Forty pages. Familiar title. New date. A filename that ends in “FINAL_v6”.

Your director asks: “Does this change anything for us?”

You could read it from scratch or ask AI for a summary. But a summary of the new version may sound almost identical to a summary of the old one, even when a crucial condition has disappeared.

Try giving AI a more specific job: compare the versions, locate the changes and help you assess their significance.

Here’s a practical workflow for policy drafts, consultation responses, position papers and coalition statements.

1. Set up the comparison properly

Give an approved AI tool the two documents, clearly labelled:

• OLD: title, date and version.

• NEW: title, date and version.

Only upload material you’re authorised to share. For long or complex documents, work section by section so you can check the output more easily.

Start with this:

Compare OLD and NEW. Identify additions, deletions and changes in wording. For each change, quote the exact old and new text and provide section references. Separate substantive changes from editorial edits. If text has moved, identify it as moved rather than deleted and added. Flag anything you cannot reliably match.

Ask for a table with four columns: location, old wording, new wording, type of change.

At this stage, keep the task factual. Establish what changed before asking what it means.

2. Look for the edits that are easy to miss

A rewritten introduction attracts attention. A deleted qualification can slip past you.

Ask AI to prioritise changes involving:

• Scope: who or what is covered.

• Responsibilities: who is expected to act.

• Timing: deadlines, transition periods and review dates.

• Conditions: exceptions, thresholds and eligibility.

• Commitments: promises strengthened, softened or removed.

Small words deserve scrutiny: “all”, “some”, “only”, “unless”, “may”.

Consider this fictional coalition statement:

Old wording:

We support the proposed timetable, provided smaller organisations receive a two-year transition period.

New wording:

We support the proposed timetable.

Shorter. Cleaner. But the condition attached to your support has vanished.

A general summary might describe both versions as “support for the proposed timetable”. A careful comparison should surface the deleted condition.

That is exactly the sort of edit you want flagged before approval.

3. Connect the changes to your priorities

A list of edits still leaves you with the main question: which ones matter to us?

Where permitted, provide your organisation’s agreed position or a short list of priorities. Be specific:

• We support the overall objective.

• We need a workable transition period.

• We oppose extending the scope to smaller operators.

• We have not agreed a position on reporting frequency.

Then ask:

Compare the identified changes with our priorities below. Which changes could affect our position, previous recommendations or planned communications? Explain each connection using the supplied documents. Separate confirmed textual changes from possible implications. Do not infer why an author made a change.

That last instruction matters. A deleted sentence is observable, unlike a claim that someone deleted it “to weaken our negotiating position” is speculation.

Useful follow-up questions include:

• Has one of our requests been reflected in the new wording?

• Does our briefing now refer to something that has changed?

• Has a compromise we accepted disappeared?

• Does this require internal agreement before we respond?

4. Verify the consequential changes

Treat the AI comparison as a review aid. It can miss deletions, misread tables or present moved text as new.

Before circulating the findings:

• Check quoted passages against both originals.

• Confirm that you compared the correct versions.

• Inspect footnotes, annexes and definitions where relevant.

• Use a conventional document comparison tool alongside AI where available.

• Ask the responsible colleague to assess implications requiring specialist judgment.

Be especially careful with “nothing significant changed.” That conclusion needs checking too.

5. Send a decision update

Your colleagues probably don’t need the entire comparison table in their inbox.

Lead with five bullets:

• What changed

• Where it changed

• Why it matters to us

• What remains unclear

• What we recommend doing next

For the fictional example above, the update could read:

The revised statement removes the condition attached to our support for the timetable. The transition-period wording has disappeared from paragraph 4. This conflicts with our agreed position on smaller organisations. We should ask whether the deletion was intentional and seek internal approval before endorsing the revised text.

That gives someone a clear issue to resolve. Keep the detailed comparison available underneath or as an attachment.

6. Pull the insights so you don’t need to do it again

When all is done, feed in the original, intermediary and final version of the document. Tell AI to ‘compare them, with the final version in mind, and generate a list of actions that could be applied to the very first version so it becomes like the final one based on specific and more generic changes next time a similar document comes in’. This way you don’t need to go through the rounds of revision and get straight to a result you will most likely appreciate.

Try it on a document you already know

Take two versions of a recent paper you have reviewed manually.

Run the comparison. Check:

• What did AI catch?

• What did it miss?

• Did it distinguish editorial changes from substantive ones?

• Could you trace every finding back to the text?

Start there before relying on the workflow under pressure.

At Influence Builders, we help public affairs and communication teams use AI to spot meaningful changes, assess their implications and turn complex information into clear next steps.

The next time “updated version attached” arrives, aim to answer one question: what do we need to reconsider?

It’s accurate, legal has no objections, everyone in the meeting understood it.

That doesn’t mean the audience will.

People skim, journalists shorten, opponents reframe.

Colleagues read words through their own priorities. A sentence that sounded perfectly reasonable in the drafting room can turn confusing, bland or unexpectedly provocative once it actually leaves the building.

Before you publish your next important message, spend 15 seconds trying to misunderstand it.

On purpose.

Start with the actual message

Use the final wording, not a summary of what you meant to say.

It could be:

• A campaign headline

• A policy announcement

• A LinkedIn post

• A press statement

• An executive quote

• A speech paragraph

• A position paper summary

• An email to members or partners

Paste it into an approved AI tool and add:

Do not rewrite this yet. Test how different audiences might (mis)interpret it. Separate reasonable interpretations from far-fetched ones.

Then run through the five tests below.

1. The five-second test

Ask:

Read this as someone scrolling quickly. After five seconds, what do you think the message is about? What do you think we want you to believe or do?

If the answer misses your main point, that’s probably not an attention span problem.

Your message might just be asking people to work too hard for it.

Now ask:

Which words carry the main meaning? Which words could disappear without changing anything?

This is a fairly ruthless way to find the decorative language.

If half the sentence could go and nothing would be lost, cut it.

2. The headline test

Ask:

Turn this message into five possible headlines:

• A neutral news headline

• A supportive headline

• A sceptical headline

• A hostile headline

• A misleading but plausible headline

Do not invent new facts.

The hostile headline is useful. The misleading one is often even more useful.

It shows you which part of your wording could be lifted out of context and made to carry a meaning you never intended.

You might decide the risk is acceptable. Fine, but at least you’ll have seen it before publication rather than after.

3. The “what does this mean for me?” test

A message can be perfectly clear and still fail because it gives the audience no reason to care.

Ask:

How might each of these audiences answer the question “What does this mean for me?”

• A policymaker

• A journalist

• An affected citizen or consumer

• An industry representative

• An NGO or campaigner

• An employee inside the organisation

Don’t expect every audience to respond positively.

Look instead for the blank spaces. If the AI can’t find any consequence for a priority audience, your message may be describing your organisation more than it’s communicating with anyone outside it.

4. The jargon alarm

Ask:

Identify every word or phrase that could be understood differently by specialists and non-specialists. Explain each one in plain language without making it childish.

Watch for phrases like:

• Strategic framework

• Stakeholder engagement

• Evidence-based

• Innovative solution

• Enabling environment

• Resilience

• Sustainable transition

• Best practice

These aren’t automatically wrong. Sometimes they’re the correct technical terms for the job.

But if three different people could walk away with three different meanings, the phrase is doing less work than you think it is.

Try swapping one abstract phrase for a concrete consequence.

Not “supporting a sustainable transition.”

What changes. For whom. By when.

5. The quote-without-context test

Finally, ask:

Select the three sentences most likely to be quoted on their own. For each one:

1. Explain how it could be interpreted without the surrounding text.

2. Identify what important context would disappear.

3. Suggest whether the sentence should be clarified, kept or removed.

This is especially useful for speeches, interviews, press releases and social posts.

The audience doesn’t control which sentence travels furthest. Neither do you, really.

Now make three decisions

Don’t ask the AI to rewrite everything.

Instead, build a simple table:

Keep | Clarify | Remove | Specific wording | Ambiguous or overloaded wording | Empty or distracting wording

Then revise the message yourself.

Once you have a new draft, go back and only re-run the tests it previously failed. There’s no prize for generating twenty-seven alternatives.

