← Video Library · AI Strategy
It’s Not Enough to Code—You Need to Speak AI
Published May 24, 2025 · 38 views on YouTube
Why do software engineers now need AI skills, not just coding skills?
Howdy.com co-founder Frank Licea says the strongest hiring trend he’s seeing isn’t demand for a separate class of AI specialist — it’s regular back-end and full-stack developers who also know how to interact with the latest AI tools, models, and simple tuning, and can integrate them with traditional systems like Elasticsearch and MySQL.
Do companies need specialized AI engineers, or just developers with AI skills?
Mostly the latter. Licea says his team expected a divergence between AI specialists and generalist engineers, but in practice they see rising demand for developers who act like a “Swiss Army tool” — capable of deploying software and features while also knowing how to work with LLMs, rather than a narrow specialist who only fine-tunes and deploys models.
What’s the biggest mistake companies make when hiring for AI roles?
The most common mistake is not thinking through a clear strategy before hiring: companies often say they want to “automate this whole workflow” without identifying the specific, smallest piece of that workflow where AI actually provides value. Howdy’s approach is to get surgical about the critical bottleneck first — for example, generating a synopsis for a software engineer — rather than building one large system meant to handle an entire workflow at once.
What does Howdy do to help teams figure out what they actually need?
Before recommending talent, Howdy learns about a company’s business objectives, existing infrastructure, and the specific problem they want AI to solve. From there, they can recommend whether a team needs a back-end engineer with AI experience or an actual AI specialist — treating “knowing how to interact with, tune, deploy, and improve these models” as being as fundamental to modern software engineering as knowing how to use GitHub.
Full video transcript
What trends are you seeing now in terms of AI talent demand, and how are your clients going to be using these AI-skilled engineers in their business? You just mentioned the tool itself — you want to set it up and train it — but there’s also this new skill set of the AI-powered software engineer. So what are you seeing in terms of trends for how businesses are hiring these types of individuals?
Yeah, the strongest trend that we’re seeing — one hypothesis that Jacqueline and I had at the beginning of this — is that we’re going to see an explosion of demand for specialized teammates who can fine-tune software and LLMs or deploy them, and that’s all they do. And we do see some of that. But actually what we’ve seen more and more in our pipeline is that people are looking for just a regular old software engineer — let’s say a Node back-end developer — who not only can deploy software and features to customers, but now also knows how to use the AI tools.
So we were expecting a bit of a divergence, and there’s still room for specialization, but much like we started with back-end developers and front-end developers and then over time, through necessity and efficiency, saw the rise of the full-stack developer who can switch between both — maybe not an expert in either one, but that Swiss Army tool, that utility tool you need on your belt to fix things as you go — we’re seeing that same demand now for back-end engineers who just know how to interact with the latest APIs, the models, do some simple tuning, and know how to integrate those things with traditional database systems.
Because what you find is a large language model is not a search engine — it’s not a text search engine. So you need to know how to marry a text search engine, the traditional Elasticsearch and MySQL stuff, and augment it with the new language model stuff. And yeah, I was pretty surprised by that. So we’re seeing it’s not enough to just be a back-end developer anymore. Now you need to be a back-end developer who knows how to interact with those tools.
The big hypothesis we’re seeing: when developers join at Howdy, we get them set up with an internal curriculum and philosophies to keep developers in the Latin American community up to date with the latest trends for joining teams on the US side. Often our engineers are even more ahead of the curve than the US engineers. And one of the things we hammer home is that if you’re a back-end developer, there’s no getting around knowing how to use GitHub — that’s where you collaborate, it’s where you put your code, it’s where you keep it safe. Very similarly, there’s no getting around knowing how to interact with, tune, deploy, and improve these models. It’s just going to be the bread and butter for any software engineer.
And so the companies that are looking to fill these types of roles — what are some of the biggest mistakes you’re seeing when they’re looking to hire out talent, whether it’s in the US, Latin America, or anywhere, hiring remote teams? How is Howdy helping them avoid these mistakes before they even happen?
Yeah, so the big mistake I think people make is not really thinking through exactly how they want to use these AI tools — thinking through a clear strategy for where they want to deploy them and the status of their work. So what we do to help companies think through this — I can speak to it from a software engineering point of view — when they come on board, we have to know a little bit about their business, a little bit about the kinds of objectives they’re trying to accomplish if they’re trying to deploy AI, and the existence of their existing infrastructure and how that works.
And we make sure it’s clear to them that, number one, they have to have all the things we talked about — the right environments, the right context, the right documentation. Number two, it’s got to be very clear the specific problem they want to use it for. Oftentimes they come to us describing “we want to automate this whole workflow.” But that’s not enough to just say you want to automate the whole workflow. When you really unpack it, you figure out that there’s a smallest piece there that provides a lot of value — the critical piece, or the critical path, of that workflow. And oftentimes, in our case, that ends up being something like creating a synopsis for a software engineer — a critical piece among a bigger workflow.
But if we tried to solve the whole “let’s automate that entire workflow” from the beginning, without really thinking about where the value is and where the bottleneck is, it would end up as this huge, expensive, clumsy system trying to reason about poor results that just never work. But because we’re surgical about where it applies — combined with knowledge about people’s or teams’ existing systems — we can, through a bit of consulting and question-answering, actually recommend the kinds of software engineers and skills they might need. They may just need a back-end engineer with AI experience, or they might need an actual AI specialist. Those are all things we can help teams reason through.