The Clarity Gap: Finding Where AI Actually Helps in Public Affairs

By Paul Shotton,

Advocacy Strategy co-Founder

"We understand AI is important, but we don't really see where it adds the most value to our work."

A PA lead at a large multinational company said this to me last week, and honestly, it was refreshing. No pretense about having it all figured out. No vendor-speak about "digital transformation." Just honest confusion about how this technology actually helps when your job is mostly about coordination and strategy, not churning out position papers.

They went on: "Everyone's talking about AI for writing and content creation. But that's maybe 20% of what we do. Most of our time goes into strategy, relationship management, coordinating across regions. Where does AI fit there?"

Great question. And based on my conversations with other PA professionals, it's the question everyone's quietly asking while publicly pretending they've got AI figured out.

The good news? We're starting to get real answers, not just hype. The Financial Times just published data showing companies getting 16.3% revenue increases from targeted AI use—not from revolutionary overhauls, but from specific, practical applications. Meanwhile, the Longitudinal Expert AI Panel (LEAP), which tracks predictions from 350 experts, offers a reality check: while tech CEOs promise human-level AI by 2026-2027, actual experts give this just a 23% chance.

The message? Stop waiting for the AI revolution and start finding practical wins.

Three Layers to Finding Your AI Value

During our conversation, I shared an approach I've been developing with various PA teams. It's not rocket science, but it seems to help cut through the confusion.

Layer 1: Actually Look at What You Do

Sounds obvious, right? But most PA teams have never really mapped out how they spend their time. When we sat down to do this, some interesting patterns emerged.

For teams focused on coordination and strategy, the workflow looks something like this: You're constantly gathering intelligence from multiple sources—what's happening in different associations and markets, who's saying what, which issues are heating up. You're synthesizing all these signals into something coherent for leadership. You're identifying which stakeholders matter for which issues. You're building coalitions and managing complex relationships across time zones and cultures. And yes, occasionally you're writing a position paper or briefing document.

The revelation for this particular team was realizing how much time they spent just trying to stay on top of information flows across their different regions. Not analyzing or strategizing—just gathering and organizing. Three people in three regions often tracking the same issues, but not sharing intelligence effectively. Sound familiar?

Layer 2: Match Technology to Actual Problems

Once you see your workflows clearly, you can start mapping technology to real pain points. And here's where that FT evidence becomes interesting—the companies seeing those 16.3% gains weren't trying to revolutionize everything. They were augmenting human work in specific, targeted ways.

For standard large language models (the ChatGPTs of the world), the sweet spots are pretty clear. They're great at first drafts of routine stuff, translating content for global campaigns, summarizing those mind-numbing 200-page regulatory documents, and turning meeting notes into something useful. Basic, but valuable.

But for PA teams doing serious strategy work, the interesting stuff is in custom AI applications:

Vote prediction models that actually work because they're trained on relevant political data, not just generic internet content.

Stakeholder and social network analysis that maps out who influences whom. Not just a contact database, but actual relationship mapping that shows you the six degrees of separation between you and that key decision-maker, and more importantly, the best path through those degrees.

Sentiment analysis that goes beyond "positive/negative" to understand nuance—is this stakeholder genuinely supportive or just being polite? Are they moveable on this issue or locked in?

Risk analysis that connects dots across multiple jurisdictions and issues. What happens if Issue A heats up in Region B while Issue C is being debated in Region D? These connections are often invisible until it's too late.

Layer 3: The Stuff You're Not Doing (But Could Be)

This is where the conversation got really interesting. I suggested we think beyond just doing current tasks better—what about things they're not doing at all?

Well, imagine having real-time coalition opportunity identification. AI scanning thousands of stakeholder positions and telling you, "Hey, this environmental group and this business association suddenly agree on Issue X—there's your opening for a conversation."

Or predictive regulatory risk scoring—models that can tell you which issues are likely to blow up six months from now based on early signals that humans might miss.

It's not just about doing our current work faster. It's about doing things we couldn't even attempt before.

Exactly. As the LEAP study predicts, by 2030 AI will assist in 18% of work hours. But those won't just be hours saved on existing tasks—they'll be hours doing entirely new types of work that create value in ways we're only starting to imagine.

Getting Real About What Works

Let's be clear about expectations. Every AI company CEO is making wild predictions, but the LEAP expert panel gives these only a 23% chance of coming true in the coming years Meanwhile, that 16.3% productivity gain from the FT study? That's happening right now with today's technology.

The key insight: don't wait for miracle AI. Start with what's proven. The companies seeing real gains implemented gradually, learned what worked, adjusted, and kept going. And despite all the job-loss hysteria, LEAP experts predict only a 2% change in white-collar employment by 2030. In PA—where relationships and judgment are everything—humans remain essential. AI just makes us better at our jobs.

Where to Start

Instead of a rigid roadmap, here's what's actually working for PA teams:

Start by picking one workflow. Just one. Maybe it's tracking stakeholder positions across multiple issues. Maybe it's synthesizing input from different regions. Maybe it's predicting which issues will heat up next quarter.

Map out how you currently handle that workflow. Be honest about the pain points. Then experiment. Try a basic AI tool for part of it. See what works, what doesn't. Learn. Adjust.

The multinational team I mentioned? They started with their regional intelligence gathering problem. Simple first step: they set up an AI system to monitor and synthesize regulatory developments across their key markets. Nothing fancy. But it freed up enough time that they could actually think strategically about the information instead of just collecting it.

The Real Question

The PA teams that thrive won't necessarily be those with the biggest AI budgets or the fanciest technology. They'll be the ones who best understand their own work, who thoughtfully apply technology to real problems, and who have the courage to explore not just efficiency gains but entirely new capabilities.

Sources:

• Financial Times: "From AI to ROI: Some Positive Evidence" (November 2025)

• The Economist: "A New Project Aims to Predict How Quickly AI Will Progress" (November 2025)