer? You can find me on X (@imjaredz).