Artificial intelligence promises substantial improvements in productivity. Federal contractors are beginning to incorporate those expected efficiencies into their technical approaches, staffing plans, and cost proposals.
But can a contractor reduce proposed labor hours based on technology that has not yet been fully developed? A recent Government Accountability Office decision provides an important lesson.
In Applied Research Associates, Inc., B-424731 and B-424731.2, decided October 28, 2026 and publicly released November 12, GAO denied a protest involving a Defense Threat Reduction Agency procurement for engineering, modeling, testing, and software-development services.
The agency selected Leidos over Applied Research Associates (ARA), even though ARA received higher ratings for its technical and management approaches.
The difference in evaluated cost was significant. ARA's most probable cost was approximately $348.7 million, compared with approximately $331.9 million for Leidos.
The AI Productivity Question
Leidos proposed using several emerging technologies to improve performance efficiency, including generative AI and other automated analytical tools.
These technologies were intended to reduce labor requirements and improve production capabilities during performance.
However, DTRA initially questioned the maturity of those tools.
Some were still in early development and had not achieved initial operational capability. The agency concluded that meaningful implementation could require more than 18 months.
That created a problem with Leidos's initial proposed staffing assumptions.
The agency questioned whether labor reductions could realistically be achieved before the technology became operational.
In other words, the proposed savings depended on a capability that had not yet been demonstrated.
The Contractor's Response Matters
During proposal interchanges, Leidos revised its approach.
It explained that it would rely upon experienced personnel, established methods, and existing commercial and Government software tools while continuing to develop newer capabilities.
Importantly, Leidos removed its previously proposed staffing reductions at the 18-month point.
The new technologies would enhance future performance rather than become essential prerequisites for meeting existing requirements.
DTRA accepted that explanation and removed the associated technical weakness.
However, the agency did not simply accept every proposed labor estimate.
For research-and-development surge requirements, DTRA increased Leidos's evaluated staffing to reflect its estimate of the likely effort required.
That adjustment increased Leidos's most probable cost by approximately $18 million.
For other operations-and-maintenance surge requirements, the agency accepted Leidos's proposed approach as adequately supported.
GAO found the agency's technical and cost evaluations reasonable and denied ARA's protest.
What Federal Contractors Should Learn
The decision does not establish a universal rule governing AI-related labor reductions.
It does demonstrate that an agency may distinguish between proven productivity improvements and efficiencies dependent upon technology that remains under development.
Contractors incorporating AI into proposals should therefore identify which capabilities already exist, which require further development, and when each is expected to affect performance.
Staffing estimates should reflect credible transition assumptions.
Companies should also prepare a fallback approach explaining how contract obligations will be fulfilled if planned AI capabilities are delayed or fail to generate expected benefits.
That is particularly important for cost-reimbursement procurements, where agencies may independently assess the realism of proposed labor hours and adjust the estimated cost used for source selection.
The broader lesson is straightforward.
AI can be a legitimate source of competitive advantage, but projected productivity must be supported by an executable performance plan.
A compelling technological vision does not substitute for credible staffing, cost, and schedule assumptions.
Recommended FedContractPros Product: GovCon BoE Blueprint™
The GovCon BoE Blueprint™ helps contractors develop defensible labor estimates, productivity assumptions, staffing plans, cost models, and documented bases of estimate when proposing innovative technical solutions.
Disclaimer:
This article is provided for general informational and educational purposes only and does not constitute legal, pricing, accounting, artificial-intelligence, or government-contracting advice. The GAO decision concerns a particular solicitation and evaluation record. Cost realism, staffing assumptions, technical risk, and evaluation requirements depend on the procurement, contract type, proposal representations, and applicable acquisition rules.