A useful final prompt

Before sign-off, ask:

Compare the original and revised messages. Has the revision become clearer without becoming less accurate? Has it become simpler without hiding an important qualification? Point out any trade-off I should review.

That last question matters more than it sounds.

Communicators are usually told to simplify. Public affairs people are usually told to preserve nuance. Good messaging has to do both at once, which is annoying but true.

AI can help you locate where those two pull against each other. A human still has to decide where the line actually sits.

The rule to remember

Don’t just ask “is this message clear?”

Ask “how could this message be misunderstood?”

Clarity isn’t what the drafting team meant. It’s what the audience can reasonably take away.

At Influence Builders, we help public affairs and communication teams use AI to test messages, sharpen arguments and prepare for real-world reactions.

Because the best time to find a second meaning is before somebody else publishes it for you.

Useful, sure. But it leaves one fairly large gap.

The other person hasn’t read your script.

They might misunderstand the proposal, challenge an assumption you thought was obvious, or ask the one question nobody prepared for. They might also just be distracted, sceptical, or interested in a completely different part of the issue than you are.

Before your next meeting, use AI to rehearse the room instead of the presentation.

Here’s a practical 5-minute exercise.

Step 1: Give AI the minimum necessary context

Start with a short factual briefing:

• Who you’re meeting

• Their role and organisation

• The purpose of the meeting

• What you want them to understand, believe or do

• What they may already know about the issue

• Their likely interests or concerns

• Any known points of disagreement

• What must not be assumed

Don’t upload confidential information or personal data. Stick to public information and your organisation’s approved internal material.

Then add:

Do not invent facts about this person or organisation. Separate publicly supported information from assumptions. Label every assumption clearly.

That instruction won’t make the output reliable on its own, but it does make the weak spots easier to spot.

Step 2: Ask AI to build the stakeholder’s mental checklist

Use this prompt:

Imagine you are this stakeholder preparing for our meeting. Based only on the information provided, list:

1. What you are likely to care about

2. What you may be suspicious of

3. What evidence you would expect

4. What would make our proposal difficult to support

5. What might make the meeting feel useful to you

Mark each point as either supported by evidence or an assumption requiring human judgement.

Review the response with the team.

Delete anything invented. Correct anything politically naïve. Keep the questions that make you uncomfortable.

Those tend to be the useful ones.

Step 3: Run the hostile version of the meeting

Now ask AI to become a difficult stakeholder:

Conduct a simulated meeting with me. You are sceptical, short on time and unwilling to accept vague answers. Ask one question at a time. Challenge unsupported claims, unclear language and unrealistic requests. Do not help me improve my answer until the simulation is over.

Answer the questions the way you would in the real meeting.

Don’t spend five minutes crafting each reply. Speak or type quickly. The point is to see how well you actually understand the case when you can’t hide behind a carefully edited briefing.

After six to eight questions, ask:

Where did my answers become vague, defensive, overly technical or unconvincing? Which question did I fail to answer directly?

This is often more useful than another round of slide editing.

Step 4: Change the stakeholder’s mood

A difficult meeting isn’t always openly confrontational. Sometimes the harder situation is a polite stakeholder who nods, thanks you, and does nothing afterwards.

Run the exercise again with a different instruction:

You are courteous but unconvinced. You will not openly disagree. Ask questions that reveal whether this proposal deserves your time, political capital or organisational resources.

This version tests whether your argument gives someone a reason to act, not just a reason to agree.

You can also try:

You broadly support our objective but think our proposed solution is unrealistic.

Or:

You are interested in the issue, but it is not among your current priorities.

The same talking points won’t land the same way in all three rooms.

Step 5: Practise the questions you hope nobody asks

Ask each team member to write down one question they’d rather not receive.

Put them into the AI simulation without discussing them first.

Then ask:

Present these questions in an unpredictable order. After each answer, ask one follow-up question based on what I said. Do not move on if I avoid the substance.

This matters because difficult questions rarely arrive in the wording we prepared for. A stakeholder listens to the answer and follows the thread wherever it leads.

Your rehearsal should do the same.

Step 6: Finish with a one-page meeting card

After the simulations, put together a short meeting card containing:

• Our objective

• The stakeholder’s likely objective

• Our main request

• Three points we must land

• Two claims requiring strong evidence

• Three difficult questions

• Our shortest credible answers

• One point on which we can be flexible

• One point we should not concede

• The next step we want agreed before leaving

If the card runs to three pages, start cutting.

Nobody needs another briefing document to read in the lift. What they need is a clear reminder of what matters once the conversation drifts away from the plan.

One important warning

AI can’t tell you what a real person thinks.

It can help you explore plausible reactions, expose gaps and make the rehearsal less comfortable. That’s valuable. But a synthetic stakeholder profile should never be presented as intelligence about an actual individual.

Treat the simulation as a set of hypotheses to test, not a prediction.

Used properly, AI doesn’t make the meeting less human. It gives the humans in the room more room to listen, adapt and respond without clinging to a script.

At Influence Builders, we help public affairs and communication teams turn AI into practical working methods for research, strategy, messaging and stakeholder engagement.

The objective isn’t to sound perfectly prepared.

It’s to stay useful once the meeting stops going according to plan.

Plausible writing travels fast inside an organisation. Headings are tidy, the tone sounds confident, the recommendations read as sensible, so nobody feels much urge to poke at it.

And yet a briefing can read perfectly well and still be politically wrong.

Maybe it treats a public statement as a firm commitment. Maybe it mixes up institutional weight with actual political influence. Maybe it points you toward someone who’s technically relevant but, in practice, on the sidelines. Or maybe it lands on that particular flavour of compromise language that upsets nobody because it doesn’t really say anything.

None of this is an argument against using AI. It’s an argument for giving the draft a hostile audience before a real one gets it.

The 2-minute red-team test

Once you’ve got a working draft, open a fresh AI conversation. Don’t ask it to polish the wording: ask it to go after the substance from four different angles.

1. The policymaker with no time

Hand the AI the draft and ask something like:

You’re a policymaker reading this between meetings. What is the author actually asking me to do? What’s unclear, irrelevant, or too hard to act on? Which claims would you want evidence for?

This one catches a problem that shows up more often than you’d think: a paper that spends pages explaining the issue and never quite states the ask.

If the AI can’t summarise the ask in a single sentence, chances are a busy human won’t find it either.

2. The sceptical opponent

Change the role:

You strongly disagree with this position. Point out the three weakest assumptions, the evidence you’d challenge, and any language you could paint as self-serving or misleading.

Don’t ask for balance here. Let it be difficult on purpose.

Most teams check whether their argument is accurate. Fewer check how easily someone with opposing interests could twist it. Those really aren’t the same test.

3. The journalist hunting for a story

Then:

You’re a specialist journalist. What headline would you put on this proposal? Which line would you quote? What obvious question has the author dodged?

This is usually where the euphemisms and the accidental subtext show up. If the imagined headline makes you wince a little, that’s worth sitting with.

The model might be misreading you. So might a journalist, an NGO, or a political opponent, and probably in the same way.

4. The insider who’s seen it all before

Last one. Ask the AI to play someone who actually knows the institution or the process:

Which parts of this briefing look procedurally naïve, politically unrealistic, or just out of date? Separate what you know from what you’re guessing, and flag anything that needs a human to verify it.

That last instruction matters more than it looks. Left to its own devices, AI tends to blur fact and inference. You have to force the line to show up.

This step doesn’t replace an experienced colleague’s read. It just means that colleague’s review goes faster.

Instead of dropping a polished document on their desk and asking “thoughts?”, you can hand them a list of the assumptions already flagged as shaky.

One rule worth keeping

Use AI to create friction before you use it to create polish.

Most teams do it backwards. They ask for tighter structure, smoother prose, a sharper close. The document gets prettier while the shaky assumptions underneath stay exactly as shaky as they were.

A red-team pass like this takes a few minutes. That’s not much to ask before something goes to a client, a senior executive, a policymaker, or a coalition partner.

It also means the quality of the review doesn’t depend on whether the right sceptical colleague happens to be free that afternoon.

