First AI Hire

Most companies get their first AI hire catastrophically wrong. Not because they hire bad people - but because they hire the wrong people for what they actually need.

Here's what typically happens: a business leader decides they need "someone for AI." They ask ChatGPT to write a job description. The AI spits out what a frontier research lab might have been looking for eighteen months ago. Suddenly, you're requiring PyTorch expertise, transformer architecture experience, and a PhD in machine learning - for a role that's actually about rolling out Microsoft Copilot and building some automations in n8n.

The result? You either can't fill the role at all, or you hire an overqualified researcher who's bored within six months because there's no novel ML work to do. Meanwhile, the practical AI adoption work sits undone.

This guide takes a different approach. We'll help you figure out what you actually need, find someone who can deliver it, and set them up for success.

Unless you're building novel AI models from scratch - training your own LLMs, developing new architectures, pushing the frontier - you don't need a machine learning researcher.

Most businesses need someone who can:

• ➢Evaluate and select AI tools
• ➢Integrate those tools into existing workflows
• ➢Build automations and internal applications
• ➢Train teams on effective usage
• ➢Establish governance and quality controls
• ➢Measure outcomes and iterate

This is valuable, skilled work. But it's not research. It's closer to a technical product manager or solutions architect who happens to specialise in AI.

The AI field is maybe three years old in its current form. GPT-3 shipped in 2020. ChatGPT launched in late 2022. The tools, techniques, and best practices are still being invented.

This means:

University degrees are already outdated. Even recent graduates learned from curricula designed before the current wave of foundation models. A 2024 CS graduate's coursework was likely finalised in 2022 - ancient history in AI terms.

Traditional credentials don't map to this work. A PhD in machine learning prepared someone for research, not for rolling out Copilot to a sales team. A software engineering background is useful, but doesn't guarantee understanding of LLM capabilities and limitations.

The most competent practitioners learned by doing. They built things. They experimented. They stayed obsessively current with a field that changes monthly. You can't learn that in a classroom.

This doesn't mean credentials are worthless - they indicate baseline capability and learning ability. But they're insufficient. And over-weighting them filters out many of the best candidates.

Capability - measured by what they've built, not what they've studied. A portfolio of working projects beats a prestigious degree. Ask: "Show me something you've shipped with AI."

Enthusiasm - this field moves fast. Someone who's genuinely excited will keep themselves current without being told. Someone who treats it as just another tech stack will fall behind within months. Ask: "What's changed in the last three months that you're excited about?"

Accountability - AI work requires navigating ambiguity, managing stakeholders, and delivering results without constant supervision. A few years of industry experience in a loosely related role (product, engineering, operations, consulting) provides evidence they can operate this way. Ask: "Tell me about a project where requirements were unclear and you had to figure it out."

You're not looking for a unicorn. You're looking for someone smart, curious, and practical who's been paying attention to AI and can get things done.