The AI Transformation Project in Public Affairs: Where Most Teams Go Wrong

There is a conversation happening inside Public Affairs functions right now that almost always starts the same way.

"We have the licences. People are using the tools. We ran a session on prompting. But

honestly? It doesn't feel like we've actually changed anything."

I hear this from Heads of Government Relations in Brussels, from VP-level PA leaders in multinationals, from Trade Association Directors managing increasingly stretched teams. And every time I hear it, my answer is the same. The reason it doesn't feel like anything has changed is because — structurally — nothing has. The tools arrived. The transformation didn't.

That is the central challenge of AI in Public Affairs right now. And it is not a technology problem. It is a process and structure problem.

The Equation You Need to Understand First

In my recent posts I have introduced what I believe is the most important framing for any Public Affairs AI transformation:

Successful Public Affairs = AI × (Data + Process + Structure + Expertise)

The critical word is multiplier. AI does not create value independently. It amplifies whatever is already in the parentheses. Strong data foundations, clear internal processes, agreed ways of working, experienced professionals — AI sharpens all of these. But if those foundations are weak, AI does not hide the weakness. It accelerates it.

This is why so many PA functions feel stuck. They added the multiplier before they built what it is supposed to multiply.

The nature of a PA AI transformation project, properly understood, is therefore this: it is a

Public Affairs management project that happens to use AI — not an IT project that lands on the PA team. Getting this distinction right from the start determines almost everything that follows.

What This Transformation Actually Involves

1. Data: Your Most Neglected Strategic Asset

Most Public Affairs functions already have a data problem. They just haven't been forced to confront it yet — because the tools have been hiding it.

Think about where your intelligence actually lives today. A monitoring platform subscription. A CRM that is partially used and partially trusted. A shared drive of documents in formats chosen by external consultants. Meeting notes sitting in individual inboxes. The accumulated institutional memory of your most senior colleagues — in their heads, nowhere else.

This is not a data strategy. It is data debt.

The functions I see operating well are taking a fundamentally different posture. They are treating data as a strategic asset rather than a by-product of activity. They are building a deliberate capture, store, use discipline that survives staff turnover and does not depend on any single supplier continuing to exist. And critically, they are in-sourcing the intelligence layer that was previously outsourced to agencies and consultancies — not to cut those relationships, but to ensure the data those relationships generate actually lives inside the organisation and compounds over time.

Without this foundation, AI tools produce output. They do not produce advantage. Generic AI applied to generic data is indistinguishable from what every competitor can produce. The competitive edge comes from applying AI to your stakeholder maps, your issue history, your institutional context. That is what makes the output proprietary.

The practical starting point for most teams is an honest AI and data audit: where does your data live, who owns it, what leaves the organisation when a consultant relationship ends, and what would you need to do to make your data actually queryable and useful? This audit alone is often clarifying in uncomfortable ways.

2. Process: Turning Data Into Impact

Data without process produces insights that nobody operationalises. Briefings that nobody reads. Analysis that sits in slides presented once and then archived forever.

The teams I see pulling ahead are not defined by the intensity of their effort — they are defined by the consistency of their execution. They have clear, repeatable processes for how they move from monitoring a policy development to analysing its business implications to deciding what to do about it to reporting the outcome internally.

For PA teams investing in AI, this has a very specific implication: you need to map your workflow before you automate it. If you automate a broken or vague process, you get a broken or vague process running faster. That is not transformation.

The GOST framework — Goals, Objectives, Strategy, Tactics — is one way to impose this discipline. It gives the function a shared internal language, separates what leadership needs to understand from what the PA team needs to manage, and creates a consistent structure for planning, execution, and reporting. When AI is applied on top of this kind of structured process, it can genuinely accelerate execution: drafting position papers against agreed messaging frameworks, generating briefings in formats internal stakeholders actually use, tracking progress against defined objectives rather than counting meetings held.

Without process clarity, AI generates motion. With it, AI generates progress.

3. Structure: Templates Are the Backbone — Not the Bureaucracy

I want to challenge a view I hear too often in PA teams. The idea that templates are somehow the enemy of good work. That they constrain judgment, flatten expertise, and reduce a sophisticated function to a form-filling exercise. I understand the instinct. I disagree with the conclusion.

Templates are not bureaucracy. They are the institutional memory of a PA function made visible. They are what separates a team that starts from scratch on every campaign from one that compounds its learning over time. And in the context of an AI transformation, they are the single most important structural investment you can make — because they are the raw material AI actually works with.

Think about what a good PA template does. It captures what the function has learned about how to frame a policy brief for the executive team. It encodes the standard of a stakeholder engagement note that is actually useful six months later. It defines what a well-constructed issue assessment looks like — the questions it must answer, the information it must contain, the format that internal stakeholders have learned to trust. Templates are the crystallisation of professional standards into something replicable.