The difference is rarely one brilliant prompt. It’s the review process built around it.

p.s.: at Influence Builders, we help public affairs and communication teams turn AI from a convenient writing tool into a reliable part of the workflow.

Unfortunately, “beautifully written” and “reliable” are not the same thing.

The problem is not always a spectacular hallucination since those are easy to catch. The more dangerous errors are less obvious:

• assumptions presented fact;

• interpretations presented as consensus;

• old positions presented as current;

• plausible predictions presented as the likely outcome.

While everything sounds confident and it all fits neatly together, nobody can immediately see which sentence is doing what.

Before circulating an AI-assisted brief, run this simple four-label check.

The four labels

Ask AI to classify every substantive claim as one of the following:

F: Fact
Directly supported by a reliable source.

I: Inference
A reasonable conclusion drawn from available evidence, but not explicitly confirmed.

A: Assumption
Something treated as true so the analysis can proceed.

U: Unknown
A gap that cannot currently be resolved with the available information.

This sounds almost suspiciously simple. That is precisely why it works.

A practical example

Suppose your draft says:

The Commission is likely to publish the proposal in October, with strong support from several Member States. Parliament will probably seek stricter provisions.

There may be three different levels of certainty hiding in those two sentences:

• The October timing may come from the Commission’s planning calendar: Fact

• Member State support may be inferred from recent Council discussions: Inference

• Parliament’s expected position may be based on previous files rather than current evidence: Assumption

Without labels, they blend into one authoritative-sounding paragraph.

With labels, you know exactly what requires verification, qualification or further research.

The 10-minute workflow

1. Paste in the draft

Use the full briefing note, stakeholder update, policy summary or internal memo.

2. Ask for claim-level classification

Use this prompt:

Review the text below and identify every substantive claim. Classify each claim as:

F: fact directly supported by evidence
I: inference drawn from evidence
A: assumption being treated as true
U: unknown or unresolved

Present the results in a table with five columns:

1. Claim

2. Classification

3. Evidence currently available

4. What could make the claim unreliable

5. Recommended action

Do not verify the claims yourself unless reliable sources have been provided. If evidence is missing, say so clearly.

The final sentence matters. Otherwise, AI may try to fill your evidence gaps with yet more confidence.

3. Focus on the risky claims

Do not spend equal time on everything.

Prioritise claims that affect:

• your recommended position;

• your assessment of political support;

• the expected timeline;

• stakeholder motivations;

• the probability of a particular outcome;

• advice given to senior management or clients.

An incorrect background detail is inconvenient. An incorrect assumption behind your strategy is expensive.

4. Decide what to do with each label

For every claim:

• Fact: retain it and cite the source.

• Inference: explain the reasoning and use appropriately cautious language.

• Assumption: verify it, qualify it or state it explicitly.

• Unknown: investigate it or acknowledge the uncertainty.

“Inference” is not another word for “wrong.” Public affairs work depends on interpretation. The point is to distinguish interpretation from confirmed information.

5. Rewrite for >honest< confidence

The goal is not to turn every sentence into nervous legalese.

Compare:

Parliament will oppose the proposal.

With:

Early reactions from the two largest political groups suggest that the proposal may face resistance in Parliament, although formal positions have not yet been adopted.

The second sentence is not weaker. It is more useful because it shows what we know, what we infer and what remains open.

Where this works especially well

Use the four-label check for:

• legislative monitoring updates;

• stakeholder maps;

• political intelligence reports;

• meeting briefings;

• scenario assessments;

• advocacy recommendations;

• summaries of consultations or position papers;

• internal notes written after informal conversations.

It is particularly valuable when several people have contributed to a document and nobody remembers where every claim originated.

One important rule

Do not ask AI, “Is this accurate?”

That usually produces a reassuring answer rather than a useful audit.

Ask it to separate the claims, classify their status and expose the missing evidence. Accuracy improves when uncertainty becomes visible.

AI is very good at removing hesitation from language.

Your job is to make sure it has not also removed uncertainty from the analysis.

Before you circulate the next brief, label the claims. Facts, inferences, assumptions and unknowns should never be allowed to wear the same suit.

P.S. At Influence Builders, we help public affairs and communications teams build practical AI workflows that improve both speed and quality, without losing professional judgement.

That is useful… but it’s all too linear.

A more valuable exercise is to ask AI to challenge your meeting before you walk into the room.

Here is a practical 20-minute workflow you can use before your next meeting with a policymaker, regulator, association, partner or difficult internal stakeholder.

Step 1: Give AI the necessary context

Start with:

I am preparing for a meeting with [person/organisation].

The purpose of the meeting is [objective].

Our position is [brief explanation].

Their likely priorities are [priorities].

The broader political or policy context is [context].

The outcome I would ideally like is [desired outcome].

Do not draft anything yet. First, identify the information you would need to properly challenge my approach.

Answer the questions it gives you. If you do not know something, say so. Those gaps are already useful preparation.

Step 2: Run the pre-mortem

Now ask:

Imagine the meeting has taken place and achieved nothing. The stakeholder remained unconvinced, no follow-up was agreed, and our position did not advance.

Identify the five most likely reasons this happened.

For each one, explain:

1. What we may have misunderstood

2. What the stakeholder may have been thinking

3. What warning sign could appear during the meeting

4. How we could prevent or respond to it

This forces you to think beyond your own message and consider the meeting from the other side of the table.

Step 3: Find your weakest assumptions

Ask:

List the assumptions behind our meeting strategy.

Separate them into:

• supported by evidence
• plausible but unconfirmed
• risky assumptions

For each risky assumption, suggest one question I could ask before or during the meeting to test it.

You may discover that your strategy depends on things you have never actually verified.

For example:

• We assume the stakeholder already understands the issue.

• We assume they have influence over the decision.

• We assume they see the problem as urgent.

• We assume our proposed solution fits their political constraints.

• We assume they are willing to act publicly.

Any one of these could derail the conversation.

Step 4: Stress-test your arguments

Paste your three main arguments and ask:

Act as a sceptical but well-informed stakeholder. Challenge each argument from political, practical, budgetary and reputational perspectives.

Do not invent facts. Flag any point that would require external verification.

Then suggest how I could strengthen each argument without making it longer.

The final instruction matters. In stakeholder meetings, stronger rarely means more slides.

Step 5: Prepare for the actual conversation

Finally, ask AI to produce a one-page meeting sheet containing:

• The meeting objective in one sentence

• The stakeholder’s likely interests

• Our three strongest points

• Three questions we need to ask

• The hardest objection we may hear

• A concise response to that objection

• Signals that the meeting is going well

• Signals that we should change approach

• The minimum acceptable outcome

• The ideal follow-up commitment

Keep this page open during the meeting. Do not turn it into a script.

One important caution

AI can help you test your thinking, but it does not know what the stakeholder privately believes. Treat its answers as hypotheses, not intelligence.

Verify factual claims. Remove confidential information. Use your own judgement.

The real value of this exercise is not that AI predicts exactly what will happen.

It is that you arrive having considered more than one version of the meeting.

And that is often the difference between delivering your message and actually advancing your objective.

p.s.: if you’re looking for AI workshops, get in touch.

That’s useful.

But there’s a much more valuable habit that almost nobody follows.

Instead of asking AI to summarise information, ask it to compare it.

Imagine you’ve just collected:

• a European Commission proposal

• European Parliament amendments

• Council of Ministers position

• your organisation’s position paper

• 3 stakeholder statements

Most people ask:

“Summarise these documents.”

You’ll receive five summaries.

Useful.

But not particularly strategic.

Instead, try prompts like these:

Compare the positions of all parties in a table.

Identify where there is broad agreement and where the biggest disagreements remain.

Which arguments appear most often across stakeholders?

Which concerns are raised by only one actor?

If I were preparing a meeting with the rapporteur, what questions should I ask based on these differences?

Use bullet points and executive style.

Suddenly AI is no longer just reducing reading time.

It’s helping you see patterns.

And that’s where much of the value in public affairs lies.

