VA: AI roleplay training for crisis-line responders (ReflexAI) — a ~$3.9M federal contract (USAspending)
The VA's Office of Suicide Prevention (OSP) ordered technical and functional support for AI-powered roleplaying simulation training for responders at the Veteran Crisis Line (VCL) call center, about $3.9 million ($3,892,500). The recipient is ThunderCat Technology (a ReflexAI solution). AI used to train responders — not to replace them.
Contract key facts
- RecipientTHUNDERCAT TECHNOLOGY, LLC
- Contract value$3,892,500 (≈$3.9M)
- Awarding agencyDepartment of Veterans Affairs
- Awarding sub-agencyDepartment of Veterans Affairs
- Award typeDELIVERY ORDER
- Period of performance2025-12-15 〜 2026-12-14
- Contract ID (PIID)36C10B26F0046
Contract scope (original)
REFLEXAI FY26 BASE. THE CONTRACTOR SHALL PROVIDE TECHNICAL AND FUNCTIONAL SUPPORT TO THE OFFICE OF SUICIDE PREVENTION (OSP) INCLUDING THE VETERAN CRISIS LINE (VCL) CALL CENTER, IN ARTIFICIAL INTELLIGENCE (AI) POWERED ROLEPLAYING SIMULATION TRAINING.
Key points
- The VA's Office of Suicide Prevention is adding AI roleplay training for Veteran Crisis Line (VCL) responders
- Recipient ThunderCat Technology (a ReflexAI solution); ~$3.9M; performance Dec 2025–Dec 2026
- AI supports the training of human responders — not the crisis conversations themselves
- It can make roleplay training a safe, repeatable environment across diverse scenarios
- A case of using AI to augment people, not replace them, in a life-critical domain
- Across the 222 federal spending records this site holds as of 2026-09-01, median amounts by period run $4.00M under a year, $1.96M for one to two years and $5.10M beyond three.
1The Veteran Crisis Line as the setting
The Veteran Crisis Line (VCL) is where veterans in distress and their families can reach out by phone or chat. In such settings, training quality is critical so responders can handle difficult moments calmly. Roleplay (simulated calls) is a proven training method, but practicing against people limits the number and variety of scenarios.
2Training responders rather than answering calls
3Length and amount do not rise together
Dividing the 222 federal spending records this site holds as of 2026-09-01 into four bands by contract period and taking the median amount in each gives about $4.00 million under a year, about $1.96 million from one to two years, about $2.09 million from two to three, and about $5.10 million beyond three.
| Contract period | Records | Median amount |
|---|---|---|
| Under one year | 49 | about $4.00M |
| One to two years | 73 | about $1.96M |
| Two to three years | 44 | about $2.09M |
| Three years or more | 56 | about $5.10M |
Longer does not simply mean larger. The band under a year sits above the one-to-two-year band because some procurement puts a concentrated sum into a short run. Period and amount are settled on separate grounds, and neither can be inferred from the other.
Why it matters
A sign that government AI is extending into supporting human training in life-critical settings. A useful example of responsible AI — augmenting people rather than replacing them.
FAQ
Does this AI handle the calls?
Why train with AI?
Sources (primary)
This article is an independent organization based on the U.S. official spending data below. Verify the exact, latest details with the official source.
- USAspending (award details)
- Contract ID (PIID)：36C10B26F0046