AI Government Contracts: How Small Businesses Win Federal Work in Artificial Intelligence

The scale of federal AI investment is not hype — it's documented in obligation data. According to a 2026 Brookings analysis of federal procurement records, AI contract obligations grew from $355 million in 2024 to $7.2 billion in 2026. That's a 966% increase in two years. The potential award value — what agencies have authorized they may eventually spend — climbed from $4.6 billion to $91.8 billion over the same period, a 1,912% jump.

The number of agencies actively buying AI grew from 17 in 2022 to 28 by 2026. That expansion means AI isn't limited to defense and intelligence anymore. Health, energy, transportation, and financial regulatory agencies are all adding AI lines to their procurement portfolios.

The policy foundation driving all of this is the White House AI Action Plan, released July 23, 2025. The plan set out more than 90 federal policy actions across three pillars: accelerating AI innovation, building U.S. AI infrastructure, and leading in international AI security. Crucially, one pillar specifically targets federal procurement — agencies are directed to buy American AI and to streamline acquisition processes for AI solutions. On June 2, 2026, President Trump reinforced this with an Executive Order titled “Promoting Advanced Artificial Intelligence Innovation and Security,” which added cybersecurity mandates for AI deployments and a voluntary framework for frontier model security. These aren't advisory documents. They fund programs and generate solicitations.

The Department of Defense dominates this market by a wide margin. The DoD grew from 254 AI contracts in 2022 to 657 in 2024 and 1,319 in 2026. The potential award value sitting with DoD is roughly $90 billion. Within DoD, the priority use cases are intelligence processing and targeting, autonomous systems, logistics and predictive maintenance, cybersecurity and threat detection, and AI-assisted mission planning for missile defense.

The Department of Health and Human Services runs a distant second with 134 AI contracts, focused on clinical data analysis, fraud detection in Medicare and Medicaid, and public health surveillance. NASA follows with 71 contracts, primarily for remote sensing data processing, mission modeling, and earth observation analytics. The Department of Commerce adds $197 million in AI contract volume, driven by NOAA weather modeling and Census Bureau data systems.

What are agencies actually buying? The contract categories fall into a few buckets:

The distinction between “AI product” and “AI services” matters for how you bid. Agencies buying a software tool — a pre-built AI platform — tend to use IT contract vehicles and simplified acquisition. Agencies buying custom AI development — building a model for their specific mission data — typically use R&D vehicles, OTAs, and SBIRs. Your approach depends on which type of work you do.

Your NAICS codes determine which solicitations you see and whether you qualify as a small business for set-aside competition. Most AI contractors need more than one code. Here are the primary codes and their 2026 SBA size standards.

541512 is the strongest primary code for most AI firms because contracting officers associate it with integrated technology solutions, and it's the code written into the most AI-related solicitations. If your work skews heavily toward building custom algorithms and models from scratch, add 541511 as a secondary code. If you offer a SaaS AI platform, 518210 covers that work stream.

Register all relevant codes in SAM.gov — agencies search by NAICS when setting solicitations, and you can miss opportunities if you're not registered under the code the CO used to classify the work. Use CapturePilot's opportunity matching to monitor which codes are driving the most AI solicitations in your target agencies. For guidance on updating your codes without losing set-aside eligibility, see our post on changing NAICS codes in SAM.gov.

Most federal AI spending does not flow through open-market SAM.gov solicitations. It flows through pre-competed contract vehicles — GWACs, IDIQs, and schedules — where agencies issue task orders only to pre-approved vendors. Getting on the right vehicles is the strategic play for long-term AI revenue.

GSA 8(a) STARS III

The 8(a) STARS III is a Best-in-Class Governmentwide Acquisition Contract (GWAC) for IT services, and it explicitly includes AI, machine learning, and data science as task areas. It's restricted to 8(a)-certified small businesses. If your firm is 8(a) certified, STARS III should be a top priority — it has a $50 billion ceiling and is the preferred vehicle for many civilian agencies buying AI services. The ordering period runs through 2030. If you're not on it yet, monitor GSA's announcements for on-ramp opportunities.

Alliant 3

GSA issued Alliant 3 Phase 1 awards in March 2026 — this is the next-generation IT services GWAC replacing Alliant 2. Unlike Alliant 2 (which closed to new vendors), Alliant 3 has both large business and small business tracks. The vehicle explicitly covers AI, cloud, cybersecurity, and digital transformation. GSA announced an on-ramp process for new vendors; watch SAM.gov and the GSA GWAC program pages for upcoming small business on-ramp announcements. Alliant 3 will be the dominant IT vehicle for the next decade.

