AI and your next move · 6 min read
Is prompt engineering a real career?
Prompt engineering as a standalone job is narrowing; as a skill inside other jobs it is spreading. Here is what that means for you and what to build instead.
The short answer
Prompt engineering exists as a skill, and some companies hire for it, but as a standalone career it is narrow and changing fast. As AI models get better at understanding plain requests, the durable value moves to related work: designing and evaluating AI workflows, knowing a domain well enough to judge outputs, and building systems around models. Treat prompting as a core skill to add to a role, not a whole career to bet on.
On this page 7 sections
Skill or job?
Writing clear instructions for an AI model, with context, examples and constraints, is a real skill. It is also one that more and more people are expected to have, much as spreadsheet skills spread from specialists to almost everyone. Skills that spread widely tend to become part of other jobs rather than jobs in their own right.
Where the durable work is moving
- AI workflow design: deciding which steps of a process AI should do, with what checks.
- Evaluation: building test sets and rubrics to measure whether outputs are good enough.
- Domain-specific AI work: legal, medical, financial or educational expertise applied to AI tools.
- AI engineering: building applications that combine models, data and tools, for people who code.
A test of whether this direction suits you
Choose a task you know well, such as summarising meeting notes into action items. Write a prompt, run it on five different sets of notes, and score each output against a checklist: did it catch every action, assign owners correctly, avoid inventing anything? Revise the prompt and score again.
If you enjoyed designing the checklist and finding failure patterns, look at evaluation and workflow design. If you enjoyed the domain judgement most, look at AI roles in your own field. If you wanted to automate the whole thing with code, AI engineering may suit you.
What a five-sample scorecard looks like (illustrative)
| Sample | Caught every action? | Owners correct? | Invented anything? | Notes |
|---|---|---|---|---|
| 1 | Yes | Yes | No | Clean |
| 2 | Missed one | Yes | No | Action buried in a side discussion |
| 3 | Yes | One wrong | No | Two people with the same first name |
| 4 | Yes | Yes | Yes | Added a deadline nobody agreed |
| 5 | Yes | Yes | No | Clean |
The revised prompt might ask the model to quote the sentence each action came from, to use full names, and to write “no deadline agreed” rather than guess. Re-scoring shows whether those changes help. This loop, define quality, measure, change, measure again, is the core of evaluation work, and it transfers across models and tools.
Reading job ads with the title
When you see “prompt engineer” in a job advert, read the responsibilities. If they mention building test sets, evaluating outputs, working with product managers or writing code, the role is evaluation, product or engineering with prompting inside it. If the role is only writing prompts with no measurement, ask how success is judged; the answer tells you how durable it is.
Who this direction tends to suit
- People who enjoy precise writing and noticing small errors.
- People who like designing tests more than producing content.
- Domain experts who can tell good answers from plausible ones.
Questions people ask
- Do companies hire prompt engineers?
- Some do, often combining prompting with evaluation, product or engineering responsibilities. Read job descriptions closely; the title can mean very different things.
- Is a prompt engineering course worth it?
- A short course can teach useful techniques. Be wary of expensive programmes promising a job. Practising on real tasks and measuring results teaches more.
- What is better to learn than prompt engineering?
- Evaluation (how to tell whether AI output is good), workflow design, domain expertise, and for technical people, building applications with AI models.
- Will prompt engineering skills still matter?
- The ability to specify what you want clearly, give context and check results is likely to stay useful. The tricks for any particular model change quickly.
Exercises are practice, not evidence of career fit or a qualification. How to read our guides and report a correction.
Sources
- Future of Jobs Report 2025: digest · World Economic Forum
How this page is written
Sources support the factual background, not a guarantee that an exercise predicts career fit. Fictional practice material is labelled. Dates record content updates; no independent expert review is implied. More about OffLadder.