SWAAP Thinking: Where AI Ends and the Bureaucracy Begins

When are we?

Since I shared The Bureaucratic Brain on GenAI, the conversation around AI has accelerated. New terms, ideas, and thinkers are sharpening my understanding and reshaping how I think about and approach AI implementation in practice.

In Australia, governments are now more AI-aware—Parliamentary committees, working groups, frameworks, pilots, and guidelines are everywhere. One 2025 report, subtitled Proceed with Caution, reminds us:

Governments have multiple roles in relation to AI, as enablers, funders, regulators, but also as users and in some cases as developers… [while] less attention has been paid to their responsibilities as users of AI.

This was the central premise in my last piece: how do we, as public servants, responsibly shape our institutional futures? Regardless of the AI-world we end up in, we each play a part in how it gets integrated into our work.

From AI as tools to AI as tempo

I initially viewed AI and GenAI more as discrete tools to reimagine workflows. By mid-2025, they feel more like a background rhythm. While adoption remains patchy (maybe even more patchy), the underlying pressure reminds me of early COVID—a pervasive hum forcing us to adapt our mindsets and framing to keep up with where the game is.

‘AI’ and ‘GenAI’ now feel too small to contain the scale of change, and too imprecise for many of the conversations that are needed. Last year I observed how I felt a great new age of question-making was arriving, and it is still something I consider a core skill for thoughtful policy work. But deploying tools well requires guided design, coaching, and team-based participation, not just better prompts.

Institutional maturity means going beyond outputs to examine interdependencies and second-order effects. The question isn’t just what we can do with AI, but why, when, and how deeply—and what the ripple effects might be.

In exploring where GenAI might enhance public service roles, I’ve seen how purpose and process are often fused. Sometimes it’s just a case of microtask decluttering. But strategic shifts require clearer purpose, task mapping, and current-state understanding. There be dragons.

Are we still even talking about AI?

Framing AI as only a technology or toolset can limit our view. It narrows the challenge to procurement and delivery and traps us between speculative AGI/ASI fears, or the more mundane. We’re also mindful of our spotty accountability track record —like Robodebt or the UK Post Office scandal - and what automation at scale may mean.

‘AI’ is often too broad to scope as a single project and too narrow to capture the full complexity. On its own, it’s become almost meaningless - except as a lightning rod for a long conversation. At a recent AI CoLab workshop, we tried to map it using a 5×5 matrix. The number of boxes tripled. Instead of clarity, we had an exploded diagram of overlapping domains.

We’ve all tried slicing it into parts—data, regulation, governance, capabilities—but the overlaps persist. Eventually, you either draw arbitrary borders or go looking for a bigger container.

I’d like to stop talking about just AI, please

That’s why I’m proposing a broader frame: Strategic Workflow Automation, Agency and Practices (SWAAP).

• It recognises that automation, agency, and practice design are interlinked, but distinct. Each has its own stakeholders, expertise, and strategic questions

• It’s more technology-neutral, and fits whether we’re replacing, removing, or reimagining workflows (processes) and our practices in how we approach design and engage them (expertise, role types)

• ‘Agency’ is a head nod to the swarms of robot ‘agents’ that will apparently replace us any minute, but also centres human agency—individual and collective—as the strategic lever. Replete with our values, feelings, and expertise (particularly as humans)

Even if we are only talking about AI, the real challenge lies in embedding it in actual institutions—not the phantom ones that look neat and tidy in frameworks and org charts.

SWAAP probably isn’t quite right, and I’m proposing something imperfect partially to make a point. We are still at a time when we need to accept that, no matter your views on ‘AI’, meaningful progress is predicated on moving without having all the answers—or even the best questions. To frame that another way: anything we do is almost guaranteed to be wrong or insufficient to some degree. Not a comfortable space for public servants to find themselves.

Hyperobjects and institutional dissonance

Hyperobjects are:

vast, non-human phenomena that stretch across planetary spatial scales and geological timeframes. They are so massive and temporally deep that they defy total human perception [my emphasis]...[and they require a new way of thinking about the boundaries of organisations, limits of agency, extent of property rights, contracts and exchange, and the meaning of externalities

In a recent paper that got a lot of attention, this is coupled with how AI confronts us with:

unavoidable differences in values and beliefs [imply that] policymakers must adopt value pluralism, preferring policies that are acceptable to stakeholders with a wide range of values, and attempt to avoid restrictions on freedom that can reasonably be rejected by stakeholders

Traditional governance systems were never designed for this. I’d like to see the matrix design capable of wrangling this into tidy silos. None of this sits too comfortably with institutional reflexes towards a matrix-based world order—rationally bounded, linear, and controllable. In my last piece I conveniently parked all the big questions to focus on the practical dimension. However, it’s clearer now how interconnected the existential and tactical questions are.

Temporal Literacy

Governments are often criticised for being slow. Internally, things can feel both too fast and too slow—which is probably healthy. Some institutions should be stable. Others must stay attuned to shifting social priorities.

Temporal literacy, and knowing ‘when’ we are, might be a key concept going ahead. A real practical challenge is distinguishing what should move slowly from what can move quickly. Government institutions are living and breathing organisations, no matter how moribund they may appear, because ultimately they are (still) composed entirely of humans, with human values doing human things, for people.

Proactively or not, we are already on a journey of redesigning and reimagining our institutions, and the short term choices will have longer term impacts on our capabilities, approaches and practices. There is plenty of room in that conversation for everyone!