How Public Affairs Teams Should Assess LLM AI Tools

By Paul Shotton, Advocacy Strategy

Start with the work, not the market noise

Public affairs teams are now surrounded by AI product launches, feature comparisons, and confident claims about which platform is moving fastest. That can be useful, but it can also be misleading. The real question is rarely which tool looks most impressive in the abstract. The real question is which tool is actually suitable for the work your team needs to do, the maturity of your current practice, and the level of complexity you are ready to absorb. That is particularly important in public affairs, where workflows are rarely generic, information is often sensitive, and teams are usually expected to turn intelligence into action rather than simply produce text more quickly.

This is why I do not think public affairs teams should approach the market by looking for one perfect tool. At least for now, that tool does not really exist. ChatGPT, Claude, Gemini, and Perplexity each offer useful but different strengths. OpenAI currently positions ChatGPT around Projects, custom GPTs, connected apps, and voice; Anthropic is pushing Claude through Projects and Skills; Google is emphasizing Gemini’s Deep Research and Workspace integration; and Perplexity continues to position itself as a real-time answer engine while also expanding its research tooling. (OpenAI Help Center)

Why this question is genuinely relevant to public affairs

In one sense, this is a broader management issue rather than something unique to public affairs. Many professional teams are trying to work out how AI fits into research, drafting, coordination, and internal reporting. But public affairs experiences this problem in a particularly demanding way. Teams are often relatively small. Workflows are unevenly defined. Information comes from a mix of internal and external sources. And the outputs are not only informational, but political and strategic. Public affairs teams are not just trying to write faster. They are trying to monitor developments, structure intelligence, prioritize issues, support engagement, and communicate internally in ways that are timely, defensible, and aligned with broader organizational goals.

That broader logic then plays out differently depending on the type of organization. A trade association has to think about member inputs, committee structures, board reporting, and the challenge of translating multiple views into a coherent internal position. A consultancy may have more variable workflows across monitoring, drafting, project management, and client servicing, but it also has to manage clear confidentiality boundaries across accounts. An in-house team may have more access to wider organizational resources, but it also operates within company-wide policies on systems, privacy, governance, and reporting. So although the underlying assessment challenge is broad, the public affairs lens is still real: the mix of small teams, sensitive information, variable maturity, and stakeholder-driven workflows makes tool selection particularly consequential.

There is no single tool that fits every public affairs need

This is why I would resist the idea that one platform should automatically become the answer to everything. In practice, most public affairs teams are more likely to end up with a dominant tool for some core workflows and one or two supporting tools for other tasks. A team might prefer one platform for conversational thinking, quick drafting, or internal structuring, another for deeper research or Workspace-connected retrieval, and another for real-time answer discovery or source-led exploration. That is not a sign of failure. It is a reflection of the current market.

The platforms themselves point in different directions. ChatGPT’s current structure supports long-running work through Projects and repeatable work through custom GPTs. Claude is increasingly explicit about reusable workflow support through Skills. Gemini is leaning into Deep Research and close links with Google tools such as Gmail, Drive, and Workspace. Perplexity’s core positioning remains centered on fast, real-time answers and research. Those differences matter because public affairs teams do not only need one type of support. They need help with quick thinking, structured recurring workflows, research, drafting, and internal reporting, and those are not necessarily best served by the same environment. (OpenAI Help Center)

Assess your maturity before you assess the tool

One of the biggest mistakes teams can make is to evaluate tools without first evaluating themselves. If workflows are poorly defined, outputs are inconsistent, information is scattered, or only one or two people understand how to use AI well, then a highly sophisticated setup may create more friction than value. In those situations, the issue is not only the quality of the platform. It is the maturity of the team’s current practice.

That is why I think tool adoption decisions should be linked to milestones along a maturity curve. At an exploratory stage, a few individuals are testing basic use cases such as summarization, note structuring, or drafting support. At a guided stage, the team begins to identify a small number of repeatable use cases and some informal good practice starts to emerge. At a structured stage, workflows are clearer, outputs are more defined, and the team can begin to think seriously about more tailored setups. At an integrated stage, AI starts to sit within a broader working environment rather than alongside it, and the questions become less about novelty and more about permissions, information management, review, and governance. The point is not to force every team into a maturity model. It is to recognize that each stage creates a different decision point, and that the right next tool depends heavily on where the team actually is.

Capability rises with complexity

This, for me, is the central trade-off. The more capable and customizable a tool becomes, the more value it may be able to create. But the more capable and customizable it becomes, the more complexity it usually introduces as well. That complexity can take several forms. It may mean more time spent setting up projects, instructions, workflows, and reusable assets. It may mean more effort to keep context current and outputs consistent. It may mean more training, more review, and stronger internal governance. It may also mean the setup becomes dependent on a small number of advanced users, which creates a risk of uneven adoption across the team.

This is also where customization enters the picture, even if it is not the central point of this article. A tool should not be assessed only on its out-of-the-box experience. It should also be assessed on whether it becomes more useful when given better context, better instructions, and better supporting material. That does not mean every team should rush into deep customization. It means serious assessment should take into account not only what the tool does on day one, but what it can do when set up properly for recurring public affairs work.

A practical framework for assessment

In practice, I think public affairs teams should ask a small set of disciplined questions. Does the tool fit the team’s most important public affairs use cases, rather than generic productivity tasks? Does it match the team’s current maturity and internal capability? Does it support the workflows the team actually uses, whether that means monitoring, briefing, reporting, knowledge retrieval, or drafting? Does it fit the information environment in which the team operates, including permissions, governance, and comfort with data handling? And what is the real implementation burden, not only in subscription cost but in time, maintenance, onboarding, and management attention?

One more question matters as well: how easy would it be to adjust course later? This is not a minor issue. The more deeply a team builds habits, instructions, workflows, and expectations around one platform, the harder it may become to change direction later. There may be technical switching costs, but there are also organizational ones. That lock-in risk matters even more in a fast-moving market, where today’s strongest feature set may not look the same in a year.

A better way to decide

So my own view is that public affairs teams should move forward, but with discipline. They should not choose tools on hype, feature envy, or the idea that one platform must become the answer to everything. They should assess them against real use cases, actual workflows, current maturity, implementation burden, and the level of complexity the team is genuinely ready to manage.

For some teams, the right next step will be a relatively simple tool used consistently for a few high-value tasks. For others, it may be a more structured environment with better context, clearer workflows, and deeper integration into the way the team already works. The important point is that there is no single tool that fits every public affairs need right now. The better question is which tool should play which role, and whether the team has the maturity and discipline to use it well.