A simple workflow

Instead of this:

Read → Summarise → Write

Try this:

Collect → Compare → Identify patterns → Decide → Draft

The comparison step often reveals things you would otherwise miss:

• unexpected alliances

• conflicting priorities

• recurring arguments

• missing evidence

• opportunities to build compromise

Try it this week

Take one legislative file you’re currently working on.

Upload three or four related documents.

Don’t ask for summaries.

Ask AI to answer questions such as:

• Where do these documents agree?

• Where do they contradict each other?

• What assumptions does each actor make?

• What evidence is missing?

• What would each stakeholder probably object to?

You’ll probably spend less time reading.

But more importantly, you’ll spend more time thinking.

And that’s where public affairs professionals create value.

p.s.: if you’re looking for AI workshops for your team, get in touch.

Every week we’d come across a new “must-have prompt” on LinkedIn or in a newsletter.

Some promised “perfect” reports.

Others claimed to unlock AI’s hidden potential. Like everyone else, we saved them.

Today, our prompt library is actually smaller.

Not because we use AI less. Quite the opposite.

We’ve simply become much more selective about the prompts we keep coming back to.

Here are five that quietly disappeared from our workflow.

1. “Act as a(n) […] expert...”

We used to write prompts like:

Act as a senior public affairs consultant with 20 years of experience...

Sometimes the output was fine.

More often, it sounded like someone trying very hard to sound like a senior public affairs consultant.

These days, we rarely bother assigning AI a role.

Instead, we explain the situation.

Who is going to read the document? What decision are they trying to make? How much time do they have? What do they already know?

We’ve found that good context beats impressive job titles almost every time.

2. “Make this more professional.”

This one almost always made us smile for the wrong reasons.

“Professional” often turned into longer sentences, more jargon and a lot of words that nobody would actually say in a meeting.

Now we’re much more specific.

We might ask AI to make the structure clearer, remove repetition, cut the length by a third or point out where the argument becomes weak.

Those are changes we can actually judge.

“More professional” means something different to everyone.

3. “What are your thoughts?”

This is another prompt we barely use anymore.

The answers tend to be broad, polite and full of observations that don’t really move the work forward.

Instead, we ask questions that force AI to take a position.

For example:

If you disagreed with this briefing, what would your strongest argument be?

Or:

Which recommendation would you remove if this had to fit on one page?

Those answers are usually much more interesting.

4. “Give us more ideas.”

We’ve realised that lack of ideas is almost never the problem.

Choosing between them is.

So instead of asking AI to generate another twenty suggestions, we’ll often ask:

Which two would you reject, and why?

Or:

If you had to defend just one of these options in front of a client, which would you choose?

That tends to produce much sharper discussions.

5. “Is this good?”

AI is generally very encouraging.

Almost everything is “clear”, “well structured” or “effective”.

That’s nice.

It’s not particularly useful.

We’ve had much better results asking things like:

Which sentence sounds generic?

Where would an experienced reader lose interest?

What’s missing that someone familiar with this topic would expect to see?

Those questions lead to much more honest feedback.

The Bigger Lesson

Looking back, we noticed something interesting.

We didn’t stop using these prompts because AI changed.

We stopped using them because we changed.

The more experience we gained, the less interested we became in clever prompts and the more interested we became in asking precise questions.

That’s probably the biggest shift we’ve seen over the past year.

Good prompting isn’t about sounding clever.

It’s about being clear about what you’re actually trying to achieve.

So here’s a question for you.

Which prompt have you quietly stopped using because it simply wasn’t worth it anymore?

It isn’t.

The quality gap between people getting mediocre results and people getting exceptional results increasingly comes down to something else:

the quality of the information they feed into the system.

The reality is that AI often knows less about your world than you might assume.

It doesn’t know:

• your organisation’s priorities

• your stakeholders

• your preferred writing style

• your previous positions

• your internal terminology

• your political sensitivities

• your strategic objectives

Which means every new conversation starts from scratch.

Many public affairs professionals respond by writing longer prompts.

A better approach is to build a simple personal intelligence system.

Step 1: Create Your Core Knowledge Base

Most of us already have one.

It’s just scattered across:

• PowerPoint presentations

• meeting notes

• strategy papers

• position papers

• stakeholder lists

• emails

• policy trackers

• previous submissions

Collect the most valuable materials in one place.

Not because AI will magically read everything.

But because you now have a structured source of truth you can continuously draw from.

Step 2: Capture What You Learn

Most valuable knowledge never makes it into official documents.

It emerges during:

• stakeholder meetings

• conferences

• political conversations

• client discussions

• internal debates

Unfortunately, this knowledge often disappears after a few days.

Instead, start capturing:

• emerging themes

• recurring concerns

• stakeholder priorities

• objections you hear repeatedly

• useful examples

The goal is not perfect documentation.

The goal is preserving insights before they vanish.

Step 3: Build Reusable Context Packs

Before asking AI to help with a task, provide context that rarely changes.

For example:

Organisation Context

Who are we?

What do we do?

What are our priorities?

Stakeholder Context

Who are we trying to influence?

What matters to them?

What are their likely concerns?

Style Context

How do we write?

What tone do we use?

What should be avoided?

Creating these once can save hundreds of explanations later.

Step 4: Feed AI Your Best Work

Many people show AI examples of average work.

Then they wonder why the output is average.

Instead:

• collect your strongest reports

• your best briefings

• your most successful submissions

• your strongest speeches

Use these as examples.

AI learns patterns remarkably well when shown quality.

Step 5: Update the System

Policy environments change.

Stakeholders change.

Organisations change.

Your intelligence system should evolve as well.

A simple monthly review is often enough:

• What have we learned?

• What assumptions changed?

• What new evidence emerged?

• What should be added?

The best AI users are increasingly behaving less like prompt engineers and more like knowledge managers.

They spend less time crafting clever prompts.

They spend more time building high-quality inputs.

And that is where the real competitive advantage is emerging.

Not in asking better questions.

In creating better context.

p.s.: if you need to train your team, get in touch

It is the lack of time.

A new European Commission initiative is published. A consultation opens. A Parliament committee adopts a report.

The information is there.

The challenge is turning it into something useful for different audiences.

Most professionals still do this manually. They read the document, write a summary, prepare talking points, draft an email, update colleagues, and perhaps prepare a stakeholder briefing.

The process works.

But it is slow.

AI can help, not by replacing expertise, but by helping you repurpose your analysis more efficiently.

Here is a simple workflow we increasingly see professionals use successfully.

Step 1: Create the master summary

Start with the source document.

Ask AI:

“Summarise this document in 10 bullet points for a German public affairs professional. Focus on policy implications, timelines, decision-makers and potential stakeholder impact.”

Review the output and correct anything that requires expert judgement.

This becomes your master summary.

Step 2: Create an executive briefing

Now ask:

“Based on this summary, prepare a one-page briefing for a senior pharma executive. Focus on strategic implications, risks, opportunities and recommended actions.”

The information is the same.

The audience is different.

Step 3: Create internal talking points

Next ask:

“Turn this into five talking points for colleagues who are not policy experts.”

This helps align teams across communications, commercial, regulatory or management functions.

Step 4: Prepare stakeholder engagement notes

Then ask:

“Which stakeholder groups are most affected by these developments? For each group, identify likely interests, concerns and engagement opportunities.”

This often produces useful starting points for stakeholder mapping and engagement planning.

Step 5: Prepare questions for the next meeting

Finally ask:

“Based on this development, what questions should we ask policymakers, industry associations and other stakeholders?”

This is one of the most underused AI applications.

Good questions often create more value than good answers.

The result

From one policy document, you now have:

• A policy summary

• An executive briefing

• Internal talking points

• Stakeholder engagement notes

• Meeting preparation questions

The entire process can take less than 30 minutes.

More importantly, it allows public affairs professionals to spend less time reformatting information and more time analysing what it means.

That is where expertise creates value.

AI is not replacing policy judgement.

It is creating more space for it.

How to write better prompts.

How to get better answers.

How to unlock the full potential of AI.

But after working with hundreds of public affairs, advocacy, and communications professionals over the past year, we have come to a different conclusion.

The most valuable AI skill is not prompting.

It is problem definition.

