What to look for when hiring an AI developer

If you've decided your business needs an AI solution, the next question is: who builds it? The market is flooded with people calling themselves AI developers, AI consultants, AI strategists. Some of them are excellent. Many of them learned to use ChatGPT six months ago and now charge £150/hour for prompt engineering.

Here's how to tell the difference.

The questions to ask

"Can you show me something you've built?"

This is the single most important question. Not a slide deck, not a proposal, not a blog post about AI trends. An actual working system. A dashboard, an API, a pipeline, a deployed tool.

The answer tells you whether they build things or just talk about building things. If they can't show you anything, that's a red flag regardless of their credentials.

"What would you NOT use AI for?"

Good developers know the limits of their tools. If the answer is "AI can solve everything," run. The best AI developers will tell you when a spreadsheet formula, a Zapier workflow, or a junior hire is a better solution than a model.

"What happens when it's wrong?"

AI systems are probabilistic. They will make mistakes. A good developer plans for this: confidence thresholds, human review queues, fallback logic, monitoring for drift. If they promise 100% accuracy, they either don't understand the technology or they're not being honest.

"What are the ongoing costs?"

A model needs hosting, API calls, monitoring, and occasional retraining. If they only quote the build cost and wave off the running costs as "minimal," push back. Get specific numbers: monthly hosting, estimated API spend, their maintenance hourly rate.

"How will you hand this over?"

Unless you're hiring a permanent team member, the developer will eventually leave. The system needs to be maintainable by someone else. Ask about documentation, code quality, and whether they'll use standard tools your team can support.

Red flags

All theory, no practice. They talk about transformer architectures and attention mechanisms but can't explain how they'd solve your specific problem in plain English. Academic knowledge is fine, but you need someone who ships working software.

Overpromising on timelines. "I can build that in a week" for a project that involves data cleaning, model training, integration with your existing systems, testing, and deployment. Two weeks minimum for even a simple ML project. More realistically, 4-8 weeks for something production-ready.

No questions about your data. If they jump straight to proposing a solution without asking about your data quality, format, volume, and accessibility, they're selling you something they've already decided on. Good developers start with the data because that determines what's possible.

Buzzword density. If every sentence includes "AI-powered," "machine learning," "deep learning," "neural networks," and "large language models" without specifics, they're probably marketing rather than engineering. The best developers use plain language.

No mention of testing or monitoring. Building a model is maybe 30% of the work. The rest is testing, validation, deployment, monitoring, and maintenance. If their proposal only covers the build phase, the project will launch broken and stay broken.

Green flags

They ask lots of questions first. About your process, your data, your team, your budget, your timeline. They want to understand the problem before proposing a solution.

They suggest the simplest option. If rules-based automation would solve your problem, they say so, even though it means a smaller project for them.

They talk about tradeoffs, not guarantees. "The model will be about 90% accurate, so we'll need a human review step for the remaining 10%" is more honest and more useful than "our AI is highly accurate."

They've worked with real data. They can talk about data cleaning, feature engineering, and the messy reality of working with business data. Not just clean datasets from Kaggle.

They include ongoing support in the quote. Because they know the system will need monitoring, occasional fixes, and retraining as your data changes.

Freelancer vs agency vs in-house

Freelancer (like me): Best for specific, bounded projects. Lower overhead, direct communication, faster turnaround. The risk is that one person has limited bandwidth and might not be available when you need urgent changes.

Agency: Best for larger projects that need multiple skill sets (design, engineering, data science). Higher cost, more process, but more capacity. The risk is that the senior person who sold you disappears and a junior does the work.

In-house hire: Best if AI is core to your business and you'll need continuous development. Expensive to recruit and retain, but you get full-time attention and deep knowledge of your domain. Only worth it if you have enough work to keep them busy.

For most small businesses exploring AI for the first time, a freelancer or a small, focused agency is the right call. You don't need a full-time data scientist for a two-month project.

What to expect on pricing

Rough UK ranges for AI development work:

• Simple automation (rules-based, no ML): £2,000-5,000
• ML model (classification, prediction, NLP): £5,000-15,000
• End-to-end system (data pipeline + model + deployment + monitoring): £10,000-30,000
• Ongoing maintenance: £500-2,000/month depending on complexity

These are wide ranges because every project is different. But if someone quotes you £50,000 for a support ticket classifier, or £500 for a predictive analytics platform, something's off.

Need help scoping a project?

If you're trying to figure out what to build, what it should cost, and who should build it, I can help with that. Book a free initial chat and I'll give you an honest view of what your project actually needs.

---

Related reading:

• How I'd run an AI audit for small business operations
• Is AI worth it for a small business? Here's how to actually decide.
• My AI consulting service: project scoping and honest assessments