--- title: "The VA's $775.72M Ambient Scribe Award Is a Ceiling, Not a Commitment" slug: "va-775m-ambient-scribe-idiq-ceiling-workflow-evaluation" author: "ibl.ai Engineering" date: "2026-10-15 11:00:00" category: "Premium" topics: "government AI procurement, IDIQ ceiling, ambient clinical scribe, VA contract, Veterans Affairs, Abridge, Knowtex, Carahsoft, evaluation science, VistA CPRS, Federal EHR, model-agnostic AI, self-hosted AI, clinical documentation" summary: "The VA's ambient scribe award is circulating as $775 million of committed AI spending. The contract record says something more useful to a buyer: it is a multiple-award IDIQ with a $775.72 million ceiling shared across every vendor eligible to compete for task orders, the contracts of record went to Knowtex and to Carahsoft rather than to the vendor named in most headlines, and the adoption figure everyone quotes is vendor-reported. Read carefully, it is an argument for holding your own measurement." banner: "" thumbnail: "" linkedin: | The VA's ambient scribe award is being shared as "the VA committed $775M to AI." The number is right and the sentence is wrong in three ways that matter to anyone writing a procurement strategy. First, the instrument. What the VA signed is a multiple-award Indefinite Delivery, Indefinite Quantity contract with a total ceiling of $775.72 million over five years, shared across every vendor eligible to compete for task orders. A ceiling is permission to spend, not money spent. It was awarded by the VA Technology Acquisition Center under solicitation 36C10B26R0006, notice dated 22 September 2026, and the task orders are still ahead. Second, the awardees. Most coverage names Abridge. The contracts of record went to Knowtex Inc. and to Carahsoft Technology Corporation, the federal reseller through which Abridge's product reaches the vehicle. Abridge won a seat to compete; Knowtex is the one whose own release claims prime awardee. If you are mapping a federal market, the distinction between a vendor and its aggregator is the whole map. Third, and this is the part worth keeping. The adoption number everyone quotes, more than 986,000 VA primary care appointments using ambient scribes since October 2025 with veterans declining less than 1% of the time, is reported by the vendors and relayed by a VA official. It may well be accurate. It is still not the agency's own instrument. That is the procurement lesson, and it is not a flattering one for anybody. The single most persuasive figure in a three-quarter-billion-dollar federal award is a number the sellers produced. Generated clinical notes become part of a patient's legal health record, so they have to be accurate and auditable. If the measurement that tells you whether that is true comes from the party being measured, you have a governance gap no contract clause closes. The VA bought COTS cloud SaaS here, deliberately. That is a reasonable choice. It just means the evaluation layer and the audit trail live somewhere you do not administer, and you should price that. With ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay by usage with no per-seat pricing. #iblai #GovernmentAI #Procurement #HealthcareAI #SelfHosted --- ## The Short Answer **The VA's ambient scribe award is a multiple-award IDIQ with a $775.72 million ceiling shared across every vendor eligible to compete for task orders, not $775.72 million committed to anyone. Its headline adoption figure is vendor-reported. That is the case for holding your own measurement: with ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing.** The widely-shared version of this award gets the figure right and the shape wrong. Correcting the shape changes what a buyer should take from it. ## What exactly did the VA award in the $775.72M ambient scribe contract? A five-year, multiple-award IDIQ for commercial cloud ambient clinical documentation, with two contract holders of record. The VA's award is a multiple-award Indefinite Delivery, Indefinite Quantity contract carrying [a total ceiling of $775.72 million over five years across all eligible competing vendors](https://hitconsultant.net/2026/09/22/abridge-awarded-va-enterprise-contract-ambient-ai-veterans-health-administration/), reported on 22 September 2026. It was awarded in September 2026 by the VA Technology Acquisition Center in New Jersey (36C10B) under [solicitation 36C10B26R0006](https://discover.govdash.com/solicitations/019d0c28-599a-75ce-b09d-5f8d2b7d8a62--da10-ambient-scribe-enterprise-idiq-va-26-00012979-award-notice), with the award notice dated 22 September 2026. The scope is the largest integrated health system in the United States. It delivers care to more than 9 million veterans across 1,380 healthcare facilities, including 170 VA Medical Centers and nearly 1,200 outpatient care sites. The technology has to work in two electronic health record generations simultaneously: the legacy VistA/CPRS architecture and the new