What does an AI developer do?

What is an AI developer?

An AI developer is a software developer who builds, tests and maintains applications in which an AI model plays a core role. Because an AI model gives a probable answer that can be wrong, a large part of the work is measuring, testing and catching errors. What should be built is a question for an AI consultant; how it's built well is the AI developer's question.

An AI developer is a software developer who builds applications where artificial intelligence plays a core role. Think of an assistant that answers questions based on internal documents, an application that extracts data from forms, or an agent that carries out tasks in other systems.

So what does an AI developer do differently from a regular developer? The difference lies in unpredictability. Ordinary software does exactly what you programme it to do. An AI model gives a probable answer, and that answer can be wrong. A large part of an AI developer’s work is therefore measuring, testing and catching: how well does it work, where does it go wrong, and what happens then?

The title is used broadly. Sometimes it means a machine learning engineer who trains models themselves. More often these days it’s a developer who deploys a vendor’s existing models and builds an application around them. An overview of the different profiles, from strategic to technical, is on types of AI specialists.

Does an AI developer train models themselves?

Sometimes: the title AI developer sometimes means a machine learning engineer who trains models themselves. Nowadays it's more often a developer who deploys an existing vendor model and builds an application around it.

Does an AI developer decide whether an application is worth the investment?

Typically not: an AI developer typically doesn't decide whether the application is worth the investment, doesn't choose between vendors on the organisation's behalf, and doesn't set up the governance. A different role is needed for that: an AI consultant works on the business case, vendor choice and governance, usually from the client's side.

Source: article by Dave Willemstein. Updated: October 2026.

Tasks

An AI developer’s tasks vary by project, but usually come down to six parts:

1. Translate the requestTurn a functional wish into a technical design with measurable requirements
2. Prepare dataUnlock, clean and structure sources so a model can work with them
3. Build the solutionChoose or train a model, write prompts and rules, write the application logic
4. IntegrateConnect the application to existing systems, including rights and security
5. EvaluateBuild test sets, measure quality, analyse and improve on errors
6. MaintainMonitor after go-live, adjust as data or the underlying models change

In practice, the centre of gravity shifts to the edges. Calling a model isn’t technically complicated. Making sure the right data is available at the right moment, that the application doesn’t see more than the user is allowed to, and that you can demonstrate how well it works: that’s where the craft sits. A messy data foundation is a common brake on that, as covered in the data foundation under AI.

An AI developer is typically not the one who decides whether the application is worth the investment, chooses between vendors on the organisation’s behalf, or sets up the governance. A different role is needed for that, covered further below.

Skills and tools

A good AI developer combines ordinary software skills with an understanding of how models behave. The main skills:

• Programming. Python is widely used in AI work, alongside knowledge of whatever languages and environments the rest of your software runs in.
• Working with data. Databases, data pipelines, cleaning and structuring. No usable AI without usable data.
• Working with models. Knowing when to use an existing language model, when to supplement it with your own documents, and when training is worthwhile. Formulating good instructions is part of that; that’s covered in prompt engineering.
• Integration. APIs, authentication, rights models and the systems you already use.
• Evaluation and testing. Building test sets, making quality measurable, and systematically analysing errors.
• Security and privacy. Knowing which data ends up where, and what GDPR and the AI Act ask of it.

On tools, an AI developer works with development environments, version control, cloud platforms, model vendors’ APIs, and frameworks for machine learning and for building applications around language models. Exactly which ones varies by organisation and changes quickly. A good developer chooses tools that fit your existing landscape, not the newest ones.

For anyone wanting to become an AI developer: a background in software development, computer science or data science is the usual starting point. After that, you learn the craft mainly by building, measuring and learning from mistakes.

AI developer versus AI consultant

An AI developer and an AI consultant work towards the same goal, but from a different question. The developer asks: how do we build this well? The consultant asks: should we build this, and if so, what, by whom, and how do we know it works?

	AI developer	AI consultant
Core question	How do we build this?	What do we build, and why?
Output	Working software	A well-founded decision, specification, vendor advice
Works on	Code, data, integrations, evaluation	Priorities, business case, vendor choice, governance
Usually sits	With the development partner or in your IT team	On the client’s side

I’m an AI consultant myself, not a developer. I don’t write code and I don’t sell implementations. In a development project, I help decide what should be built, guide the choice of a development partner, and check that the delivery does what was agreed. The building itself is done by your own team or by a development partner you contract directly. What a consultant does beyond that is covered in what does an AI consultant do.

That separation isn’t a formality. Whoever advises on what should be built, and earns from the build themselves, has an interest in a bigger scope. An independent adviser alongside a good developer gives you both: craftsmanship in the build, and a counter-voice on your side of the table.

For a good working relationship with AI developers, it helps to settle three things upfront: a specification with measurable requirements, a test set on your own data that stays yours, and agreements on maintenance and handover. Whether to use off-the-shelf software or have something built is covered in choosing or building AI software. How the whole process runs is on the page about AI development.

About to bring in an AI developer or development partner and want to talk through the brief first? Get in touch for a thirty-minute conversation.