AI is making a stormy advance in the communications department. But only a handful of professionals actually achieve a real return on it. The key? Understanding that AI is not a magical helper, but a powerful tool – provided you direct it well. And that starts with your prompt. With this three-part guide, you will master working with AI for communications:
1. Personalise → 2. Produce → 3. Perfect.
AI changes communications, but communications changes AI right back. That is the core. Any communications professional who learns to think in roles, context, and goals – instead of just formats or text – gets more out of every tool. And that begins with the baseline attitude: not 'the machine makes something for me', but: 'I am directing a creative professional, in hyper-acceleration'.
Why good prompts are indispensable
A prompt is an instruction in text form with which you direct AI like ChatGPT. Think of it as a briefing for an ultra-critical freelancer: the better your input, the better the output. Communications departments that master this language can speed up repetitive tasks, raise quality, and arrive at sharp concepts faster. Prompting is therefore not a gimmick, but a strategic skill.
Thinking and producing faster and more strategically
The crux? Prompting forces you as a professional to ask better questions. Instead of automatically typing what you have always done, you have to pause and consider your target audience, the goal, the structure, the tone, the timing. That is the essence of communications. AI thus helps us not only produce faster, but also think more strategically. Prompt design is communication strategy in concentrated form.
Step 1: personalise – make the AI your own
Your first task: setting up the AI in such a way that it works for your reality. You feed the model with context, instructions, and nuance. The better you execute this phase, the more intelligent the output.
Compare it to a good briefing for a content writer. If you only provide a topic, you get mediocrity. If you provide a goal, target audience, tone of voice, structure, and examples? Then you get quality. It's no different with AI. Below are the 13 steps for a super prompt:
Role Put the AI into a specific role: ‘You are a senior communications consultant with a legal background.’ Or: ‘You are a copywriter at Coolblue’. The more concrete the role, the better the perspective.
Goal What do you want to achieve? Inform? Persuade? Activate? State it explicitly. Without a goal, the AI doesn't know what to work towards.
Target audience Be specific. Not ‘young people’, but: ‘female, 17-22 years old, vocational level, active on TikTok and responsive to humour and visual language.’ That’s how you get content that truly resonates.
Context AI does not know your organisation, industry, or project. So explain who you are, what you do, what challenges are at play, and what the project's goal is.
Task Describe what you expect from the AI. Be precise. For example: ‘Write a clear summary of this report of max. 300 words, at B1 language level, with bullet points.’
Format Indicate which format you need: blog post, press release, carousel, script, etc. Explain which elements you want to see included: headline, intro, subheadings, CTA, etc.
Style Describe the tone of voice, writing style, brand values, and desired form of address. Do you want it down-to-earth, inspiring, activating, or informative? Provide examples.
Examples Input texts that already work well. AI learns from style examples. This increases the chances of consistency and quality.
Data Input market figures, target audience data, competitor analyses, and trends. The AI uses this to generate timely, relevant, and credible content.
Constraints Specify restrictions: max number of words, words that must be avoided, sensitive topics. This helps prevent blunders.
[Brackets] Use square brackets for info that serves only as context and is not intended for publication. For example: [target audience: healthcare professionals in elderly care]
Sources Explicitly ask for source attribution for facts or claims. Indicate which sources you consider reliable (CBS, RIVM, SCP, etc.). Check everything.
Question Ask at the end: Do you understand the assignment? Do you have any questions? Can you improve this prompt? That will take your prompt from good to excellent.
Pro tip:
Write your prompt in Word first. Prevent accidentally hitting Enter. And: save your prompt, this way you build a reusable library.
Bonus:
Work with layers. First an instruction about tone of voice, then a separate prompt about content. Split it up. That gives you finer control.
Step 2: produce – start the real work
Once you have set up Chat properly, this is the easiest step. You've given the briefing, now you explicitly ask for the first output. Start your sentence with a verb: ‘Write…’, ‘Generate…’, ‘Create an outline…’.
Examples:
‘Write a LinkedIn post about our new strategy, max. 1,200 characters, with a call to action.’
‘Create an outline for a whitepaper on water safety for municipal policymakers.’
‘Generate five angles for a campaign against food waste among students.’
Be clear, pay attention to timing
The AI can do a lot, but only if you ask it.
Also pay attention to the timing. When in the process do you deploy AI? Not just at the end for a text, but also at the beginning for an angle. Or halfway through, for counterarguments. Think iteratively, not linearly. That leads to more creative content and better choices.
Step 3: perfect – optimise your output
The AI has produced. What next? That's when the real work begins. Because a first draft is rarely the best. Time for feedback, deepening, and evaluation.
Concept variations Ask Chat for alternatives: ‘Give me three variants with a different angle.’ Or: ‘Can you rewrite this text for a more critical audience?’ That prevents tunnel vision.
Adding layers Give targeted feedback. For example: ‘Paragraph 2 is too vague. Rewrite it, add an example, and keep it under 50 words.’ Or: ‘Make this more active and shorter, in the style of De Correspondent.’
Testing alternatives Let Chat spar: ‘What are the strengths and weaknesses of this intro?’ Or: ‘Which of these three tones of voice best fits the target audience?’ AI thus becomes your co-editor.
No shortcuts Don't rewrite it yourself, no matter how tempting. Let AI do it. That is the only way you learn to steer, fine-tune, and push boundaries. Those who skip this step remain stuck with mediocre output.
Evaluation Use the AI as a copy editor. Ask: ‘Check for passive sentences, inconsistencies, and spelling errors.’ Or: ‘Assess readability at B1 level, provide suggestions for improvement.’
Reliability analysis Ask for a star rating system: 0 stars = unsubstantiated, 5 stars = fully verified. That's how you keep a grip on facts.
Audience test Ask: ‘How would this text come across to an 18-year-old vocational student?’ or: ‘What would an alderman for healthcare think of this?’ You use AI as a simulation engine for a real audience. And that often delivers surprising insights.
The magic recipe summarised
You get the best AI output with this triptych:
Personalise: give Chat the right context and instructions
Produce: ask clearly and concretely for the content
Perfect: refine, improve, deepen
This is not a luxury, but a necessity. AI only does what you want once you know what you want.
These three steps are not a one-off action, but a learning process. The more often you go through this cycle, the sharper you become at prompt design. And the more you discover: what works for your brand, your target audience, your communication challenge?
Data & AI: the strategic side
Do you want to use AI not just for content, but also for strategy? Then use it for:
Sentiment analysis in media or social channels
Discovering blind spots in stakeholder thinking
Creating target audience segmentations
Risk analyses for reputation issues
Content gap analysis compared to competitors
Spotting new trends in search behaviour
This way, AI becomes not only your editor, but also your advisor.
Want to know more? We'd love to think along with you.
