Aubrey Vaughan, vice president of government strategy at Celonis, explains why DoD can’t just lather AI over top of legacy systems to improve financial audits.
April 28, 2026 3:22 pm
4 min read
Artificial intelligence is becoming Washington’s favorite shortcut.
But no algorithm can repair a process that was broken long before the model was trained.
The federal government is racing to adopt advanced AI tools while roughly 80% of its IT systems still run on legacy infrastructure. Some are more than 50 years old. That is not modernization. That is bolting a Ferrari engine onto a 1970s transmission and hoping it holds.
Recent coverage of the Pentagon’s early AI efforts underscores the risk. Leaders pushed advanced analytics into environments defined by siloed systems, inconsistent data and manual workarounds. The technology was sophisticated. The foundation was not. The result was predictable.
]]>
Across government, agencies are learning the same lesson. The Energy Department and the FBI are modernizing their approach to AI only after confronting the limits of fragmented legacy systems. Public sector AI initiatives routinely stall for one simple reason: They try to automate complexity without first understanding it.
For defense leaders, this is not just a technology issue. It is a readiness issue.
When procure-to-pay transactions are split across disconnected systems, obligations go unmatched. When journal vouchers bypass controls, reporting risk increases. When disbursements fail to reconcile with source documentation, audit findings pile up. Layer AI on top of that environment and you do not get transformation. You get faster confusion.
The Pentagon’s audit challenges are not primarily data problems. They are process visibility problems.
CFOs and inspectors general already know the pain points. Duplicate vendors. Split payments. Late approvals. Expiring funds. Journal vouchers posted after period close. Segregation of duties violations buried in manual workflows. Unmatched disbursements that linger for months.
These are not edge cases. They are systemic symptoms.
This is where process intelligence changes the conversation.
]]>
Before deploying AI to detect anomalies, agencies must map how transactions actually flow across enterprise resource planning (ERP) and feeder systems. They must reconcile mismatches in real time. They must establish a complete, time-stamped audit trail for every obligation, approval, modification and disbursement.
The answer to this problem is process mining, a data-driven technology that analyzes digital footprints from business systems to create a real-time, objective view of how processes actually run. With that foundation, organizations can move from reactive oversight to proactive control. It surfaces improper payments before they become findings, flags expiring funds before they lapse, identifies orphaned assets and unsupported journal vouchers before auditors do, and monitors segregation of duties violations as they occur. In short, it enables continuous control monitoring instead of periodic panic.
Only then does AI become powerful.
When machine learning models are applied to reconciled, contextualized transaction data, anomaly detection becomes meaningful. Continuous auditing becomes achievable. Root cause analysis replaces repetitive corrective entries. Audit readiness becomes a byproduct of daily operations rather than a once-a-year fire drill.
The defense community does not lack AI ambition. It lacks operational transparency at scale.
If we are serious about strengthening national security and safeguarding taxpayer dollars, we cannot treat AI as a silver bullet. It is a force multiplier. But multipliers only work when there is something solid to multiply.
The path forward is not AI or modernization. It is disciplined process reform paired with intelligent automation. The following are three practical steps the Pentagon can take immediately:
Create a single, enterprisewide view of how work actually happens.
Break down system and organizational silos to establish a unified, real-time understanding of financial workflows across the department.
Shift from compliance exercises to operational discipline.
Move beyond periodic reviews and check-the-box controls toward embedded, day-to-day accountability within core processes.
Institutionalize continuous improvement at the process level.
Focus less on correcting individual errors and more on identifying and fixing the underlying process breakdowns that create them. By embedding feedback loops into operations, the department can reduce recurring issues, strengthen financial integrity and build resilience over time.
Until the Pentagon can see how work truly happens across its financial and operational systems, no algorithm will deliver the efficiency gains Washington promises.
At the moment, AI will not fix broken processes. But with the right foundation, it can finally help the defense community move from reactive remediation to proactive, mission-ready execution.
]]>
Aubrey Vaughan is vice president of government strategy for the public sector at Celonis, where he helps drive government modernization through process intelligence. He brings more than 25 years of experience in advanced technologies, including leadership roles supporting the Department of Defense and intelligence community.
Copyright
© 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.