Federal EHR running on Oracle Health. One detail in the award notice deserves more attention than it gets. The vehicle is for "Commercial-Off-The-Shelf (COTS), cloud-based" solutions and specifies a **Software as a Service delivery model**. The VA chose multi-tenant SaaS on purpose. ## Who actually holds the VA ambient scribe contracts? Knowtex Inc. and Carahsoft Technology Corporation, which is not the same list as the one in most headlines. This is where the reporting and the contract record diverge, and it is worth being precise because the difference is instructive. The [award notice](https://discover.govdash.com/solicitations/019d0c28-599a-75ce-b09d-5f8d2b7d8a62--da10-ambient-scribe-enterprise-idiq-va-26-00012979-award-notice) names the two contract holders as Knowtex Inc. and Carahsoft Technology Corporation. Carahsoft is a government aggregator and reseller, not an ambient scribe vendor. Abridge's product reaches the vehicle through Carahsoft's contract. HIT Consultant's own account notes Abridge was selected "through a distribution partner," and headlines its award as winning a **seat** on the contract. The asymmetry shows up in the vendors' own language. Only [Knowtex](https://www.globenewswire.com/news-release/2026/09/25/3369134/0/en/knowtex-the-first-frontier-ai-lab-for-healthcare-named-prime-awardee-on-va-s-nine-figure-ambient-scribe-enterprise-contract.html) claims "prime awardee," in a release dated 25 September 2026. Both companies can compete for task orders. But if you are mapping a federal market, a vendor and its aggregator are different entities, and reading the vendor's press release as the contract record is how that distinction disappears.
Parameter What the award record says
Instrument Multiple-award IDIQ, not a single-vendor award
$775.72 million Total ceiling across all eligible competing vendors, over five years
Contract holders of record Knowtex Inc. and Carahsoft Technology Corporation (Abridge's federal reseller)
Money obligated to a vendor Determined later, per task order
Delivery model COTS, cloud-based, Software as a Service
EHR surface Legacy VistA/CPRS and Federal EHR (Oracle Health)
## Does a $775.72 million IDIQ ceiling mean the VA committed $775.72 million? No. An IDIQ ceiling is the maximum the government may spend under the vehicle, not an obligation to spend it, and here it is shared. [Nextgov's reporting](https://www.nextgov.com/artificial-intelligence/2026/09/va-selects-abridge-ambient-scribe-under-new-enterprise-contract/416140/) states the ceiling covers "all vendors who are eligible to compete for task orders." The competition for the actual work happens after the award, task order by task order. This distinction is not pedantry. It changes the arithmetic a vendor or an agency should run. A headline reading "the VA committed $775M" implies a budget already allocated. What exists is permission to spend up to that figure across multiple vendors across five years, with each increment still to be competed. For a buyer, the practical consequence is that winning a seat on the vehicle is the beginning of the sales cycle, not the end of it. ## What did each ambient scribe awardee actually bring to the VA vehicle? Two different profiles: an installed footprint across VA sites in one case, and clinical immersion plus a published evaluation method in the other. A caveat first, because it is the kind of thing that gets asserted without a source. The VA has not published its source-selection rationale for 36C10B26R0006, and no outlet reports one. What follows is what each company brought, not a claim about why the VA chose them. Abridge's stated advantage was integration reach. Per [Nextgov on 14 September 2026](https://www.nextgov.com/artificial-intelligence/2026/09/va-gradually-expand-medical-ai-scribe-following-successful-pilots/415971/), the company says its technology is "deployed at 70 VA sites and over 300 U.S. health systems." Other coverage renders this as more than 75 VA medical centers. Running against two EHR generations at once is an integration achievement regardless of which figure you take. Knowtex's route was domain immersion. Founders Caroline Zhang and Jocelyn Kang met as Stanford freshmen studying AI and linguistics, founded the company in 2022, and [spent a combined year in scrubs as medical scribes](https://www.fiercehealthcare.com/ai-and-machine-learning/knowtex-launches-frontier-healthcare-ai-lab), working alongside clinicians by day and building at night. They also [published a clinical AI benchmark, KnowBench](https://www.prnewswire.com/news-releases/knowtex-launches-the-frontier-ai-lab-for-healthcare-bringing-research-grade-clinical-ai-to-health-systems-to-maximize-intelligence-per-patient-encounter-302878575.html), on 15 September 2026, reporting that its own platform scores 97.99% on it. Note the date: the week before the award notice. A self-published benchmark carrying a self-reported score is worth reading