Which Contract Vehicle Should You Buy AI On? A Decision Framework for Federal Buyers
There is no single best vehicle for buying AI. There’s a best vehicle for your timeline, dollar value, scope maturity, and socioeconomic goal posture — and most of the time it’s whichever pre-competed multiple-award contract you can already order off. Standing up a new open-market procurement is the slow, expensive default; ordering off an existing vehicle saves a contracting officer roughly 145 days and an estimated $37,000 per order. The move most buyers miss: you can set aside an order under nearly any of these vehicles for SDVOSBs — with no separate justification and no public posting — and earn full goal credit at vehicle speed.
This piece is a decision framework: the factors that should drive the choice, the one mechanism that combines speed with goal credit, and the real 2026 status of every major FAR vehicle (which matters, because the landscape is unusually unsettled this year). It stays within FAR vehicles. The standalone SDVOSB sole-source contract is one option in the framework, not the whole story — its mechanics live in the $5M sole-source how-to, the FAR Part 19 rewrite behind it in how the FAR Part 19 overhaul lets you direct-award AI, and year-end timing in year-end funds, fast. The broad picture is in the pillar.
Start here: the order-level SDVOSB set-aside
Before the vehicle tour, the single most useful thing to know. FAR 16.505(b)(2)(i)(F) lets a contracting officer (CO) set aside an individual order under a multiple-award vehicle for SDVOSBs — and FAR 16.505(b)(2)(ii) confirms no justification is needed for that set-aside, with no public-posting requirement either. FAR 8.405-5 provides the parallel authority for GSA Schedule orders and Schedule BPAs.
That single mechanism resolves the tension every buyer under goal pressure feels. You don’t have to choose between the speed of an existing vehicle and the SDVOSB goal credit your agency needs to chase the now-5% governmentwide target. You set aside the order to the SDVOSBs already on the vehicle, compete it among them (satisfying fair opportunity), and the award counts toward your SDVOSB goal — at order speed, with bounded protest exposure. If you take one thing from this article, take that.
What should drive your vehicle choice?
Run these factors in order.
Timeline. If you need award in weeks, order off a vehicle you already have access to. Schedule orders under FAR 8.4 are typically fastest — GSA has already determined Schedule pricing fair and reasonable, and Parts 6 and 15 competition rules don’t apply (FAR 8.404(d)). GWAC task orders under FAR 16.505 follow streamlined fair-opportunity rather than full source selection. A standalone open-market acquisition — including a FAR 19.14 SDVOSB set-aside or sole-source — adds synopsis, full evaluation, and (for sole-source) a justification, and is materially slower. For year-end deadlines, the practical speed hierarchy is: a BPA call, then a Schedule order with an 8.405-5 SDVOSB set-aside, then a GWAC task order, then a standalone sole-source (see year-end funds, fast).
Dollar value. Thresholds shape both your authorities and your protest exposure (current 2026 figures):
- ≤ $15,000 (micro-purchase): maximum flexibility.
- ≤ $350,000 (simplified acquisition threshold): small-business set-aside presumed under FAR 19.502-2.
- ≤ $5M services / $8.5M manufacturing: the FAR 19.1406 SDVOSB sole-source ceiling (mechanics in the $5M sole-source how-to). Set-asides themselves have no upper ceiling.
- $10M civilian / $35M DoD: the task-order protest bar lifts above these values.
- $50M+ AI programs: effectively require a proven multiple-award vehicle, usually broken into competed task orders.
Scope maturity. Well-defined scope → fixed-price order on MAS or a GWAC. Evolving AI scope (model not selected, metrics not baselined) → performance-based techniques (a statement of objectives plus a surveillance plan), which OMB M-25-22 explicitly encourages, and a vehicle that permits cost-reimbursement if development risk is genuinely shared (OASIS+ and VETS 2 both do). For the policy backdrop, see OMB M-25-21 reads like a buying spec.
Competition posture. FAR 16.505(b)(1) requires fair opportunity for task orders above the micro-purchase threshold; the only legal off-ramps are the exceptions at 16.505(b)(2)(i) — urgency, single source, logical follow-on, minimum guarantee, statute, and the (F) small-business set-aside. The (F) set-aside is the cleanest: it satisfies fair opportunity (competed among the SDVOSBs on the vehicle) while guaranteeing an SDVOSB award.
Access and protest risk. GWACs require a Delegation of Procurement Authority (typically 1–2 weeks from GSA, or use GSA Assisted Acquisition Services as your ordering CO); Schedules require none. On protest exposure, ordering off a vehicle is far safer than open-market: below $10M civilian / $35M DoD, the only protest ground is that the order exceeds the underlying contract’s scope, period, or value.
Which vehicle, then? The 2026 landscape
Status as of May 2026 — verify against the current eLibrary listing before you issue an RFQ, because several of these are moving.
GSA Multiple Award Schedule (FAR 8.4). Live, mature, broadest contractor base, no DPA. SIN 54151S (IT Professional Services) ran about $12.9 billion in FY2025 across roughly 5,150 contractors; SIN 518210C (Cloud and Cloud-Related IT Professional Services) about $1.8 billion across roughly 1,130. Schedule BPAs (FAR 8.405-3) are the right home for recurring AI work. Order-level SDVOSB set-asides available under 8.405-5.
GSA VETS 2. The only governmentwide SDVOSB-exclusive GWAC — every task order produces SDVOSB credit by definition. Ceiling raised to $6.1 billion; ordering through February 22, 2028; 0.75% access fee. The cleanest path for SDVOSB IT-services AI work, full stop.
