Change is inevitable; and when industry changes occur, they can be fast, real, and powerful. Take for example the change from physical media to streaming platforms: industries moved from reels to VHS, VHS to laserdisc, disc to video rental, rental to streaming services, and so on. Once dominant platforms and participants either adapt or lose relevance.
Public sector entities face similar choices today in the technology space with the rise of AI and efficient automation, but keeping up doesn’t have to be a done in a single “big bang.” It can begin with automating repeatable processes, innovating ways to manage existing data, and establishing responsible guidelines.
AI has quickly become a priority for organizations, across almost every industry, seeking to modernize operations. Yet, successful modernization doesn’t start with forcing AI into every process. It starts by understanding how business processes move, where organizations encounter friction, and what combination of automation, data integration, and human input can improve the outcomes being produced.
The Problem is Not a Shortage of Technology
Inefficient processes, work overload, siloed systems, operational bottlenecks, frequent and redundant activities, and limited insight are just some of many headwinds facing organizations trying to modernize.
AI alone will not combat those headwinds, and adding AI incorrectly may only serve to produce a faster version of the same problem. The broader opportunity involves an orchestration of enterprise: connecting systems, data, decision making, controls, and people so that the work can move seamlessly, consistently, and reliably across organizational roadblocks.
“I suppose it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail.”
When AI becomes the default answer to every problem, organizations overlook simpler and more dependable options. The strongest solution is not necessarily the most sophisticated one.
AI Belongs to the Broader Automation Toolbox
Put simply, automation is a broad approach to improving how work gets done. It is not a single product, platform, or technology. Instead, it can include rules-based workflows, robotic process automation (RPA), system integration, low-code/no-code applications, document intelligence, AI, and custom development. Much like tools in a toolbox, the tool used should be the right one for the task at hand. The processes, risks, and outcomes should all be considered when matching the “tool” to the opportunity.
Organizations need to act with purpose and intent when considering automation. Applicable examples include:
- Rules-based automation for deterministic, repeatable work
- RPA for structured tasks on specific screens or websites, especially legacy systems
- Low-code applications for workflows involving intake, approvals, or fieldwork
- Machine learning for constrained tasks involving decisions, such as document extraction and classification
- AI for summarization, content analysis, and well-defined, guided multistep tasks
- Custom development for cloud services that require more detailed customization
There are levels of autonomy that exist as well. Rules-based automation, RPA, and even machine learning are at the bottom of the pyramid and widely used today. The next level involves partial autonomy, or multistep work for well-defined use cases, which is where a heavier prevalence of AI use may be apparent. High and full autonomy are the top levels, where full workflows and processes are handled without any human intervention. This last level is still considered aspirational in current day, but we are moving closer each and every day.
Guardrails are Part of the Design
AI can undoubtedly accelerate work, but organizations must account for several important constraints when working with it, such as hallucinations, limited context, contextual drift, non-deterministic outputs, model bias, and the absence of human judgement.
For public sector entities, many of these constraints are increasingly important to manage due to protected and public information, policy interpretations and effects, and/or consequential decision making.
Guardrails sound like a no-brainer, and they should be implemented alongside the design stage. Effective AI governance includes testing and evaluation, security reviews, context engineering, prompt engineering, human-in-the-loop review, training, change management, and trust and ethics.
Additional considerations to follow when implementing AI include:
- Using approved information sources and clearly defining the context of the solution
- Applying role-based security access and data protection controls
- Testing a wide range of expected and unexpected scenarios
- Retaining human review wherever decisions require judgement or carry higher risk consequences
- Creating audit trails for actions and approvals
- Monitoring the solutions and updating controls as the process changes and grows
An important rule of thumb is that AI may be the agent, but people must be the pilot.
Choosing the Right Process Before The Right Approach
The goal for an AI use case is to automate the processes where automation makes sense and brings value. Six criteria can be used to help evaluate candidates for automation:
- Repetitiveness. Does the process include the same steps performed again and again?
