The AI Agent Dilemma: Should We Build Them — or Wait for Them to Build Themselves?

By Paul Shotton, Advocacy Strategy

Over the past year, the conversation around artificial intelligence has shifted dramatically. Many professionals began by experimenting with tools like ChatGPT or Claude to draft emails or summarise documents. Now the conversation has moved to something more ambitious: agents.

In fields like public affairs, it is easy to imagine agents that analyse legislation, produce monitoring reports, or help structure advocacy strategies. The idea is attractive because agents promise to automate complex knowledge workflows.

But the excitement around agents also raises an important strategic question. Should organisations actually be racing to build them? Or are we entering a race that may not be ours to run?

Building Agents Also Means Becoming a Technology Operator

When organisations talk about building AI tools or agents, the conversation often focuses on development. But building technology is only the beginning. Once a system exists, it must be operated and maintained. Data, infrastructure, and governance must all be managed.

In other words, building an AI tool means taking responsibility for running it. For many organisations, this represents a significant shift. Building a custom AI system creates an ongoing operational responsibility that requires technical expertise and continuous oversight.

This raises a deeper strategic question: are organisations in fields like public affairs actually trying to become technology companies?

Understanding the AI Ecosystem

Innovation research shows that major technologies evolve through ecosystems in which different actors play different roles. Some organisations build core platforms, others build complementary tools, and many focus on applying those technologies within specific domains. Artificial intelligence is evolving in exactly this way.

The emerging AI landscape can be understood as a stack of different layers.

Layer 1 — Foundation Model Developers

At the top of the ecosystem are the organisations building the core models. These include companies such as Google, Meta, OpenAI, and Anthropic. They invest enormous resources in research, computing infrastructure, and engineering talent to develop the models that power most modern AI applications.

Layer 2 — Enterprise Integrators

Below them sit large consulting firms and technology integrators. These organisations focus on embedding AI capabilities into enterprise systems and organisational infrastructures. They translate the capabilities of foundation models into solutions that can operate within complex organisations.

Layer 3 — Sector-Specific AI Developers

Further down the ecosystem are companies developing specialised tools for particular industries. These organisations build applications tailored to specific industries, often relying on the capabilities of foundation models.

Layer 4 — Professional Users and Workflow Designers

At the base of the ecosystem are the organisations and professionals applying AI in their daily work. This includes corporate teams, consultancies, advocacy organisations, and public affairs departments. Their expertise lies not in building AI models, but in applying them within complex professional workflows.

Where Does Public Affairs Fit in This Ecosystem?

Once the ecosystem becomes visible, the strategic question becomes clearer: where do we actually fit?

Most public affairs professionals are trained in policy, law, economics, or communications. Their expertise lies in understanding political systems, interpreting legislation, analysing policy dynamics, and developing advocacy strategies. Artificial intelligence does not replace these competencies. Instead, it introduces tools that can support how these capabilities are applied.

Seen through the lens of the AI ecosystem, public affairs professionals are unlikely to compete in the layers where foundational models or large-scale AI infrastructure are developed. Those layers require levels of investment and technical expertise far beyond the scope of most policy organisations or consultancies.

Instead, public affairs organisations operate closer to the application layer of the ecosystem. Their value lies in understanding policy processes, structuring analytical workflows, and translating complex information into strategic insights.

How Far Up the Value Chain Should We Go?

This raises an important strategic question: how far up the AI value chain should organisations in our field try to move?

Some organisations may build specialised tools. Others may partner with developers. Many will focus on integrating AI into their existing workflows. Each of these choices represents a different position within the ecosystem.

Business strategy research often suggests that organisations should position themselves in parts of the value chain where their capabilities provide the greatest advantage. For public affairs professionals, that advantage typically lies not in developing AI infrastructure but in applying technology within complex policy and advocacy processes.

The Exploration Dilemma

Management research describes a tension between exploiting existing capabilities and exploring new opportunities. Organisations must continue delivering the work they are already good at while also investing time and energy in understanding emerging technologies.

Exploring AI too slowly risks leaving organisations behind in a rapidly evolving profession. But investing too heavily in building technologies outside one’s core competencies may also divert attention and resources away from the capabilities that make the organisation valuable in the first place.

This tension is increasingly visible across the public affairs profession as firms experiment with AI tools, partner with developers, or offer technology-related services while remaining fundamentally organisations built around policy expertise and strategic advisory work.

The Emergence of the Public Affairs Technologist

Within this evolving landscape, a new hybrid role may be emerging.

Some professionals are becoming fluent in both policy processes and AI technologies. They experiment with tools, design workflows, and collaborate with technologists to adapt AI systems to professional practice.

These individuals are not necessarily building the underlying technology themselves. Instead, they act as translators between technology and professional expertise. In that sense, the emerging role may be something closer to a public affairs technologist — someone who understands both the policy domain and the technological tools that can support it.

Choosing Which Race to Run

Artificial intelligence is often described as a race. But not everyone in the ecosystem needs to run the same race.

Technology companies are racing to build more powerful models. Large consulting firms are racing to integrate AI into enterprise systems. Startups are racing to build specialised tools.

Public affairs professionals may be running a different race entirely. Their challenge is to understand how AI can support policy analysis, mapping, monitoring, and advocacy.

In that sense, the most important capability may not be building agents. It may be understanding the workflows and professional processes that those agents could support.

And that brings us back to the central dilemma.

The question is not simply whether we should build AI agents. The question is whether building them is the race we actually want to run.