The right tool improves the speed, cost and outcome of any job; that’s why we drive nails with hammers and screws with screwdrivers and don’t swap tools for each (usually). Similarly, federal agencies should consider widening their focus on AI implementations beyond just generative AI to add other proven approaches — particularly deterministic AI, which offers distinct advantages for upgrading and improving federal IT systems.
Deterministic AI empowers federal agencies to eliminate persistent IT modernization challenges, slash support costs and lower total cost of ownership. It also fosters innovation by freeing funds to leverage new technologies and improve operations. These capabilities lead directly to tangible benefits for taxpayers, such as enhanced efficiency and competitiveness, reduced risks and increased ability to innovate even further.
Combining multiple AI approaches ensures agencies have the right tools to get the job done when it has never mattered more. Agencies in 2025 are under extreme pressure to demonstrate their value. AI is an omnipresent buzzword as a potential panacea to improve speed and efficiency, augment operations, and cut costs.
With stakes this high, agencies would do well to remember AI is not monolithic — many approaches exist, each with its own strengths. Moreover, understanding which AI approach best suits particular needs is essential for agencies to successfully modernize their IT systems, offer innovative new services, and continue to serve the American people.
Two different approaches
Generative AI deserves praise as a go-to approach to AI for doing many things well, such as producing human-like written prose from large volumes of disparate information. It shines at creating documentation and training materials or summarizing a year’s worth of interactions for an annual performance review, for instance. But if you have a tight deadline to translate two million lines of COBOL into Java for a mission-critical IT system, deterministic AI is the way to go.
Generative AI assembles new content based on mathematical probability, meaning the system doesn’t always give the same output to a given input, and sometimes it hallucinates or provides incomplete or misleading information. This is the reason we review every piece of content generated by Copilot or ChatGPT for its accuracy and contextual applicability. Similarly, many new AI code conversion tools that depend solely on LLMs fail miserably at modernizing complex code. This is where deterministic AI comes to rescue.
Deterministic AI is designed for consistency, accuracy and security. Deterministic AI focuses on semantics — understanding the original intent behind existing code and precisely replicating it in new code. It’s like expert human developers ensuring outputs work exactly as intended, every time. In that respect, deterministic AI’s strengths play directly to the mission needs of federal agencies looking to modernize and enhance their IT systems.
Made for IT modernization
Code built or modernized with deterministic AI excels at repairing software errors, resolving security vulnerabilities, and preventing data breaches. It is more maintainable and auditable, making it more reliable and secure for critical operations.
Deterministic AI helps streamline automation and futureproof systems by baking in the ability to easily update them to meet evolving technologies and requirements. One of its greatest boons is significantly accelerating the IT modernization process, replacing outdated systems with modern architecture in months, not years.
Deterministic AI provides a long-awaited suite of capabilities to tackle one of the most intractable challenges in federal IT: modernizing legacy applications. These systems can be frustratingly hard to manually integrate into single systems, especially as they often are decades old and lack documentation and subject matter experts (SMEs) to explain how they work.
This situation often leads to a “Don’t touch it!” attitude toward aging mission-critical systems, out of fear of breaking irreplaceable relics while attempting to upgrade them. Meanwhile, those systems’ drawbacks continue to waste valuable time, money and opportunities for improvement.
Three months to success
Deterministic AI overcomes these obstacles by understanding the intent across multiple applications — either in one agency or across many — and discovering what needs to happen so things keep working and don’t break. It then rationalizes the myriad applications into a single modern application.
Case in point, the U.S. Air Force in 2024 applied deterministic AI to upgrade its web application framework from outdated Angular JS to the modern Angular framework. The project required fast, secure, error-free conversion of old code into new code — requirements tailor-made for deterministic AI.
The Air Force completed a prototype in only three months without any available documentation or SME involvement. The prototype modernized their legacy system and empowered strategy-to-execution planning, enhancing the efficiency of mission-critical operations. That success has encouraged the Air Force to actively explore expanding its use of deterministic AI to modernize other applications.
Looking ahead
To get the right AI tools to nail delivery of mission-critical capabilities, federal agencies should:
- Know what they need: Leaders should review their programs and the technical viability of available technologies — whether deterministic AI, generative AI or one of the many other types of AI — to most efficiently deliver envisioned capabilities and outcomes.
- Look in the right place: Accelerated IT modernization is not just about code, it requires expedited procurement as well. History abounds with projects in limbo from procurements taking years. Fortunately, the U.S. Department of Defense’s Tradewinds solutions marketplace portal is dedicated to cutting red tape and rapidly putting vetted IT solutions, including AI, where they can do the most good. The Air Force leveraged Tradewinds to award the contract for its deterministic AI-enabled prototype.
- Find the right partner: Agencies should look for capabilities such as semantic understanding of code, ability to repair errors and resolve security vulnerabilities, and comprehensive support for any language across any stack. They should also assess vendors’ experience and past performance to ensure optimal fit and results.
It’s never been more urgent or important for federal agencies to demonstrate they can efficiently provide continually improving services at lower cost. Integrating AI, especially deterministic AI, will help federal agencies deliver not just on the promise of AI, but their own promise to serve the American people.
Rajiv Gidadhubli is chief transformation officer at Macro Solutions, a division of Alpha Omega, a leading provider of AI and other advanced technology solutions for the federal government.
Copyright
© 2025 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.