- Volume. Is the work performed frequently or by multiple people?
- Prone to error. Can manual mistakes create rework, delay, or additional risk?
- Scalability. Does the current method become harder to sustain as the process grows?
- Employee impact. Does the process consume employee capacity inefficiently?
- Cost savings. Does automation offer a meaningful opportunity to save or reallocate cost?
Moving From Opportunity to a Working Solution
A simple three-step process can help organizations move from opportunity to implementation:
- Identify: Ask questions and listen to understand. Define the problem, the desired outcome, the people affected, and the processes to be improved.
- Discover: Engage the people who perform the work and review the outcomes. Document systems, data, handoffs, timelines, decisions, and controls. Determine whether the proposed outcome is realistic and viable.
- Implement: Build, test, iterate, and deliver. In other words: build it, break it, fix it, send it. Review the adoption process and measure outcomes to determine whether the solution provides lasting value.
When starting the automation journey, it may be helpful to follow three principles: start small, win small, and fail fast. Begin with low-hanging fruit that can generate a quick win or an easily identifiable stopping point.
Real-World Automation Use Cases
Below are a few instances where automation made an impact for public sector entities.
- Example 1:Bank Reconciliation For a Local City Government
- Scenario: A local city government performed bank reconciliation through a highly manual daily process.
- Solution: RPA extracted general ledger data from a web-based enterprise resource planning (ERP) system, then combined it with a bank transaction file.
- Outcome: The department shifted from reviewing every transaction manually to only managing exceptions. The solution saved the team more than five hours a week.
- Example 2:Streamlined Digital Filing
- Scenario: A client monitored an email inbox for signed documents, downloaded them, and manually filed them into SharePoint.
- Solution: A Power Platform solution monitored the inbox, extracted information from signed document attachments, and automatically routed and sorted the files into SharePoint.
- Outcome: This saved an employee up to 520 hours annually for a projected 128% return on investment (ROI) in the first year.
- Example 3:Payroll Audit at Scale
- Scenario: A client needed to validate living wage rates under New York’s 421A Affordable Housing Program. Two full-time employees transcribed data into an Excel workbook every month from 40 to 60 contractors.
- Solution: A custom web application ingested timesheet data and used AI to identify inconsistencies or missing information.
- Outcome: The monthly validation process was reduced from 30 days to three.
Measure More Than Cost Savings
Cost savings and ROI should consist of hours saved, employee engagement, reallocation of capacity, scalability, and efficiency. The value of automation and AI is in allowing employees to focus on other opportunities and supporting growing workloads without significant staffing growth. Consistency and reduction in error follow.
Build Momentum Through Disciplined Wins
Public sector modernization doesn’t have to begin with a large, multi-year transformation. Organizations can start with one process, establish a baseline, apply appropriate controls and guardrails, and measure the outcome. A successful first project can create proof-of-concept, develop internal skillsets, improve trust in technology, and reveal further opportunities.
AI and automation are most effective when treated as tools and resources. Not every process has to be “intelligent.” Instead, AI and automation can be used to make work more connected, responsible, accountable, and easier to navigate.
How Forvis Mazars Can Help
If your organization is exploring how to modernize processes, improve data integration, or responsibly implement AI and automation, professionals at Forvis Mazars can help support your efforts.
Our AI Strategy & Integration services can help you:
- Assess automation and AI readiness across people, processes, data and technology
- Identify use cases where automation, AI, or other solutions can create measurable value
- Develop practical road maps aligned with organizational goals, risk considerations, and available resources
- Design responsible governance, security, and human-in-the-loop review structures
- Implement and scale solutions that improve efficiency, consistency, and insight
Connect with us today to discuss how your entity can take a practical, responsible approach to AI and automation modernization. Download our Public Sector Priorities for additional insights into embracing change and preparing for what’s next.