NIH CIO-SP3 Small Business

The CIO-SP3 Small Business vehicle, administered by NIH, covers health IT and data science work across civilian agencies. It's been extended through October 2026 for task order ordering. The CIO-SP4 successor is in development — watch for solicitation announcements if you work in health AI, clinical informatics, or federal data systems.

GSA Multiple Award Schedule (MAS)

GSA's MAS program — Schedule 70 equivalent — is the easiest vehicle to get on and covers AI products and services under Large Category IT. GSA Refresh #31 (early 2026) introduced new AI-specific contract terms and expanded Transactional Data Reporting requirements for AI schedule holders. Getting a GSA Schedule for your AI offerings is the foundational step before pursuing GWACs. It takes 90–120 days and gives you access to the entire civilian agency market without competing for a spot on every vehicle.

Other Transaction Authority (OTA)

OTAs have become a primary tool for DoD AI acquisitions because they bypass standard FAR procurement rules. Consortia like NSTXL, NCMS, and AFWERX manage OTA vehicles that give small businesses direct access to DoD AI opportunities. Joining one of these consortia costs $500–$5,000 annually and can put you in front of program offices that never post openly on SAM.gov. For more on OTAs, see our complete OTA guide.

Set-aside certifications reduce your competition from thousands of vendors to a qualified pool of dozens. For AI contracts specifically, four certifications create the most opportunity.

For AI work specifically, the 8(a) certification is the most powerful because of 8(a) STARS III — a $50 billion GWAC with AI explicitly in scope. If you qualify for 8(a), apply before pursuing any other certification strategy. For a deep dive on each program, see our posts on 8(a) sole-source contracts and which certifications actually help you win.

Not sure which programs you qualify for? Use CapturePilot's Quick Checker to get a fast eligibility assessment based on your business profile.

If your AI firm is R&D-oriented — building novel models, doing applied research, or developing AI for a specific technical problem — the SBIR/STTR program is the best entry point into federal AI work. It's structured as grants, not contracts, which means you retain IP rights while getting paid to de-risk your technology.

The program had a significant disruption: SBIR/STTR authorization expired September 30, 2025, freezing new solicitations across all 11 participating agencies including DoD, NIH, NSF, DOE, and NASA. As of mid-2026, Congress reauthorized the program and DoD published its first new Broad Agency Announcements (BAA) — the 26.1 solicitation. If you were waiting on SBIR, it's back.

DoD releases three SBIR BAAs per fiscal year (numbered 26.1, 26.2, 26.3), each containing hundreds of topic areas from Air Force, Army, Navy, DARPA, MDA, and other components. AI, autonomy, machine learning, and computer vision topics appear in nearly every component's topic list. Phase I awards typically run $50,000–$250,000 for a 6-month feasibility study. Phase II awards run $750,000–$1.75 million for a 2-year development effort. Phase III is where the commercial and government contracts flow — often sole-source.

The key to winning SBIR topics is responding to the specific technical challenge described in the topic — not just describing your technology broadly. DoD program managers write the topics because they have a real problem. Show you understand the operational context, not just the algorithm. For a full breakdown of the SBIR process, see our SBIR and STTR grants guide.

Federal buyers are sophisticated enough to know that “AI” is not a differentiator anymore. Everyone claims AI. What they're evaluating is: do you understand my specific problem domain, can you prove you've solved problems like it before, and can you show your approach works with government data constraints?

Your capability statement and technical proposals need to address three things that generic AI marketing misses:

Your CapturePilot capability statement builder lets you create agency-specific versions that emphasize the right past performance and differentiators for each target buyer. For AI contractors, having separate versions for DoD, civilian health, and civilian civilian IT is worth the time investment.

Most small businesses lose their first few AI bids not because their technology is wrong, but because they misread the acquisition structure. These are the five patterns that show up repeatedly in debriefs.

Winning AI government contracts consistently is a pipeline problem. The agencies spending the most on AI — DoD components, HHS, NASA — plan their acquisitions 12–24 months before an RFP hits SAM.gov. If you're responding to RFPs cold, you're already behind the vendors who shaped the requirement.

Managing this pipeline — tracking opportunities, expiration dates, incumbent data, and agency relationships — is exactly what CapturePilot's pipeline management is built for. You can set alerts for Sources Sought notices in your NAICS codes, track incumbent contract awards, and score your probability of win on each pursuit before you commit proposal resources.