The Prompting Trap

Many professionals approach AI like a search engine with a larger text box.

They start typing immediately.

“Summarise this report.”

“Write a stakeholder map.”

“Draft a position paper.”

“Prepare talking points.”

Sometimes the results are surprisingly good.

Often they are disappointing.

The instinctive reaction is usually:

“I need a better prompt.”

Not necessarily.

In many cases, the real issue is that the user has not clearly defined the problem they are trying to solve.

AI cannot compensate for unclear thinking.

In fact, it often amplifies it.

The Difference Between Tasks and Problems

Consider these two requests.

Task-focused request:

“Create a stakeholder map for the proposed legislation.”

Problem-focused request:

“Our objective is to prevent amendments that would restrict market access during the Council negotiations. Which stakeholders are most likely to influence the national positions of Germany, France, and Poland, and where should engagement efforts be prioritised?”

Both requests concern stakeholder mapping.

Only one is connected to a real-world objective.

The second produces far more useful outputs because it gives AI a clear problem to work on.

The same principle applies across public affairs.

The quality of the answer depends heavily on the quality of the question.

Why This Matters More Than Ever

Most AI tools are becoming easier to use.

Prompting techniques that seemed advanced a year ago are increasingly built into the systems themselves.

Models are becoming better at interpreting vague instructions.

That means prompting is becoming less of a competitive advantage.

The ability to define the right problem is not.

The professionals who will benefit most from AI are not necessarily those who know the latest prompting tricks.

They are the people who can clearly articulate:

• What decision needs to be made

• What objective needs to be achieved

• What constraints exist

• What success looks like

• What information is actually needed

In other words, they think before they prompt.

A Simple Framework

Before opening any AI tool, ask five questions:

1. What decision am I trying to support?

2. What outcome am I trying to achieve?

3. Who will use this output?

4. What constraints matter?

5. What would make this output genuinely useful?

6. What does success look like?

Only then should you start prompting.

You may find that your first prompt changes completely.

More importantly, the quality of the answer improves dramatically.

The Real Opportunity

The public affairs profession has never suffered from a lack of information.

It suffers from a lack of clarity.

AI can help process information faster.

It can help draft, summarise, analyse, and brainstorm.

But it cannot decide what matters.

That remains a human responsibility.

The future belongs neither to those who reject AI nor to those who blindly embrace it.

It belongs to those who combine good judgment with powerful tools.

And that starts long before the prompt.

It starts with defining the problem.

(and if you need AI workshops for your team, get in touch)

Artificial intelligence is changing the way public affairs consultants work.

Research that once took hours can be completed in minutes. Policy developments can be monitored more efficiently. Stakeholder intelligence can be gathered faster. Drafts can be produced almost instantly.

Speed is not the most important change.

The real opportunity for public affairs consultants is to rethink how they create value for clients.

As AI becomes better at handling routine tasks, the consultants who will stand out are not necessarily those who produce more work.

They are the ones who make better judgements, ask better questions, identify better opportunities, and help clients make better decisions.

This session was designed specifically for consultants working in public affairs agencies who want to understand how AI can strengthen their advisory role, improve their workflows, and increase the impact they deliver to clients.

Moving beyond productivity

Most conversations about AI focus on efficiency.

How can we draft faster?

How can we summarise more documents?

How can we automate repetitive tasks?

These are useful questions.

But they only scratch the surface.

The most successful consultants are beginning to use AI for much more than productivity.

They are using it to:

• identify emerging policy risks and opportunities earlier;

• strengthen stakeholder and political analysis;

• challenge assumptions and test strategic thinking;

• improve the quality of client recommendations;

• create more structured and repeatable workflows;

• spend less time producing information and more time interpreting it.

In other words, they are using AI to amplify expertise rather than replace it.

The changing role of the public affairs consultant

Clients do not hire consultants simply to receive information.

They hire them to make sense of complexity.

That distinction matters.

AI can collect information.

AI can summarise documents.

AI can generate first drafts.

But it cannot fully understand a client’s political context, organisational realities, risk appetite, or long-term objectives.

The consultant’s role therefore becomes even more valuable.

The challenge is no longer finding information.

The challenge is knowing what matters.

This requires judgement, experience, contextual understanding, and strategic thinking.

AI can support these capabilities.

It cannot replace them.

Creating more value for clients

One of the central themes of this session is how consultants can move beyond measuring success through hours worked or outputs delivered.

AI encourages a different mindset.

Instead of asking:

“How can I produce more work?”

A better question might be:

“How can I create more value?”

This may mean improving the quality of advice.

It may mean identifying opportunities that others have missed.

It may mean spending more time on strategy and less time on administration.

Or it may mean redesigning weekly workflows so that routine tasks consume less attention and energy.

The firms that succeed in the coming years will be those that use AI to strengthen client outcomes, not simply increase activity.

Learning from real-world consultancy practice

The session draws on practical lessons from training and working with some of the leading public affairs consultancies.

Rather than focusing on theory, it explores concrete examples of how consultants are already integrating AI into their daily work.

What is working?

What is not?

Where does AI create value?

Where does human expertise remain essential?

And how can firms adopt AI responsibly without compromising quality, confidentiality, or professional judgement?

Watch the recording

If you work in a public affairs consultancy, this recording will help you think more strategically about the role of AI in your work.

It is particularly useful if you have already experimented with AI tools and want to move beyond simple prompting towards more structured and impactful use.

You can also use it as a starting point for discussion within your team:

• Which parts of our workflow create the most value for clients?

• Where could AI free up time for higher-value work?

• How can we improve the quality of our advice, not just the speed of delivery?

• What skills will define successful consultants in an AI-enabled profession?

AI will not replace public affairs consultants.

It will change what clients expect from them.

The consultants who learn to combine AI capabilities with strong judgement, political insight, and strategic thinking will be the ones who create the greatest value.

Watch the recording and explore how AI can help you become a more effective, impactful, and future-ready consultant.

The teams getting the most value are rarely the ones with the best tools.

And definitely not the ones generating the most content.

They are the ones redesigning workflows.

This sounds abstract until you look at what is actually happening inside many organisations today.

A typical knowledge workflow still looks something like this:

• information scattered across emails, Teams chats, PDFs, meeting notes and SharePoint folders

• multiple people manually checking the same sources

• senior staff repeatedly answering the same questions

• briefings recreated from scratch

• institutional knowledge trapped inside individuals

• nobody fully sure which version is the latest

• “quick drafts” turning into endless editing cycles

Then AI gets added on top.

Usually as:

• summarisation

• drafting

• brainstorming

• rewriting

Useful? Absolutely.

But often it accelerates the wrong system.

This is why many organisations feel simultaneously:

• excited about AI

• disappointed with outcomes

• overwhelmed by tools

• unclear on ROI

The issue is often not the model.

It is workflow design.

The organisations seeing real gains are usually doing a few specific things differently

Not necessarily sophisticated things.

Just structured ones.

1. They separate information gathering from judgement

This is one of the biggest shifts.

Many senior professionals still spend enormous amounts of time:

• collecting updates

• cleaning information

• formatting material

• preparing first drafts

• consolidating repetitive inputs

AI is exceptionally good at handling parts of this process.

Not the strategic judgement.

The preparation.

For example, instead of manually reviewing 20 different policy sources every morning, some teams now use structured AI monitoring workflows that:

• scan selected institutions, stakeholders and topics

• filter repetitive updates

• cluster developments by relevance

• produce digest-style briefings

• highlight what has materially changed

The gain is not “automation.”

It is cognitive bandwidth.

Senior staff spend less time assembling information and more time interpreting it.

That distinction matters.

2. They stop treating prompting as improvisation

A surprising amount of AI usage still looks like this:

“Can you summarise this?”
“Can you rewrite this?”
“Give me ideas.”

Repeated thousands of times.

But mature teams increasingly build reusable prompting systems around recurring workflows.

For example:

• stakeholder mapping

• consultation analysis

• board briefings

• event preparation

• speech drafting

• media monitoring

• policy comparison

• scenario planning

This creates consistency across teams.

It also reduces one of the biggest hidden inefficiencies in AI adoption:

Every employee reinventing workflows individually.