Now think about what happens when AI meets a well-developed template library. The function can prompt an AI model with a stakeholder briefing template and get a first draft that is already structured the way your organisation expects — not a generic output that needs to be pulled apart and rebuilt. It can generate internal reporting that follows the Development-Action-Impact framework every time, automatically. It can produce position paper drafts that begin from your own agreed messaging architecture rather than from a blank page. The quality of the AI output is directly proportional to the quality of the template it works from.

The reverse is equally true. If your function does not have shared templates — if every colleague produces briefings in their own format, if stakeholder notes are wherever people happen to save them, if there is no agreed standard for what a good issue assessment contains — then AI has nothing of substance to anchor to. It produces output that looks credible and is structurally hollow.

I have written before about the importance of developing your own PA 'Way' — a branded, internal methodology that defines how your function works. Templates are the practical expression of that Way. They cover every core PA output: issue trackers, stakeholder maps, engagement notes, position papers, internal reporting dashboards, briefings for senior leaders, campaign plans, monitoring summaries. Each one represents a decision about what good looks like in your organisation.

Building this library is not glamorous work. But it is the work that pays the highest long-term dividend — both for your AI transformation and for the function's overall quality and consistency. Teams that have invested in it onboard new colleagues faster, maintain standards across markets, and are far better positioned to capture the genuine efficiency gains that AI promises.

One practical test: if you removed your three most experienced people tomorrow, how much of their knowledge would survive in your templates and documented processes? The answer to that question tells you a great deal about where your function actually stands.

4. Expertise: More Important, Not Less

There is a widespread anxiety in the PA profession that AI will reduce the value of experienced professionals. I think the opposite is true — with one important caveat.

AI can process volume at extraordinary speed. It cannot read political nuance. It cannot judge when not to act. It cannot build the kind of trust with a policymaker that takes years of consistent, credible engagement. It cannot tell you that the beautifully structured output it just generated is quietly, confidently wrong because it missed the context only a senior professional would have.

What AI is doing is changing where expertise matters most. The skills that used to be about production — drafting, summarising, monitoring — are increasingly baseline expectations that tools can support. The skills that are becoming more decisive are evaluative: knowing what is wrong with the draft, knowing when the stakeholder map misses the three people who actually decide, knowing when a political moment requires restraint rather than engagement.

This is a profound shift in what we should be hiring for and developing. In practice it means that PA AI transformation is also, necessarily, a capability-building project. You need to invest in your team's ability to work critically with AI outputs — not just to use the tools, but to interrogate them, adapt them, and know when to discard them entirely.

The Rules of Running This Project Well

Based on what I am seeing across organisations and what the PA & AI equation tells us, here are the principles that separate transformations that land from those that stall:

Start with the workflow, not the tool. Map what your team actually does — from intelligence gathering through to internal reporting — before evaluating any technology. The use cases that are worth automating will become obvious. The ones that look attractive but require a stronger data or process foundation will also become obvious.

Make it a team project, not an individual experiment. The most consistent pattern I see is AI knowledge accumulating with individuals and their personal subscriptions, never making it into the function's institutional practice. AI adoption at the team level requires deliberate co-ordination: shared prompts, shared use cases, shared standards for what good output looks like. This does not happen organically.

Govern the shadow AI gap. Many PA professionals are already using AI tools that their organisation has not sanctioned, because the sanctioned tools are less capable or too restricted. This is understandable — and manageable — if it is acknowledged and governed. Ignoring it means losing the learning that individual experimentation generates. Addressing it means building a policy and a pathway: here is what you can use, here is what you cannot, here is how we are going to learn together.

Protect proactive capacity. AI should, over time, free up meaningful time from transactional tasks. The question is what you do with that time. If it simply gets absorbed by additional reactive requests — more briefings, faster turnarounds, wider monitoring coverage — the transformation produces efficiency but not strategic progress. The goal should be to redirect recovered capacity toward the proactive PA work that has the highest long-term impact: relationship-building, early policy shaping, the coalition work that takes months to bear fruit.

Measure what matters, not what is easy. The temptation when investing in AI is to measure outputs: documents produced, time saved, tools adopted. These are not the right metrics for a PA function. The right metrics remain the same ones that always mattered — progress against objectives, policy outcomes, business impact. AI should make these things easier to achieve and easier to demonstrate. If it is not doing that, something in the equation is still missing.

Where to Start

If you are leading a PA function and this landscape feels overwhelming, the most useful place to start is an honest assessment of where you are in the equation.

How structured is your data? How repeatable is your process? How developed is your template library — and does it actually reflect how your best people work? How well is your team able to evaluate, rather than just produce?

The answers will tell you where your weakest link is — and that is where the transformation needs to begin. Not with the AI tool. With what the AI tool needs in order to actually work.

The competitive advantage in the next phase of Public Affairs will not come from adopting AI. It will come from being ready for it.