GSA OASIS+. Services MAC, $50 billion, ten-year period, six socioeconomic pools including a fully active SDVOSB pool, across 13 service domains after the January 12, 2026 continuous-open-enrollment expansion. Best fit for professional-services AI (data strategy, AI governance, M-25-22 advisory). Order from the SDVOSB pool and credit follows automatically.
VA T4NG2. The VA’s anchor IT vehicle — $60.7 billion ceiling, 30 awardees with 15 reserved for SDVOSBs. First choice for VA AI and health-IT modernization.
GSA 8(a) STARS III. $50 billion; base ordering period through July 1, 2026, with a three-year option to July 1, 2029; directed sole-source orders up to $4 million; an explicit Emerging Technology sub-area covering AI, ML, and RPA. (8(a), not SDVOSB — relevant if your goal posture is 8(a).)
GSA Alliant 3. Phase 1 Notice to Proceed issued March 10, 2026 — 43 awards from 133 proposals, with up to 76 awardees planned, no contract ceiling, primary NAICS 541512. Strong for large, complex AI integration once fully populated; confirm awardee capability before relying on it.
GSA Polaris. Small-business GWAC for IT and emerging tech, with an active SDVOSB pool (plus HUBZone and WOSB pools). An emerging option for SDVOSB AI work as ordering ramps up.
NASA SEWP V / VI. SEWP V extended through September 30, 2026 (options potentially to April 30, 2027); SEWP VI awards are delayed by GAO protests, so plan to use SEWP V into late 2026. Best for hardware-heavy AI infrastructure (GPU, compute) rather than services.
NIH CIO-SP4 — cancelled. NITAAC cancelled the $50 billion CIO-SP4 solicitation in January 2026; CIO-SP3 is extended through April 29, 2027, with GSA absorbing the portfolio. Don’t plan new work around CIO-SP4.
What about just buying AI tools?
For commodity AI — model seats, copilots, evaluation environments — start with USAi (no cost) for piloting, then move to GSA’s OneGov enterprise AI agreements for rollout, layered through your existing MAS access. As of May 2026, GSA reported more than 120 orders placed against OneGov AI offerings, reaching roughly 3.4 million users governmentwide. These sit on top of FAR vehicles (the underlying SINs are on MAS); they don’t replace the vehicle decision for custom AI development.
M-25-22 follows the work, not the vehicle
One point that trips buyers up: the AI-specific contract terms required by OMB M-25-22 (testing and monitoring rights, vendor lock-in mitigation, data-use restrictions, IP scoping, a U.S.-developed preference, disclosure of unanticipated vendor AI use) apply to AI acquisitions regardless of which vehicle you use. They attach to the order, not the contract. Pick the vehicle for speed and competition posture, then build the M-25-22 terms into the SOW or task order — GSA’s AI-acquisition repository, live since early 2026, provides sample clauses. The vendor-side counterpart — what to verify on the FedRAMP and security posture before award — is in how to verify an AI contractor’s FedRAMP posture.
What’s different at the VA?
VA-specific. Under the Supreme Court’s Kingdomware decision and VAAR 808.405-70, a VA contracting officer must run a Rule-of-Two analysis before any Schedule order, AI orders included — so the SDVOSB-first question is already built into the VA’s Schedule workflow, not optional. The VA’s tiered evaluation methodology (upheld in Land Shark Shredding) is the operational mechanic, and T4NG2 with its 15 SDVOSB reserves is the natural anchor vehicle.
The decision, mapped to scenarios
| Scenario | Recommended path |
|---|---|
| Defined AI scope <$5M, SDVOSB partner identified, no vehicle access | Standalone FAR 19.1406 SDVOSB sole-source. If the partner holds VETS 2 or STARS III, a single-source or directed order off that vehicle is faster. |
| Scope $5M–$10M, want SDVOSB credit | Order-level SDVOSB set-aside on MAS, OASIS+, Alliant 3, SEWP V, Polaris, or T4NG2 under 16.505(b)(2)(i)(F) or 8.405-5. Below the $10M/$35M protest bar, exposure is bounded. |
| Scope >$10M or mission-critical infrastructure | Compete a task order under VETS 2 (SDVOSB-exclusive), OASIS+, Alliant 3, or T4NG2, with full M-25-22 terms. Plan for protest exposure. |
| Recurring AI services (monitoring, retraining, evaluation) | A Schedule BPA under FAR 8.405-3, ideally set aside for SDVOSBs at the BPA level, with fixed-price calls. |
| Commodity AI seats and tools | USAi to pilot, OneGov to scale. |
For the credentialing argument behind any of these paths, see SDVOSB is leverage. Use it.
Frequently asked
Can I direct an order to an SDVOSB without a sole-source justification?
What's the fastest vehicle?
Is CIO-SP4 still an option?
Which vehicle gives the cleanest SDVOSB credit?
Do M-25-22 AI clauses depend on the vehicle?
Working with Truvisory
The fastest compliant path to award usually isn’t a new procurement — it’s the right order off the right vehicle, set aside correctly. Truvisory helps federal buyers scope AI work that maps cleanly to a FAR vehicle, including order-level SDVOSB set-asides under FAR 8.405-5 and FAR 16.505(b)(2)(i)(F), with M-25-22 terms built in from the start.
Verify it all in SAM.gov and at search.certifications.sba.gov. When you’re ready, start with a scoping call — and see the $5M sole-source how-to, year-end funds, fast for year-end timing, and the federal AI modernization pillar for the full path.