# Government AI Contract Vehicles and What They Don't Decide

> Source: https://ibl.ai/resources/guides/government-ai-contract-vehicles-mas-oasis-stars
> Last updated: 2026-08-19


*MAS, OASIS+ and the GWACs determine how quickly you can award — not whether the agency ends up owning what it paid for*

Reading time: 10 min read | Difficulty: intermediate

**On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing — so you can deploy anywhere, from your own cloud to a fully air-gapped network.**

## How do you government AI Contract Vehicles and What They Don't Decide?

Federal agencies rarely buy AI through an open competition. They buy through pre-competed vehicles — the GSA Multiple Award Schedule, OASIS+, and IT GWACs such as 8(a) STARS III — structuring the purchase as a task order or a Blanket Purchase Agreement.

That choice determines lead time, competition pool, socioeconomic credit, and fee structure. It is a genuine acquisition decision and it is usually made carefully.

What the vehicle does not determine is the thing that decides whether the programme succeeds: what the agency holds when the period of performance ends.

Every vehicle discussed here can carry either arrangement — a services engagement that constructs a platform the contractor retains rights to, or a licence to a platform that already exists with source rights transferring to the agency. The vehicle is silent on that. It is set by the requirement and the clauses, and it is the question most often left to the default.

## Prerequisites

- **Your vehicle access and its scope:** Confirm which vehicles the agency can order from and whether AI services sit within scope, rather than assuming from prior use.
- **The requirement separated into platform and integration:** This distinction drives both the contract type and the rights position, and it is invisible if the requirement is written as a single services statement.
- **A stated software and data rights requirement:** Decide what the agency must be able to operate, modify and inspect independently before the solicitation is drafted.
- **Deployment constraints as requirements:** On-premise, GovCloud, or air-gapped operation eliminates whole categories of offering and belongs in the requirement, not in evaluation.

## Step 1: Pick the vehicle for speed and competition, not for outcome

Vehicle selection is about award timeline, competition pool, and fee. Treating it as the substantive decision is how requirements end up under-specified.

- [ ] MAS for commercial products and services with established SINs
- [ ] OASIS+ for complex professional services requirements
- [ ] GWACs such as 8(a) STARS III for IT services with socioeconomic goals
- [ ] Consider a BPA where recurring ordering is expected

**Tips:**
- A vehicle shortens the path to award. It does not improve a requirement, and it cannot supply a rights position the requirement omitted.

## Step 2: Write the rights position into the requirement

This is the step that determines what the agency owns, and no vehicle supplies it by default. It is also far cheaper to state up front than to negotiate at closeout.

- [ ] State whether source-code access is required, and at what level
- [ ] State whether the agency must be able to modify and redeploy independently
- [ ] Address ownership of custom developments explicitly
- [ ] Address data — inputs, outputs, embeddings, logs — and its deletion

## Step 3: Choose a contract type per portion

Vehicles carry multiple contract types. The estimability of platform construction and of integration usually differs, and the structure can reflect that.

- [ ] Firm-fixed-price for a platform licence against a defined deliverable
- [ ] Firm-fixed-price or capped effort for integration against named endpoints
- [ ] T&M only for the genuinely exploratory remainder, with a D&F

**Tips:**
- Splitting the structure by portion is usually easier to justify than a single T&M award covering everything.

## Step 4: Evaluate on the things that persist

Evaluation criteria drive proposals. Criteria that reward staffing depth get staffing depth; criteria that reward delivered capability and rights get those instead.

- [ ] Demonstrated production deployments matching your deployment constraint
- [ ] Rights offered, evaluated as a discriminator rather than a checkbox
- [ ] Independent operability at closeout, with an acceptance test
- [ ] Model provenance and the ability to change models later

## Common Mistakes

### Treating vehicle selection as the substantive decision

**Consequence:** A fast award of an under-specified requirement, which does not shorten time to capability.

**Prevention:** Spend the effort on the requirement and the rights position; vehicles are the easy part.

### Leaving rights to the default clauses

**Consequence:** The agency funds a system it cannot independently modify or redeploy.

**Prevention:** State the rights position in the requirement and evaluate it as a discriminator.

### One contract type for the entire award

**Consequence:** T&M covering work that could have been fixed-price, or fixed-price covering work nobody can estimate.

**Prevention:** Assess estimability per portion and structure accordingly.

### Evaluating deployment constraints instead of requiring them

**Consequence:** Proposals that cannot meet an air-gap or residency requirement consume evaluation time and sometimes win.

**Prevention:** Put binding deployment constraints in the requirement.

## FAQ

**Q: How do federal agencies buy AI services?**

Predominantly through pre-competed vehicles — the GSA Multiple Award Schedule, OASIS+, and IT GWACs such as 8(a) STARS III — structured as a task order or a Blanket Purchase Agreement rather than as an open competition.

**Q: Which contract vehicle is best for AI?**

The vehicle should be chosen for award timeline, competition pool and fee, because all of them can carry AI work. It does not determine what the agency ends up owning, which is set by the requirement and the clauses.

**Q: Does the contract vehicle determine data and software rights?**

No. Every vehicle discussed can carry either a services engagement in which the contractor retains rights or a licence with source rights transferring to the agency. The requirement decides, and defaults apply when it is silent.

**Q: Can an agency buy an AI platform licence through MAS?**

Yes, where the offering is on schedule under an appropriate SIN. A firm-fixed-price licence against a defined deliverable is a preferred contract type and avoids the justification burden that T&M carries.

**Q: How should AI proposals be evaluated?**

On demonstrated production deployments matching your deployment constraint, the rights actually offered, independent operability at closeout, and model provenance including the ability to change models later. Staffing-weighted criteria produce staffing-weighted proposals.

**Q: How does ibl.ai fit in?**

ibl.ai is available as a perpetual licence with the full source code, deployable on-premise, in GovCloud, or fully air-gapped, with forward-deployed engineers handling integration as bounded work. You own all the code and the data, run it model-agnostic across any LLM, with no per-seat pricing.


## Can you do this on infrastructure you own?

**ibl.ai is the agentic AI platform where you own all the code and the data. You self-host the entire stack inside your own perimeter, run it model-agnostic across any LLM and switch anytime, and pay by usage with no per-seat pricing — so you can deploy anywhere: your cloud, on-premise, GovCloud, or fully air-gapped.**

- **You own all the code and the data.** Full source code under a perpetual license, running on your infrastructure. Not API access to someone else's platform — the stack itself is yours.
- **Model-agnostic.** Run any LLM — Claude, GPT, Gemini, Llama, Command, or your own fine-tune — and switch providers without rewriting the platform.
- **No per-seat pricing.** Usage-based billing against a budget cap you set. Cost tracks what your organization actually uses, not how many people you employ.
- **Deploy anywhere.** Your cloud, your VPC, on-premise, GovCloud, or a fully air-gapped network with no outbound connectivity.

1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.

ibl.ai is family-owned and operated from New York, NY — a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on.