For a long time, technology and linguistics seemed to belong to different worlds.

On one side were programmers, engineers and data specialists.

On the other were linguists, translators, writers, philologists and communication professionals.

Artificial intelligence is making that boundary much less clear.

Today, a machine can summarize a legal document, translate a conversation, analyse thousands of customer comments, generate code from a written instruction or hold an almost natural conversation.

And the better technology becomes at understanding human language, the more important it becomes to understand how that language actually works.

The real novelty may not simply be entirely new professions.

It may be the new intersections between disciplines that used to work separately.

Computational linguistics did not begin with ChatGPT.

It is a long-established field combining linguistics, computer science, statistics, artificial intelligence and natural language processing.

Its applications include language analysis and generation, machine translation, speech systems, information retrieval, text classification, information extraction and model evaluation.

The Association for Computational Linguistics maintains an international research and professional community around the field, and in 2026 it is explicitly examining how large language models are changing work and professions.

A good language system does not need good programmers alone.

It also needs people who understand how language behaves, what an expression means in context and what can be lost when a machine turns human intent into data.

“

A machine can process words. Understanding why we choose one word rather than another is a different problem.

For years, we talked about big data as if having enormous amounts of information were enough.

It is not.

We need to understand what the data contains, how it is structured, how it has been labelled, how reliable it is, what biases it may contain and which languages it represents.

This increases the importance of roles such as data analysts, data scientists, data curators, data annotators, language-data specialists and AI evaluation specialists.

Many of these jobs are not completely new. They are evolving professions.

What has changed is the context.

The U.S. Bureau of Labor Statistics projects 35% growth for data scientists between 2025 and 2035, far above the projected average across occupations.

The linguistic side is becoming more formal too. ISO is developing ISO/IEC TS 26320, a project focused on principles and methods for developing and maintaining corpora used in natural language processing systems and assessing their quality.

Few terms have become as popular as “prompt engineer”.

The idea sounds simple: a person who specializes in writing instructions for artificial intelligence models.

But the reality is more interesting.

A 2025 study analysed 20,662 AI-related LinkedIn job postings and found only 72 positions with the exact title “Prompt Engineer”. That was fewer than 0.5% of the sample.

This does not mean prompt engineering is an empty trend.

It may mean something more interesting: prompting is moving from a standalone function into a skill embedded across many professions.

Designers use prompts.

Data analysts do too.

Researchers.

Developers.

Marketing professionals.

Linguists.

Product designers.

Teachers.

The important skill is not knowing a magic formula for the perfect prompt.

It is understanding the problem, providing context, defining criteria, checking results and recognizing when an AI answer is not good enough.

“

The prompt is a tool. Judgement is still human.

As technology becomes more complex, there is a growing need for people who can connect different ways of thinking.

A company has a problem.

A user explains it in their own words.

A designer thinks about experience.

A developer thinks about architecture.

An analyst thinks about data.

The AI model works through patterns.

Someone has to make those worlds understand one another.

This is where profiles such as AI product specialist, conversation designer, AI trainer, AI evaluator, knowledge engineer, language technology specialist and AI implementation specialist are emerging or evolving.

Some are already used as job titles.

Others are functions appearing inside existing roles.

They may not all have the same name five years from now.

But they share an important characteristic:

they work across disciplines.

Chatbots and AI assistants have turned conversation into a new kind of interface.

A conversational system has a voice, rules, boundaries, errors and states.

It needs to know when to ask, when to explain, when to admit uncertainty and when to stop talking.

Designing these conversations combines linguistics, UX, cognitive psychology, product thinking and technology.

Conversation can be designed too.

And the more we use language-based interfaces, the more important it becomes to think carefully about how machines speak and respond.

It is not enough for an answer to be grammatical.

It needs to be useful, appropriate, clear and consistent with its context.

“

When conversation becomes an interface, words become part of the design.

Generating an answer is relatively easy.

Determining whether the answer is actually good is much harder.

Is it correct?

Relevant?

Safe?

Too long?

Appropriate for the audience?

Equally effective across languages?

Able to understand a regional expression?

Consistent when the same question is asked differently?

In 2026, NIST launched its Artificial Intelligence Technology Evaluation programme to evaluate AI models using sequestered test data, diverse tasks and common metrics.

ISO is also developing specific methods for evaluating the quality of natural language processing systems.

That shows an important shift.

The question is no longer only:

“Can it do it?”

It is becoming:

“How do we know that it does it well?”

An AI system can work extremely well in English and be less consistent in another language, even when its grammar looks correct.

Languages are not simply different dictionaries.

They change structure, register, politeness, irony, humour, cultural references and professional conventions.

The amount and quality of available data also vary between languages.

UNESCO has warned about the risk that systems trained mainly on dominant languages can reproduce linguistic hierarchies and reduce diversity when they should be learning from it.

That is why multilingual AI needs more than automatic translation.

It needs linguistic and cultural knowledge.

For people who combine language expertise with technology skills, that combination could become a major professional advantage.

“

Translating words does not always mean transferring meaning.

The World Economic Forum identifies AI and big data among the fastest-growing skills toward 2030, while also highlighting analytical thinking, creative thinking, flexibility, resilience and lifelong learning.

That combination matters.

Not everyone needs to become a programmer.

Not everyone needs to become a linguist.

The opportunities appear in the intersections.

Linguistics + AI.

Design + data.

UX + conversation.

Research + automation.

Communication + models.

Technology + culture.

AI is also creating an interesting paradox: the better it becomes at generating output, the more valuable human judgement may become when distinguishing an acceptable answer from a genuinely good one.

Perhaps we are not only creating new professions.

Perhaps we are creating new intersections.

Language and programming.

Data and communication.

Design and artificial intelligence.

Linguistics and machine learning.

UX and conversation.

Research and automation.

Technology and culture.

The Association for Computational Linguistics is already examining how large language models, task automation and AI systems are changing work.

At the same time, recent research is beginning to use language models to detect new skills emerging in labour markets.

This points to something we can already observe:

Professional boundaries are becoming more porous.

Some professions will disappear.

Many will change.

Others do not have names yet.

But one skill keeps appearing:

knowing how to work between worlds.

Stanford HAI reports that organizational AI adoption reached 88% in 2025 among surveyed organizations and that generative AI was used in at least one business function by 70% of them.

That means knowing how to use an AI tool will probably stop being a rare advantage.

What will make the difference is what comes next.

Asking better questions.

Structuring information.

Detecting errors.

Providing context.

Building workflows.

Evaluating results.

Working with specialists from other disciplines.

And, above all, knowing when not to trust the answer.

Technology is becoming better at understanding our words.

Now we need to become better at understanding technology.

Not to speak like machines.

But to make machines work better with us.

Stanford HAI — 2026 AI Index Report.

World Economic Forum — Future of Jobs Report 2025.

U.S. Bureau of Labor Statistics — Data Scientists, Occupational Outlook Handbook.

Association for Computational Linguistics — Future of Work in the Age of LLMs and computational linguistics research resources.

Vu, An & Oppenlaender, Jonas — Prompt Engineer: Analyzing Hard and Soft Skill Requirements in the AI Job Market.

ISO/IEC — Corpus development and maintenance for natural language processing systems, ISO/IEC AWI TS 26320.

NIST — Artificial Intelligence Technology Evaluation (AITE).

UNESCO — work on AI, linguistic justice and language diversity.