Award and Recompete Intelligence from USAspending

Full report

On this page

Bottom line

The strongest bounded account-development candidates from USAspending award records are: Treasury/IRS modernization and cloud managed services around Accenture, CMS cloud modernization and identity management around ALCOR, VA data-platform work around Palantir, and HHS data-platform work around Palantir. A smaller but still credible AI-specific watchlist exists at USPTO around Accenture’s patent-search AI work, and an architecture-specific watchlist exists at VA OIG around Deloitte’s enterprise-architecture support. Every item below is an existing federal contract award or parent IDV in USAspending, not an open solicitation.

The practical answer to the commercial question is that repeat demand is clearest where an agency, a recognizable technical category, and an incumbent recur across multiple USAspending records or across a parent vehicle plus multiple orders. In this review, the repeat patterns are strongest in modernization, cloud, enterprise data and analytics, identity/access management, and architecture support rather than in one-off innovation buys.

Signal construction and guardrails

I used USAspending award and vehicle records surfaced from the official USAspending search/index, with emphasis on terms tied to modernization, cloud modernization, AI, software, data, reliability, identity, and architecture. USAspending’s keyword and award search tools are designed to expose federal contracts, grants, loans, and related award records, and the API documentation confirms that award search is one of the core public data functions.

For timing, I treated period-of-performance current end date and potential end date as recompete-monitoring signals, not as proof that a new solicitation is imminent. Treasury’s federal spending transparency definitions state that a procurement potential end date is the date the award would end if all pre-negotiated options were exercised, and that these dates can change as awards are modified.

For value, I used only deterministic public values that the official USAspending result exposed directly, usually the current award amount. I did not substitute ceilings or estimates when the official snippet did not expose them, and I mark those cases as unknown instead of backfilling from non-official mirrors. That matters because USAspending and FPDS expose multiple amount concepts, and contract reporting is not the same thing as future spend.

I also excluded grants, SAM notices, policy memos, and local/state contracts from the priority list below. The CRS overview on USAspending is a useful caution here: searches can be incomplete or contain inaccuracies, so these signals are best used for relationship research and monitoring, not for claiming that an opportunity is open.

Families that justify bounded investment

Treasury and IRS modernization and cloud managed services with Accenture Federal Services look like the strongest modernization family in this sample. USAspending shows a Treasury/IRS delivery order to Accenture, 2032H522F00497, for “IRS Enterprise Cloud Managed Service …” at $192,221,741, running from September 1, 2022 to September 30, 2025. A second Treasury/IRS order, 2032H522F00006, is still in progress with a current award amount of $49.7 million and a potential end date of November 20, 2026. Both sit under Accenture’s broader GSA vehicle 47QTCK18D0036, whose USAspending IDV page shows $2.4 billion in combined current award amounts, which is a strong repeat-demand signal rather than a single isolated modernization buy. Recurring categories: modernization, cloud managed services, systems design, software engineering. Provenance: USAspending award pages 2032H522F00497 and 2032H522F00006, plus parent IDV 47QTCK18D0036. Human-owned research action: assign [insert human name] to approve exactly 8 person-hours to map IRS CIO stakeholders, confirm whether Accenture’s base work is centralized or office-specific, and identify likely teaming gaps before any larger spend is authorized.

CMS cloud modernization and identity management with ALCOR Technical Solutions also clears the threshold. USAspending shows CMS order 75FCMC25FJ069 to ALCOR with a current award amount of $17.4 million and a potential end date of July 24, 2030. The parent vehicle GS35F595GA explicitly references ongoing support, modernization, and innovation for CMS identity, credentialing, and management work, and the same vehicle shows $65.5 million in combined current award amounts, including an earlier CMS order 75FCMC20F0048 under the same family. That is enough to treat CMS cloud modernization and ICAM work as a repeat account, not a one-time event. Recurring categories: cloud modernization, identity and access management, ongoing support and innovation. Provenance: USAspending order 75FCMC25FJ069 and parent IDV GS35F595GA. Human-owned research action: assign [insert human name] to approve 6 person-hours for a CMS buyer map covering the Office of Acquisition and Grants Management, the program-side identity/cloud owners, and any subcontracting footprint under the ALCOR family.