The organisations progressing fastest are quietly standardising internal AI processes long before formal governance catches up.

3. They use AI to challenge thinking, not just generate text

This is probably the most underrated use case.

Especially for experienced professionals.

The value is often not:
“Write something for me.”

It is:
“Stress-test my reasoning.”

Some of the most useful prompts are surprisingly simple:

• What assumptions are weak here?

• Which stakeholder perspective is missing?

• What would a hostile regulator challenge?

• Where is the argument too vague?

• What unintended consequences are ignored?

• Which parts sound persuasive but contain little substance?

• What evidence would strengthen this position?

Used properly, AI becomes less of a content machine and more of a structured thinking partner.

This is particularly useful in public affairs, advisory work and strategic communications, where weak logic often hides beneath polished language.

4. They reduce organisational memory loss

One of the biggest invisible costs inside organisations is repeated rediscovery.

Teams repeatedly:

• relearn old lessons

• recreate past analyses

• lose stakeholder context

• forget earlier positioning

• duplicate research

• restart work from zero when people leave

AI will not magically solve knowledge management.

But it can significantly reduce friction around institutional memory.

Some organisations are beginning to create structured internal systems that retain:

• client preferences

• stakeholder sensitivities

• recurring policy interests

• previous strategic positions

• historical project context

• approved language patterns

• recurring analytical frameworks

This creates continuity that many organisations currently lack.

Especially fast-moving teams where institutional knowledge is fragmented across individuals.

The uncomfortable reality

Many organisations currently have workflows designed for a pre-AI environment.

That environment assumed:

• information scarcity

• slower production cycles

• high drafting costs

• limited analytical assistance

• fragmented access to expertise

AI changes those assumptions very quickly.

Which means the real question is often not:

“What AI tool should we buy?”

But:

“Which parts of our current way of working no longer make sense?”

Because simply adding AI on top of chaotic workflows often produces:

• faster confusion

• more content

• inconsistent quality

• duplicated effort at scale

The organisations creating real advantage are usually redesigning systems, not just adopting tools.

And interestingly, many of the most effective changes are operational rather than technological.

Clearer workflows.
Better information flows.
Reusable structures.
Defined review stages.
Smarter allocation of human judgement.

The technology matters.

But the operating model matters more.

It is usually won before the meeting starts.

Public affairs professionals know this. But in practice, meeting preparation often happens under time pressure:

• a calendar invite appears

• the topic is broad

• the stakeholder is important

• the internal team has different priorities

• someone needs a briefing “quickly”

• and the meeting is tomorrow.

This is where AI can help. Not by replacing your judgement, and not by inventing political intelligence. But by helping you prepare faster, more systematically, and with fewer blind spots.

Here is a practical 20-minute workflow you can use before a stakeholder meeting.

Before you start: gather the input

Do not begin with a vague prompt like:

Prepare me for a meeting with this stakeholder.

Instead, gather the basic material first.

You may need:

• the stakeholder’s public role and institution

• the meeting topic

• your objective

• your organisation’s position

• previous contact notes, if appropriate and safe to use

• recent public statements or policy positions

• the legislative or policy context

• any sensitive issues to avoid

Important: do not paste confidential or sensitive information into tools that are not approved for that type of content. If needed, anonymise or summarise the input first.

Now you can start.

Minute 1-3: clarify the meeting objective

Many meetings are weak because the objective is vague.

“Discuss the file” is not an objective.

“Introduce our concerns” is better, but still incomplete.

A stronger objective defines what you want the stakeholder to understand, feel, consider, support, question, or do after the meeting.

Prompt:

I am preparing for a stakeholder meeting on [topic] with [stakeholder / institution / role].
Our organisation’s position is [short summary].
Help me define 3 possible meeting objectives:

• a minimum objective

• a realistic objective

• an ambitious objective

For each, explain what success would look like.

This helps the team align before the meeting.

It also prevents the conversation from becoming a general exchange with no clear outcome.

Minute 4-6: map the stakeholder’s likely perspective

Next, use AI to think through the stakeholder’s possible priorities.

Prompt:

Based on the stakeholder’s role and the policy context, what are their likely priorities, constraints, sensitivities, and questions on this topic?
Organise the answer under:

• political priorities

• institutional constraints

• likely concerns

• possible incentives

• questions they may ask us

Do not invent private information. Base the analysis only on the role, public context, and reasonable assumptions.

This is not a substitute for real intelligence.

But it gives you a structured starting point.

It can also help junior colleagues prepare for the logic of the conversation, not just the content.

Minute 7-9: prepare your message hierarchy

A stakeholder meeting should not contain 15 messages.

It should have a clear hierarchy:

• one main message

• two or three supporting points

• evidence or examples

• a clear ask or next step

Prompt:

Help me build a message hierarchy for this meeting.
Include:

• one main message

• three supporting points

• one proof point or example for each

• one clear ask

• one softer fallback ask if the main ask is not appropriate

This is especially useful when several colleagues are attending the same meeting.

It gives everyone the same structure and reduces the risk of mixed messages.

Minute 10-12: anticipate difficult questions

Stakeholder meetings often go off-script.

That is not a problem if you are prepared.

Prompt:

Act as a sceptical but fair stakeholder. What are the 10 most difficult questions you could ask us in this meeting?
For each question, suggest a concise answer that is credible, non-defensive, and politically aware.

Then add:

Flag any questions where we should avoid answering too strongly without further evidence.

This is important.

AI can make answers sound more certain than they should be. You need to know where to be cautious.

Minute 13-15: identify risks and red lines

Some meetings carry reputational, political, legal, or relationship risks.

AI can help you identify them before the meeting.

Prompt:

Review this meeting context and identify possible risks:

• messages that could be misunderstood

• claims that may need evidence

• politically sensitive wording

• topics we should avoid

• areas where we should listen rather than push

• possible follow-up commitments we should not make too quickly

This is particularly useful for complex or sensitive files.

It can also help teams avoid overpromising in the room.

Minute 16-18: create the meeting brief

Now ask AI to turn the preparation into a concise briefing note.

Prompt:

Create a one-page meeting brief using this structure:

1. Meeting purpose

2. Stakeholder profile

3. Likely stakeholder perspective

4. Our main message

5. Supporting points

6. Difficult questions and suggested responses

7. Risks and sensitivities

8. Proposed ask

9. Desired follow-up

Keep it concise and practical. Use bullet points. Do not include generic background.

This gives you a usable internal document.

Not perfect. But usually much better than starting from a blank page.

Minute 19-20: prepare the follow-up before the meeting

This step is often forgotten.

But a good follow-up email is easier to write before the meeting, when the objective is still clear.

Prompt:

Draft a short follow-up email template for after this meeting.
Include:

• thanks

• a brief restatement of the discussion

• our key message

• any agreed next step

• a placeholder for documents or evidence to send

Keep it professional, concise, and easy to adapt.

You will still need to update it after the meeting.

But having the structure ready saves time and improves consistency.

The full 20-minute workflow

Here is the sequence in one place:

1. Clarify the meeting objective.

2. Map the stakeholder’s likely perspective.

3. Build your message hierarchy.

4. Anticipate difficult questions.

5. Identify risks and red lines.

6. Create the meeting brief.

7. Prepare the follow-up email template.

This can be done in 20 minutes if the input is ready.

For high-stakes meetings, of course, you should spend more time. But even then, the same workflow applies.

What AI should not do

AI should not decide your strategy for you.

It should not invent stakeholder intelligence.

It should not replace official sources, previous relationship knowledge, or internal alignment.

It should not be trusted blindly on political nuance.

And it should not be given confidential information unless your organisation’s tools and policies allow it.

The human role remains essential:

• checking facts

• adjusting tone

• adding political judgement

• understanding the relationship

• deciding what to say and what not to say

• reading the room

A practical rule

Use AI to prepare the structure.

Use human judgement to prepare the strategy.

That distinction matters.

AI can help you avoid blank-page preparation, generic messages, and missed objections.

But the quality of the meeting still depends on your understanding of the stakeholder, the policy context, and the relationship.

The bottom line

Stakeholder meetings are too important to prepare casually.

