# Writing a Determination and Findings for AI Services > Source: https://ibl.ai/resources/guides/writing-a-determination-and-findings-for-ai-services > Last updated: 2026-08-19 *The D&F must establish that no other contract type is suitable — a test that AI acquisitions increasingly fail, because the platform portion can now be bought* Reading time: 11 min read | Difficulty: advanced **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 writing a Determination and Findings for AI Services? Before a time-and-materials or labor-hour award, the contracting officer signs a Determination and Findings establishing that no other contract type is suitable. FAR requires it to describe the market research conducted, establish that it is not possible to accurately estimate the extent or duration of the work or to anticipate costs with reasonable certainty, and address why a cost-plus-fixed-fee term, other cost-reimbursement, incentive, or fixed-price arrangement is not appropriate. For AI services this document has quietly become harder to write well, and the reason is not legal. It is that the underlying market changed faster than acquisition templates did. A D&F asserting that AI capability cannot be estimated is defensible when the contractor must build a platform. It is considerably weaker when a production platform can be licensed with source rights, because then the unestimable portion shrinks to integration against named systems — which is exactly the kind of work a fixed-price arrangement handles. This guide covers what the document must contain, the specific test to apply for AI, and the market-research step that dates faster than any other part of an acquisition plan. ## Prerequisites - **Current market research, not inherited research:** For AI infrastructure, research more than roughly two years old is likely to be wrong about what can be bought rather than built. - **The requirement separated into platform and integration:** These have completely different estimability profiles, and merging them is what makes an unestimable-work assertion look true. - **An independent government estimate of the integration portion:** If integration against named systems can be estimated, the fixed-price alternative has to be addressed seriously. - **Your software and data rights position:** Contract type and rights are separate decisions, and the D&F is a good forcing function for settling the second. ## Step 1: Conduct market research on the platform portion specifically The most common defect in an AI D&F is market research conducted at the level of 'AI services' rather than at the level of the platform capabilities the requirement actually needs. - [ ] Enumerate the platform capabilities the requirement implies — Retrieval, evaluation, guardrails, access control, audit logging, model routing, agent orchestration. - [ ] Identify existing products that provide them, including source-available options - [ ] Record deployment constraints as evaluation criteria — On-premise, GovCloud, air-gapped operation. - [ ] Document the date of the research — This field matters more here than in most acquisitions. ## Step 2: Apply the estimability test to each portion separately The standard is that it is not possible to accurately estimate the extent or duration of the work. Applied honestly, that is often true of building a platform and often false of integrating one. - [ ] Platform construction: can the effort be estimated? — Usually not, which is why T&M is reached for. - [ ] Integration against named systems: can the effort be estimated? — Usually yes, given a defined endpoint list. - [ ] State the conclusion for each portion rather than for the acquisition as a whole **Tips:** - If the answer differs by portion, the acquisition strategy should differ by portion too. A hybrid structure is frequently the defensible outcome. ## Step 3: Address each alternative contract type on its merits The requirement is to explain why cost-plus-fixed-fee, other cost-reimbursement, incentive, and fixed-price arrangements are not appropriate. Generic language here is what makes a determination fragile on review. - [ ] Firm-fixed-price: address it against the platform-licence option explicitly - [ ] Cost-plus-fixed-fee: address the oversight burden it carries - [ ] Incentive arrangements: address whether an objective measure exists — For AI, an evaluation-set pass rate often supplies one. - [ ] Avoid boilerplate; a reviewer can recognise it immediately ## Step 4: Set the ceiling from the estimate, and secure approvals The ceiling belongs to the determination's logic. A ceiling derived from available budget contradicts the assertion that the work was carefully assessed. - [ ] Derive the ceiling from the independent estimate - [ ] Confirm contracting officer signature before executing the base period - [ ] Obtain head of contracting activity approval where base plus options exceeds three years - [ ] Document the surveillance approach the contract type requires ## Common Mistakes ### Reusing the market research section from a prior acquisition **Consequence:** The determination rests on a description of a market that no longer exists. **Prevention:** Re-run research against the specific platform capabilities the requirement implies, and date it. ### Asserting unestimability for the acquisition as a whole **Consequence:** A finding that is true of one portion and false of another, which is exactly what a reviewer will separate. **Prevention:** Assess platform construction and integration independently and state both conclusions. ### Dismissing fixed-price with boilerplate **Consequence:** The alternatives analysis reads as pro forma and weakens the entire determination. **Prevention:** Address firm-fixed-price against the platform-licence option specifically. ### Setting the ceiling to the budget line **Consequence:** The document asserts careful assessment while the ceiling shows otherwise. **Prevention:** Derive the ceiling from the independent estimate and show the derivation. ## FAQ **Q: What must a Determination and Findings contain for a T&M award?** A description of the market research conducted, an establishment that it is not possible to accurately estimate the extent or duration of the work or anticipate costs with reasonable certainty, and an explanation of why cost-plus-fixed-fee, other cost-reimbursement, incentive, and fixed-price arrangements are not appropriate. **Q: Who signs and approves the D&F?** The contracting officer signs it before execution of the base period or any option periods. Where the base period plus options exceeds three years, the head of the contracting activity approves it before the base period is executed. **Q: Why is a D&F harder to write for AI now?** Because production AI platforms can be licensed with source rights, which shrinks the unestimable portion to integration against named systems. That makes an assertion that no other contract type is suitable considerably harder to sustain. **Q: Can an AI acquisition use a hybrid structure?** Frequently that is the defensible answer. Estimability often differs by portion — construction is hard to estimate, integration against a named endpoint list is not — so the strategy can differ by portion too. **Q: How do you measure AI performance objectively enough for fixed-price?** With a held-out evaluation set drawn from agency data and an agreed passing threshold. That supplies the objective measure usually said to be missing when incentive and fixed-price arrangements are dismissed. **Q: How does ibl.ai fit in?** ibl.ai is a production platform available under a perpetual licence with the full source code, deployable on-premise, in GovCloud, or fully air-gapped. That changes what market research finds, and generally makes a firm-fixed-price structure available for the platform portion. 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.