VA enterprise data and analytics platforms with Palantir are strong enough to justify direct monitoring. On Palantir’s GSA vehicle GS35F0086U, USAspending shows $1.37 billion in combined current award amounts. That vehicle includes VA order 36C10B21F0353 for the “Enterprise Data & Analytics Platform Project” at about $93.3 million to $93.6 million, with a current and potential end date of September 29, 2025. USAspending also shows a later VA order, 36C10B25F0313, as in progress, which suggests continuity of demand in the same buyer/recipient family even though the official result returned here did not expose its public value directly. Recurring categories: data platforms, analytics, SaaS/software. Provenance: USAspending parent IDV GS35F0086U and VA orders 36C10B21F0353 and 36C10B25F0313. Unknowns: the returned official snippet for the later order did not expose a current award amount, so I am not presenting one. Human-owned research action: assign [insert human name] to approve 8 person-hours to identify whether VA’s analytics platform work is centralized through TAC or split across mission offices, and to decide whether this is a platform-replacement, platform-expansion, or sustainment relationship play.

HHS data-platform demand with Palantir is a second Palantir family worth bounded investment because the repeat signals span multiple HHS components, not just one office. USAspending shows FDA order 75F40121F80001 for an “Enterprise Data & Analytics …” platform at $48,154,420, running from November 24, 2020 to February 1, 2024. Separately, the NCATS vehicle 75N95022D00025 shows $63.4 million in combined current award amounts, and the official USAspending snippet for that vehicle lists delivery order 75N95023F00007 at $7.7 million from September 28, 2023 to September 27, 2024 for “NIDAP NCI Instance POP.” A follow-on NCATS vehicle, 75N95023D00027, remains open with $69.4 million in combined current award amounts. Taken together, that is clear repeat HHS demand in enterprise data, scientific data access, and platform support. Recurring categories: enterprise data and analytics, secure/scientific data platforms, platform operations. Provenance: USAspending records 75F40121F80001, 75N95022D00025, and 75N95023D00027. Human-owned research action: assign [insert human name] to approve 6 person-hours to separate FDA-style enterprise data buying from NCATS research-platform buying, since the buyer map and incumbent threat model are different even though the recipient family is the same.

USPTO patent-search AI with Accenture Federal Services is the cleanest AI-specific watchlist in this review. USAspending shows DOC/USPTO contract 1333BJ24C00280005 for “Patent Search Artificial Intelligence Models and Labor” with a current award amount of $28.4 million, current end date March 31, 2027, and potential end date September 30, 2029. USAspending also surfaces a prior related contract, 1333BJ23C00280006, with the same patent-search AI description, which is a good sign that the agency is iterating rather than experimenting once. The caveat is that the returned official snippet for the earlier contract did not expose its public value, so I am not filling it in from non-official sources. Recurring categories: AI models, search, software/labor blend, likely operationalization rather than research-only AI. Provenance: USAspending contracts 1333BJ24C00280005 and 1333BJ23C00280006. Human-owned research action: assign [insert human name] to approve 4 person-hours to determine whether USPTO is buying model development, search workflow integration, human-in-the-loop review support, or all three; that distinction will determine whether the family is worth expanding beyond a watchlist.

Watchlist and known unknowns

A credible architecture-specific but smaller family sits at VA OIG with Deloitte Consulting. USAspending shows order 36C10M25F50103 for “VA OIG Enterprise Architecture” with a current award amount of $1.2 million, start date September 17, 2025, and current end date September 16, 2026. The related Deloitte vehicle context also references enterprise architecture solutions and VA activity, which suggests a real buyer pattern, but at the values visible here it is better treated as a watchlist account, not a large pursuit thesis. Human-owned research action: assign [insert human name] to approve 3 person-hours to confirm whether this is a niche advisory stream or the front edge of a broader EA portfolio.

The biggest unknowns in this review are not whether these awards existed, but how much of each family is renewal, option exercise, additional scope, or a true successor buy. USAspending gives strong public signals on incumbent, buyer, amount fields, and performance dates, but those signals do not by themselves reveal acquisition strategy, contractor performance, or whether an agency will recompete versus extend. The CRS guidance also warns that USAspending search results can be incomplete or inaccurate, so the right posture is monitor and research, not assume an open pipeline.

A second unknown is the exact public value of some successor orders when the official returned snippet did not expose it. I left those values as unknown instead of substituting amounts from commercial aggregators. That is deliberate and keeps the output inside the user’s “public source” and “deterministic costs only” guardrails.

Named human decision gate

If the goal is a bounded account-development investment, the rational cut is:

Approve now for: IRS/Accenture, CMS/ALCOR, VA/Palantir, and HHS/Palantir. These families show the clearest recurrence through either multiple orders, parent vehicles with multiple related task orders, or both.

Watch but do not scale yet for: USPTO/Accenture AI and VA OIG/Deloitte architecture. Both are real and relevant, but the repeat signal is narrower and the public evidence in the official snippets here is thinner than the top four families.

The required human decision should be explicit: no account-development spend beyond 27 person-hours total unless [insert a named human approver] signs off on the four “approve now” families and names an owner for each. Without that named human owner, these remain research signals from public award data, not a capture plan.