A simple AI-supported workflow can help public affairs teams become faster, clearer, and more consistent.

Not by automating the relationship.

But by improving the preparation behind it.

AI is no longer a side topic for communications teams.

It is becoming part of how we research, plan, write, test ideas, understand audiences, and make better decisions under pressure.

But for professionals working in policy communications, advocacy, public affairs, and strategic communications, the challenge is not simply using AI more.

The real challenge is using it well.

That means knowing when AI can help, when it can mislead you, how to protect sensitive information, and how to turn generic outputs into work that is accurate, persuasive, and useful.

It was designed specifically for communications professionals working in policy and advocacy contexts who want to improve the way they use AI tools at work — not just for faster drafting, but for better thinking, stronger messaging, and more confident decision-making.

Beyond basic prompting

Many professionals have already experimented with AI.

They have asked it to draft an email, summarise a document, rewrite a paragraph, or generate a few social media posts.

That is a useful starting point.

But the real value of AI for communications teams goes much further.

Used well, AI can help you:

• structure complex ideas more clearly;

• identify angles you may have missed;

• adapt messages for different audiences;

• test whether your argument is persuasive;

• prepare for difficult questions;

• analyse policy or advocacy materials faster;

• turn scattered information into usable communication assets.

The best communications teams are not using AI as a shortcut to avoid thinking.

They are using it as a thinking partner.

Why this matters for policy communications and advocacy

Communications in policy and advocacy environments are rarely simple.

You are often dealing with complex topics, multiple stakeholders, sensitive issues, fast-moving developments, and high expectations for accuracy.

A message that sounds good is not enough.

It needs to be credible.

It needs to be adapted to the audience.

It needs to reflect the political, institutional, or reputational context.

And it needs to hold up under scrutiny.

This is where AI can be genuinely useful, if it is used with judgement.

It can help you move faster, but it should not replace your expertise.

It can help you generate options, but it should not decide the strategy for you.

It can help you improve a draft, but it cannot know by itself what is politically sensitive, institutionally appropriate, or strategically wise.

That is why the focus of this session is practical: how to integrate AI into your communications work in a way that strengthens, rather than weakens, professional judgement.

Watch the recording

If you work in communications, public affairs, advocacy, policy, or stakeholder engagement, this recording will give you practical ideas you can start testing immediately.

It is especially useful if you already use AI occasionally, but want to move beyond basic prompts and start using it in a more structured, responsible, and strategic way.

You can also use it as a starting point for discussion within your team:
- Where could AI help us work better?
- Where do we need clearer rules?
- Which tasks should we automate, support, or leave firmly in human hands?

AI will not replace strong communications professionals.

But communications professionals who learn how to use AI well will have a serious advantage.

Watch the recording and start experimenting with a more practical, responsible, and strategic approach to AI in communications.

“Summarise this.”
“Write me a LinkedIn post.”
“Draft an email.”

And then they conclude: “AI is kind of OK… but very generic.”

The problem is not the tool: it’s the workflow.

The teams getting real value from AI are no longer treating it as a chatbot. They are treating it as a junior strategist, researcher, analyst, editor, and sparring partner, all at once.

Here are 5 practical ways public affairs and communications teams can immediately improve how they work with AI.

1. Stop Prompting Once. Build Iterations.

Most people ask AI one question and judge the output immediately.

Strong users iterate.

Instead of:

“Summarise the Energy Efficiency Directive.”

Try:

“Summarise the Energy Efficiency Directive from the perspective of an industrial trade association concerned about compliance costs.”

Then:

“What arguments would NGOs likely use against this position?”

Then:

“Turn this into 3 stakeholder-sensitive talking points for meetings with MEP assistants.”

This is where AI becomes genuinely useful: not for producing one output, but for accelerating thinking cycles.

2. Use AI Before Meetings, Not After

Most teams use AI to write meeting notes.

Far fewer use it to prepare strategically.

Before a stakeholder meeting, ask AI:

• What are likely pressure points for this person?

• Which recent legislative files may influence their position?

• What objections could they raise?

• Which framing would resonate most with them?

• What questions should we ask to gather intelligence?

You can even paste:

• previous meeting notes,

• public speeches,

• amendments,

• LinkedIn posts,

• consultation responses.

The result is often a far more focused and strategic conversation.

3. Turn AI Into a “Challenge Function”

One of the most valuable uses of AI in public affairs is not content creation.

It is pressure-testing your own thinking.

Try prompts like:

“Act as a skeptical journalist and challenge this position.”

“What are the reputational risks in this message?”

“Which parts of this argument sound defensive or unconvincing?”

“How would a policymaker misinterpret this sentence?”

This is especially valuable when teams become too internally aligned and stop seeing weaknesses in their own messaging.

AI can simulate friction surprisingly well.

4. Build Reusable Prompt Frameworks

The biggest productivity gains rarely come from one brilliant prompt.

They come from repeatable systems.

For example, create internal prompt templates for:

• stakeholder mapping,

• consultation analysis,

• amendment comparison,

• media scanning,

• briefing note drafting,

• event preparation,

• speech refinement,

• social media repurposing.

A good prompt framework saves time every single week.

Most teams still reinvent the wheel every time they open Copilot or ChatGPT.

5. Use AI to Think Structurally, Not Just Faster

The strongest professionals are not using AI merely to “save time.”

They use it to see patterns.

For example:

• identifying recurring narratives across consultations,

• detecting message inconsistencies,

• mapping coalitions,

• spotting emerging themes,

• comparing positioning across stakeholders,

• identifying gaps in engagement strategies.

This is where AI starts shifting from operational support to strategic advantage.

And this is likely where the biggest long-term divide will emerge between teams.

Not between those who “use AI” and those who don’t.

But between:

• teams that use AI for execution,
  and

• teams that use AI to improve judgement, positioning, and strategic clarity.

The public affairs profession is changing quickly.

The question is no longer whether AI “matters”.

The core question is: how do we integrate it without losing the human judgement, political instinct, and credibility that make public affairs valuable in the first place?

How is your team currently using AI? Operationally, strategically, or somewhere in between?

Let us know if we can help.

That’s not helpful.

So here’s a concrete walkthrough of how you can go from: There’s a new EU file, I don’t fully understand it” to: “Here is a structured, client-ready briefing note”

…using AI as a thinking partner, not just a writing tool.

Step 1: Start with a “structuring prompt” (not a question)

What most people do:

“Summarise the Energy Efficiency Directive”

What actually works:

You are an EU public affairs consultant. 
Help me structure my understanding of the revised Energy Efficiency Directive.

Break it down into:
1. Objective of the file
2. Key policy changes vs previous version
3. Stakeholders affected (by category)
4. Timeline and decision-making stage
5. 3 strategic implications for companies

Keep it structured and concise.

Why this works: You’re not asking for information, you’re asking for a mental model.

Step 2: Turn output into a client briefing skeleton

Now take the output and refine it:

Turn this into a 1-page client briefing.

Add:
- A short executive summary (5 lines max)
- Clear section headers
- Bullet points only (no long paragraphs)
- Neutral, professional tone

You now have ~70% of a usable briefing.

Step 3: Add strategic value (this is where most people stop too early)

Now push the AI further:

Based on this briefing, what should a company in the electrification sector do?

Provide:
- 3 risks
- 3 opportunities
- 3 immediate actions for a public affairs team

This is the difference between: ❌ AI user and ✅ AI-enhanced consultant

Step 4: Pressure-test the output (critical step)

AI is useful, but not reliable by default.

Use this:

Challenge your previous answer:
- What might be incorrect or uncertain?
- What requires verification?
- Where are you making assumptions?

This step alone will make you better than 90% of AI users in policy.

Step 5: Turn into internal talking points

Final prompt:

Turn this into talking points for an internal meeting.

Keep it:
- Clear
- Action-oriented
- Easy to present verbally

Key Takeaways

• Think in workflows, not prompts

• Use AI to structure thinking, not just generate text

• Always add strategic interpretation

• Always include a verification step

AI won’t replace public affairs professionals.

But professionals who know how to structure their thinking with AI will replace those who don’t.

