Rethinking Street-Level Bureaucracy in the Age of Agentic AI

The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.

By Bill Brantley
July 18, 2025

Rethinking Street-Level Bureaucracy in the Age of Agentic AI
Michael Lipsky’s (1980) Street-Level Bureaucracy argues that frontline public service workers—such as police, teachers and caseworkers—significantly influence policy outcomes through their discretionary decisions. These street-level bureaucrats must navigate limited resources, unclear rules and high workloads, often relying on routines to balance demands.

The emergence of agentic AI (AI that acts independently, pursues goals and adapts) necessitates rethinking SLB theory. Beyond automating tasks, agentic AI impacts decisions, resource distribution and human judgment. This article explores how such AI is changing street-level bureaucracy and its implications for public administration.

Defining Agentic AI
Unlike traditional automation driven by set rules, agentic AI operates with independence and intentionality. Built on agent-based systems and often using reinforcement learning, these systems adapt to context, set goals and make trade-offs in real time. Agentic AI can act with “bounded autonomy” within policy constraints, serving as actors rather than mere tools in administrative tasks.

The Traditional Role of Discretion in SLB Theory
Street-level bureaucracy highlights how frontline workers use discretion to implement policy. SLBs interpret mandates based on their judgment and experience, which is essential in complex situations. However, the exercise of discretion may result in outcomes that are inconsistent, biased or inefficient, as street-level bureaucrats may ration services or employ informal practices to address constraints. Discretion thus represents both a critical asset and a significant challenge in the delivery of public services.

How Agentic AI Reconfigures SLB Discretion
Agentic AI systems are increasingly being used to supplement or substitute for human judgment in SLB contexts. Examples include:

AI-driven eligibility assessments used in welfare programs

Predictive policing tools that recommend patrol routes or identify individuals assessed as higher risk

Automated triage systems implemented in public health and emergency services

AI-based education platforms that tailor instruction and assess student work

These technologies perform more than routine tasks; they can structure available choices, filter information and limit the scope of human decision-making. As algorithms are used to mediate access to services, they play a growing role in shaping policy implementation, a function previously managed by SLBs.

Some scholars contend that algorithmic discretion merely shifts decision-making from street-level bureaucrats to AI system developers rather than eliminating discretion altogether.

Three Impacts of Agentic AI on SLB Theory
Displacement of Discretion – Agentic AI streamlines service delivery by limiting discretion via algorithms, boosting consistency but reducing flexibility and missing nuances that human SLBs might catch. This creates a paradox: decisions become more uniform yet often lack humanity and context sensitivity.

Redefinition of the Bureaucratic Actor – Agentic AI introduces ambiguity regarding the distinction between human and non-human agents within public administration. The question arises as to whether “street-level” bureaucrats are still involved when autonomous systems operate in these environments. As AI systems undertake tasks such as identifying potential abuse, recommending sanctions or denying benefits, they increasingly function in quasi-bureaucratic roles, leading theorists to reconsider concepts of agency and responsibility in public service.

Amplification of Accountability Dilemmas – Agentic AI systems introduce complexities to established accountability frameworks. Determining responsibility becomes challenging when, for example, a predictive policing algorithm disproportionately affects certain communities.

Similarly, when a chatbot provides inaccurate unemployment guidance, questions arise about whether accountability lies with the AI itself, its developers or the deploying agency.

These considerations highlight the necessity for a revised administrative ethics that address both the opacity inherent in algorithms and the distributed character of decision-making within AI-integrated bureaucratic systems.

Implications for Public Administration Scholars
Agentic AI doesn’t negate SLB theory but calls for its expansion. Public administration should now address hybrid human-machine discretion within socio-technical systems.

This shift invites new research questions:

How do SLBs adjust their roles in AI-enhanced settings?

What training and oversight support responsible agentic AI use?

How can algorithm design be democratized to mirror public values?

How might agentic AI return discretion to SLBs by managing routine tasks?

A Cautionary and Constructive Path Forward
Agentic AI in public service offers potential benefits such as improved efficiency, scalability and precision. However, it is important for scholars and practitioners to exercise institutional humility. AI is intended to support rather than replace human elements within governance, and discretion should be considered both a technical and ethical concern.

Michael Lipsky cautioned that lacking proper safeguards, street-level bureaucrats may unintentionally enforce injustice. Similarly, agentic AI acts as a policy actor and requires regulation, oversight and democratization.

As public administration evolves, we need to rethink digital bureaucracy to ensure it continues to meet people’s needs. As public administration scholars, we must be in front of the evolution of digital bureaucracy.

---

Author: Dr. Bill Brantley is the President and Chief Learning Officer for BAS2A, an instructional design consultancy for state and local governments. He also teaches at the University of Louisville, the University of Maryland and Franklin University. His opinions are his own and do not reflect those of his employers. LinkedIn: https://www.linkedin.com/in/billbrantley/

 

(3 votes, average: 4.33 out of 5)
Loading...