as a statement of method, not as independent validation, and it is too recent to have decided anything. ## Is the VA's ambient scribe adoption data independent evidence? Not strictly. The most-quoted figures in the story are reported by the vendors and relayed by the agency. The numbers are genuinely impressive. More than [986,000 VA primary care appointments have used ambient scribes since October 2025](https://www.cdomagazine.tech/us-federal-news-bureau/va-awards-new-ai-scribe-contract-with-lessons-for-governing-ai-at-scale), and veterans have declined the technology less than 1% of the time. Read the attribution carefully. Nextgov's wording is that the appointment total comes "with vendors reporting veterans have declined the use of the technology less than 1% of the time" since October 2025. Dr. Rebecca Gladding, acting deputy chief of psychiatry at VA Greater Los Angeles Healthcare System, is quoted on the under-1% figure. The agency relayed it; the vendors produced it. This is not an accusation. The figure may be exactly right, and a sub-1% refusal rate across nearly a million appointments would be a real result. But it is the single most persuasive number in a three-quarter-billion-dollar federal award, and the parties being evaluated are the parties who measured it. That is a governance gap, and no contract clause closes it. ## Why does a buyer want the evaluation layer and audit trail under its own administration? Because the artifacts are legal records, and because measurement supplied by the vendor being measured is weak evidence. The governance requirements here are concrete. [Reporting on the award](https://www.cdomagazine.tech/us-federal-news-bureau/va-awards-new-ai-scribe-contract-with-lessons-for-governing-ai-at-scale) identifies three: mapping where data originates, setting access rules and consent protocols for sensitive patient conversations, and building auditability into the system as it scales. The third one carries the weight. Once a generated note becomes part of a patient's legal health record, it has to be accurate and auditable, which turns "how was this produced" into a compliance question rather than an engineering curiosity. To be clear about what this does and does not imply: the VA bought SaaS here deliberately, and a well-run multi-tenant service can absolutely provide audit logs and evaluation tooling. Vendors do. The difference is who administers the instrument, who can change it, and what happens at contract exit. When the scoring and the trace run in a vendor's tenant, your evidence about the vendor depends on the vendor's continued cooperation. This is the same structural point that governed [the Pentagon's move from AI pilots to procurement-grade deployment](/blog/pentagon-13-4b-ai-budget-fy2026-government-procurement): at procurement scale, control over the controls starts to matter. On ibl.ai, that layer is part of the product you self-host. You run evaluations with model-graded scoring plus human annotation, keep transcripts and an audit trail of every query, and export the results, inside your own infrastructure. ## What should a government or enterprise buyer do differently after the VA award? Read the instrument before the headline, check who actually holds the contract, and ask who produced every number in the business case. Three changes follow from the record above. First, a multiple-award IDIQ with a $775.72 million shared ceiling is a different fact from $775.72 million awarded, and vendor marketing will not make the distinction for you. Second, verify the awardee list against the award notice. Here the contract holders of record are Knowtex and a reseller, which is not what the headlines say. Third, label every figure in your evaluation with its source. The VA's own flagship adoption statistic is vendor-reported, and that is in a federal award with far more procurement rigour than most private buyers apply. 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. ## The Edge A $775.72 million ceiling shared across competing vendors is permission to spend, not spending. The contract holders of record are Knowtex and Carahsoft, a reseller, rather than the vendor most coverage names. And the adoption statistic carrying the whole story, under 1% veteran refusal across 986,000+ appointments, was produced by the vendors and relayed by the agency. None of that makes the award a bad one. It makes it a precise illustration of where procurement evidence actually comes from. The reproducible lesson is that the numbers deciding an AI purchase tend to originate with the seller, and that generated clinical notes enter a patient's legal record, where accuracy and auditability are obligations rather than features. Both pressures point the same way: toward measurement you administer. With ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing. ## Why does owning the AI stack matter? **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.