Very few are actually making the most of its potential.

That gap is where the risk is.

Why this matters now

AI adoption is moving faster than most teams can structure themselves around it.

Teams are already:

• drafting press releases

• summarising briefings

• monitoring media

But they’re doing it without guardrails.

That’s how you end up with:

• confidential strategies pasted into public tools

• AI-generated content going out unchecked

• inconsistent quality across teams

The pattern we see often isn’t “AI-skepticism”.

A larger risk is rushed adoption without laying the foundations first.

1. Map your workflows before using any tool

Before you choose tools, understand your workflow.

Map your core processes:

• drafting

• monitoring

• stakeholder reporting

• internal comms

Then ask two questions:

Where can AI accelerate this?
Where can it cause damage?

Drafting? Sure.
Confidential legal or financial work? High-risk.

Yet many teams skip this step.

2. Write your AI policy before someone breaks a rule (accidentally even)

AI policies are often treated like legal documents.

In fact, they’re operational guides.

A good policy answers:

• Which tools are approved

• What data can (and cannot) be used

• Disclosure rules

• Ethical red lines

• What happens when something goes wrong

One rule is non-negotiable:

No sensitive/confidential information must go into public AI tools.

Keep the policy short.

If it takes 30 minutes to read, no one will follow it.

3. Don’t leave this to comms alone

AI governance is not a communications-only topic.

You need a small group that includes:

• communications

• legal

• IT

• (ideally) ethics

This group:

• approves tools

• maintains the policy

• runs periodic checks

• answers team questions

It doesn’t need to be heavy.

But it needs to exist and have authority.

4. Get leadership buy-in with reality, not hype

“AI transformation” doesn’t convince anyone anymore.

What works:

• time saved

• cost per output

• error reduction

And equally:

• reputational risk

• regulatory exposure (including the EU AI Act)

• data security gaps

The teams that get fast approval are the ones that show both sides clearly.

5. Train people before they touch real work

This is where most teams fail.

They give access to tools… before building capability.

High quality prompting is not intuitive.

Your team needs:

• hands-on sessions

• real use cases

• comparison between AI output and manual work

We consistently see the same issue:

People underuse AI because they don’t know how to guide it.

One structured session can change that (just get in touch).

6. Pilot a few tools and cut fast

Don’t evaluate tools in theory.

Test them on real tasks:

• internal summaries

• newsletters

• monitoring outputs

Run 3–5 tools in parallel.

Evaluate based on:

• quality

• security

• usability

• cost

• policy compliance

And define failure upfront.

If a tool doesn’t meet the bar, drop it.

7. Make experimentation part of the job

AI evolves too fast to track passively.

You need structured experimentation.

Block time for it.

Then keep it visible:

• one win

• one failure

• one open question

This turns experimentation into team learning.

8. Keep the human in control

AI output should never go out unreviewed.

AI can sound convincing while being:

• wrong

• incomplete

• tone-deaf

Human review is not optional.

On disclosure: be transparent where it matters.

Trust is easier to build than to rebuild.

9. Treat this as change management, not a tool rollout

Most resistance to AI is not irrational.

It’s about:

• job security

• quality concerns

• lack of clarity

Ignoring it makes it worse.

The teams that succeed:

• explain why

• show how

• create feedback loops

They treat AI adoption like what it is: an organisational change process.

So, is there a shortcut?

I wish.

There is only sequence:

• policy before tools

• skills before deployment

• testing before scaling

The teams that move fastest long-term are the ones that slowed down first.

If your team is starting this journey, we run hands-on AI workshops designed specifically for communication professionals who need to get this right.

No hype. No generic tools. Just practical systems that actually work.

How can AI support better decision-making inside EU agencies and international organisations, without compromising confidentiality, control, or institutional integrity?

In this webinar, we move beyond the hype and into practical application.

We explore how AI can help senior policy and communications professionals align outputs across teams, retain organisational knowledge, and improve the quality of analysis and messaging, while operating within strict regulatory and internal policy frameworks.

What you’ll learn:

• How to use AI to support day-to-day decision-making in institutional contexts

• Applying AI to improve coordination, consistency, and knowledge retention across teams

• Strengthening policy analysis, scenario planning, and strategic communications

• Identifying risks and using AI in a controlled, transparent, and compliant way

• Practical examples of AI tools, workflows, and prompts tailored to EU and international organisations

This session focuses on real-world use cases and actionable insights drawn from training over 2,000 professionals working in regulated environments.

This is not about outsourcing thinking to AI; it’s about using it to think better, stay in control, and deliver higher-quality outcomes.

Who it’s for:

Senior policy and communications professionals in EU agencies and international/global organisations working with sensitive information and complex decision-making processes.

If you’re looking to use AI responsibly to improve alignment, strengthen analysis, and enhance the impact of your work, you can now watch the full session (or get in touch with us for a custom workshop).

They need better thinking, faster execution, and fewer mistakes.

That’s where AI comes in, if you use it properly.

Why this matters now

Policy communications has always been a precision discipline.

One wrong word in a briefing.
One misread stakeholder signal.
One delayed response in a fast-moving file.

That’s all it takes.

AI doesn’t remove those risks.

But it does compress the time between information and action, idea and output, signal and response.

That’s why most of the teams we work with aren’t asking if they should use AI.

They’re asking:

How do we use it without exposing ourselves?

The answer starts with understanding where AI actually adds value, and where it doesn’t.

1. Draft faster. Think harder.

Yes, AI can draft a speech, op-ed, or briefing in minutes.

That’s not the real value.

The value is what you do after.

Policy communications requires:

• factual accuracy

• institutional awareness

• political nuance

AI doesn’t guarantee any of these.

The teams getting the most out of AI treat drafts as raw material, not finished output.

They don’t debate whether to use AI.

They build editorial review into the workflow.

2. Turn complexity into something people will actually read

Most stakeholders will never read:

• a 200-page impact assessment

• a trilogue compromise text

• a technical policy annex

But they still need to understand it.

AI is extremely good at:

• extracting key points

• simplifying language

• adapting content to different audiences

Briefings, summaries, LinkedIn posts, talking points, all from the same source material.

The interpretation is still human.

But the translation layer becomes dramatically faster.

3. Stop monitoring everything. Start seeing what matters.

Monitoring used to mean:

• scanning newsletters

• tracking media manually

• paying for expensive subscriptions

Now, AI can:

• surface legislative developments

• track sentiment across sources and languages

• flag emerging issues early

The key word is surface.

AI finds the signal.

You decide what it means.

What disappears is the time spent digging for it.

4. Stress-test ideas before they cost you

Most campaigns fail long before they launch.

The problem is: no one sees it early enough.

AI can act as a thinking partner:

• challenge your messaging

• simulate stakeholder reactions

• highlight weak points

Used well, it helps you kill bad ideas early, or strengthen good ones before they go public.

That’s a much cheaper place to learn.

5. Prepare for crises you can already see coming

Good crisis communication is rarely reactive.

It’s prepared.

AI makes scenario planning scalable:

• map likely crisis scenarios

• identify early warning signals

• draft response frameworks in advance

You’re not predicting the future.

You’re making sure you’re not surprised by it.

6. The human is not optional

Let’s be clear.

AI does not understand:

• institutional sensitivities

• political dynamics

• timing in a live policy environment

It doesn’t know:

• which Commissioner is sensitive on a file

• which MEP will push back

• which message will land badly this week

That’s human judgment.

Always.

AI handles volume, structure, and speed.

You handle meaning, positioning, and risk.

That’s the division of labour.

So, can AI make you better at policy communications?

Yes.

But only if you use it to amplify expertise, not replace it.

The gap we see across teams isn’t about tools.

It’s about capability.

The teams using AI well:

• know where to intervene

• know what to question

• know when to ignore it entirely

The teams using it badly:

• move faster

• produce more

• and get worse outcomes

From working with policy and public affairs teams across 8 countries, one thing is consistent:

Training, not access, is what makes the difference.

If you want your team to use AI in a way that is practical, safe, and genuinely effective, we run hands-on workshops specifically for policy communications professionals.

No theory. No hype. Just real workflows that work.