Governments love shiny new toys. Right now, artificial intelligence is the shiniest toy in the room. Public sector leaders are racing to adopt automated systems to sort through benefits applications, manage traffic, and predict infrastructure failures. But there is a massive problem. Most of these public management programmes focus entirely on the technology while ignoring the actual human beings who have to use it, manage it, and live with the consequences.
Tech vendors sell a dream of flawless efficiency. The reality is often messy, biased, and alienating for citizens. When public agencies implement algorithms without heavy human oversight, things break. True innovation in public administration does not come from building the most complex model. It comes from embedding human empathy, ethics, and accountability directly into the system. Meanwhile, you can read related stories here: The Unit Economics of Military Autonomous Systems Analysing the US Army P550 Procurement.
The Dangerous Trap of Algorithmic Bureaucracy
Public management programmes that treat AI as a magic wand usually fail. They fail because government work is fundamentally different from private business. A tech startup can afford to move fast and break things. A public welfare agency cannot. If a streaming service recommends a bad movie, you lose two hours. If a public management algorithm wrongly flags someone for fraud, they lose their home.
Look at real historical data to see how this plays out. Between 2013 and 2020, the Netherlands used an algorithmic system to detect childcare benefit fraud. The system relied on indicators like dual nationality and low income. It incorrectly flagged tens of thousands of families, forcing them into catastrophic debt. The Dutch government collapsed in 2021 specifically because of this scandal. It was a failure of technology overriding human judgment. To explore the full picture, check out the excellent analysis by Engadget.
Algorithms are trained on historical data. That data reflects past human biases and systemic inequalities. When you feed that data into a machine learning model, you do not eliminate bias. You automate it at scale.
- Machines lack context. They see numbers, not human struggles.
- Automated systems do not have empathy. They cannot bend a rigid rule for a desperate citizen.
- Black-box models hide accountability. It is incredibly difficult for an ordinary citizen to challenge a decision made by a hidden piece of code.
Public officials must stop treating AI as an objective truth. It is a tool. Nothing more.
Reclaiming the Human Element in Public Sector AI
So how do we fix this? Public management programmes must pivot toward a human-centric framework. This means training public servants not just to use software, but to question it, audit it, and override it when necessary.
Human-centric public management means establishing strict protocols where machines assist humans, rather than replace them. The ultimate decision-making power must always rest with a person. If an algorithm flags a tax return or a housing application for rejection, a human expert needs to review the case from scratch before any action is taken.
We also need to bring diverse voices into the development process. Most government tech contracts are handed over to external corporations staffed by engineers who have never set foot in a public housing office or a community clinic. Public managers need to co-design these tools alongside frontline workers and citizens. The people who know the community best should have the final say on how the technology behaves.
Building the Human-Centric Toolkit
Shifting to a human-focused approach requires changing the way public sector workers are trained. We need a new curriculum for public administrators that prioritises algorithmic literacy, data ethics, and civil rights alongside traditional management skills.
First, public managers must learn how to demand transparency from tech vendors. You should never purchase a proprietary system that does not allow your internal teams to inspect the underlying source code and training data. If a company tells you their model is a trade secret, walk away.
Second, continuous auditing must become standard practice. You cannot just deploy a system and forget about it. Public agencies need independent oversight boards composed of ethicists, data scientists, and community advocates to regularly test models for drift, bias, and accuracy.
Start small. Instead of overhauling an entire department overnight, pilot automated tools in low-stakes environments. Use them to schedule internal meetings or draft basic administrative emails. Observe how the staff interacts with the tool. Learn where the friction points are before you even think about using technology to deliver critical public services. Ensure your frontline staff feels empowered to say no to a machine.
