How government agencies can beat the odds on AI pilots

The government AI landscape shifted recently when the General Services Administration announced federal agencies could access various AI chatbots for just $1. This bold move drastically lowers the barrier for agencies — and in many cases, state, local, and tribal governments that can leverage GSA procurement vehicles — to test and deploy AI. It also reflects a larger trend: agencies are rushing to apply AI to everything from procurement and finance to citizen services.
But cheap and fast access to AI does not guarantee success. In fact, a recent MIT study found that 95% of AI pilots fail to deliver expected returns. For governments navigating tight budgets, workforce shortages, and growing public expectations, there is no margin for wasted investment. AI is too important to get wrong.
The problem isn’t the technology itself. AI too often fails in government because it’s deployed without clarity on the problems it is meant to solve. Data is inconsistent, processes are undocumented, and agencies assume that layering on an algorithm will generate value. Instead of correcting inefficiencies, AI ends up reinforcing them — locking in waste, frustrating staff, and fueling public skepticism.
When AI fails in government, it isn’t just a wasted pilot project. It means missed opportunities to improve service delivery, reduce costs, and rebuild trust in government. The stakes are too high for trial-and-error experiments.
Process Intelligence (PI) addresses this gap. Where most agencies rely on static process maps or assumptions, PI gives leaders a real-time, data-driven view of how workflows actually function. It uncovers bottlenecks, highlights hidden costs, and pinpoints where AI will create measurable return on investment.
Think of PI as a process “X-ray” that shows how work gets done today — and where reforms will stick tomorrow. It lets leaders test changes in a digital environment before making costly investments. By grounding AI adoption in this clear picture of operations, PI ensures that modernization efforts deliver real results.
The State of Oklahoma offers a powerful example. The Office of Management and Enterprise Services (OMES) has partnered with Celonis for over a year to examine procurement and financial processes. By digging into data and identifying inefficiencies, Oklahoma unlocked more than $10 million in value.
The reforms didn’t stop there. With this process foundation in place, the state introduced an AI Copilot that now gives procurement specialists and leadership quick, accurate insights into operations. Instead of spending days piecing together data, staff can make smarter decisions in real time. Together, PI and AI have streamlined workflows, reduced costs, and maintained service quality — all while protecting taxpayer dollars.
This isn’t just a story about Oklahoma; it’s a playbook for governments everywhere.
Every agency — federal, state, and local agency — faces the same modernization challenge: how to do more with less while meeting rising expectations for accountability and service. AI can help, but only if applied strategically.
- Procurement: Agencies can use PI to cut cycle times, reduce duplicative reviews, and then layer AI on top to recommend faster sourcing strategies.
- Fraud detection: PI helps agencies identify gaps in benefits delivery, giving AI the structured data it needs to spot fraudulent claims without adding unnecessary friction.
- Customer experience: Under Executive Order 14058, federal agencies are tasked with improving the customer experience. PI can highlight where services break down, while AI-enabled chatbots and digital assistants provide citizens faster answers — without creating new inefficiencies behind the scenes.
These are not futuristic scenarios — they are practical opportunities available now.
Why this matters now
With tools like GSA’s $1 chatbot making AI accessible across government, the risk is clear: We could see a surge of pilots that fail to scale, reinforcing the perception that AI is overhyped. That would be a mistake. AI is here to stay, and government agencies cannot afford to be left behind.
The lesson is simple: AI on its own will not deliver the reforms agencies need. But AI paired with Process Intelligence can.
A call to action
Now is not the time to abandon artificial intelligence. It is the time to implement it responsibly, strategically, and with taxpayer value in mind. Governments don’t have the luxury of trial-and-error experiments. With Process Intelligence as the foundation, agencies can cut waste, reduce risk, and ensure AI delivers lasting reform.
The public expects modernization. Taxpayers demand accountability. Leaders across federal, state, and local government now have the tools to deliver both.
AI is no longer optional — but without Process Intelligence, it’s unlikely to succeed.
Aubrey Vaughan is vice president of strategy & business development for public sector at Celonis.