CareSource Foundation's $1 Million AI Care RFP: One Award, Character-Limited Concepts Due January 6, and a 30% Technical Score
November 15, 2026 · 7 min read
Granted Research Team · Editorial policy
The CareSource Foundation is running a two-phase national competition with board-level review, a ten-section full proposal, and scored responsible-AI criteria. All of that leads to one grant.
The RFP, "Advancing AI-Enabled Models for Care and Community Transformation," was released November 10, 2026. It offers up to $500,000 a year for two years, for a total of $1,000,000, and the award structure section says plainly: "Anticipated one award." Phase 1 concept proposals are due January 6, 2027, at 11:59 p.m. EST. Invited applicants submit full proposals by February 12, 2027. Final award notification is scheduled for March 18, and the project launches in Q2 2027. The RFP page and the full RFP document are the controlling sources.
With one winner, the size of the field matters as much as the size of the check. The application process also differs from most foundation RFPs in ways that will catch unprepared applicants.
A Health Plan Foundation Buying a Partner, Not Just a Project
CareSource is a nonprofit managed care organization headquartered in Dayton, Ohio. Its foundation was established in 2006 and says it has awarded "more than $39 million to nonprofits" since then. In July 2026 the foundation moved its regular grantmaking to an invitation-only process, with Signature Grants of $100,000 or more and Community Grants below that. This RFP is an open call during a period when the foundation has otherwise closed its doors. It is a rare cold-entry point for organizations without an existing CareSource relationship.
The RFP is clear that the foundation is looking for a working relationship. The selected partner will "collaborate with the CareSource Foundation to explore the art of the possible," and the foundation may provide "strategic collaboration," "connections to community partners," and "access to data (where appropriate)." Section 11 describes an advisory structure with "strategic guidance and checkpoints," "performance and risk monitoring," and "responsible AI oversight discussions." Payments are "milestone-based disbursements tied to deliverables" rather than annual lump sums.
In practice, this is closer to a structured pilot with a payer than a typical program grant. The word "member" appears throughout the RFP, defined as "an individual served by or eligible for health care coverage, services, or supports," and one stated goal is improving member outcomes while enhancing "operational effectiveness." The phrase "return on investment" appears in the list of things solutions should demonstrate. Applicants should read these signals literally. The foundation wants a model that could plausibly be adopted by a Medicaid plan after the grant ends, and Section J of the full proposal asks how the solution "will continue beyond the CareSource Foundation funding."
CareSource president and CEO Erhardt Preitauer framed it in the announcement: "The real promise of AI in healthcare isn't the technology itself. It's what that technology can help us do for people."
Who Can Apply, and the Partnership Requirement That Screens People Out
The lead applicant must be a 501(c)(3) "in good standing for at least three years," serving as lead or fiscal sponsor. The RFP explicitly includes "affiliated nonprofit entities of academic institutions," which opens the door to university research foundations and academic medical center nonprofits. Leads must also show:
- Demonstrated experience serving Medicaid, dual-eligible (Medicare and Medicaid), and/or underserved populations
- "Proven organizational capacity and financial stability"
- The ability to manage data responsibly and comply with privacy standards
The most consequential requirement is the partnership mandate. Applicants "must demonstrate a cross-sector collaborative partnership" that combines the expertise needed to implement, evaluate, and scale the solution, and the Phase 1 concept must name the partners, their roles, and what each brings. Applicants without that partnership "may be deemed ineligible for advancement to Phase 2."
This shapes who is competitive. A technology vendor cannot apply directly; it needs a nonprofit lead. A community health organization with no technical partner will struggle to show an implementation-ready AI solution. The strongest configurations will probably pair a community-based or safety-net provider nonprofit, which supplies the Medicaid population and the workflow, with a technology partner that already has a working product and an academic or research partner that supplies evaluation. If you are a vendor, find a nonprofit lead now. If you are a nonprofit, the next seven weeks are about recruiting partners, not drafting prose.
The RFP does not limit eligibility by state. CareSource's foundation FAQ says it "prioritizes its support for programs within CareSource service areas but will consider additional locations on a case-by-case basis." An applicant outside CareSource's markets is not barred, but it is starting at a disadvantage that the concept proposal should address directly, for example by explaining how the model would transfer to a CareSource market.
Phase 1 Is a Character Count, Not a Page Count
The concept proposal is short and strictly bounded. Beyond eligibility and organizational information, it has six scored narrative fields, each with a hard character limit that includes spaces:
| Section | Limit |
|---|---|
| Problem and Target Population | 1,000 characters |
| Proposed AI-Enabled Solution | 2,550 characters |
| Existing Capability and Readiness | 2,550 characters |
| Anticipated Impact | 1,700 characters |
| Scalability and Sustainability | 1,700 characters |
| Partnership Opportunity | 850 characters |
That is about 10,350 characters of narrative in total, or roughly 1,600 to 1,800 words. The screening rules are strict. Proposals may be deemed non-responsive if they "exceed the stated character limits" or "include supplemental materials, appendices, or other additional materials not expressly requested." Content over the limits "will not be reviewed or scored."
Two of those fields deserve the most attention. "Existing Capability and Readiness" gets the same space as the solution description, which tells you the foundation weights readiness as heavily as novelty. The RFP repeatedly asks for "implementation-ready solutions" and "existing capabilities or demonstrated models that can be operationalized." A concept built on a model that has not yet been built will read as weak next to a team that can point to a prior deployment, even a small one. The 850-character "Partnership Opportunity" field asks why working with the CareSource Foundation would help. A generic answer wastes it. A specific answer names what CareSource data, community connections, or plan workflows would let you validate that you cannot validate alone.
Phase 1 review runs January 6 to 18, and Phase 2 invitations go out January 18. That leaves invited applicants about 25 days to produce the full proposal.
How the Full Proposal Is Scored
The Phase 2 rubric is published with weights, which is unusual and useful:
- Solution Description and Technical Approach: 30% (20,000 characters, plus a required work plan attachment)
- Outcomes, Measurement, and Learning: 15%
- Data, Privacy, Security, and Responsible AI: 15%
- Organizational Overview and Experience: 10%
- Statement of Need: 10%
- Budget and Budget Narrative: 10%
- Project Team and Resource Requirements: 5%
- Scalability and Sustainability: 5%
Each criterion is scored on a 1-to-5 scale. Two things stand out. First, responsible AI is worth as much as outcomes measurement. The section asks for bias mitigation, human oversight, transparency, explainability, model validation, and "ongoing performance monitoring." The RFP says AI should "support—not replace—appropriate human judgment and oversight." Teams that treat this as boilerplate are giving away 15 points. Bring a real model card, a validation plan stratified by population subgroup, and a named human-in-the-loop design.
Second, the technical section requires applicants to "distinguish between existing solution capabilities and components that would require adaptation, customization, or further development." The section also asks for references the foundation may contact. Reviewers will check whether a claimed capability actually exists.
The budget section asks for a "high-level total cost of ownership, including people, process, and technology considerations as well as implementation, ongoing operating, licensing, maintenance" costs. That is procurement language. Budget for post-grant licensing honestly; a model that costs more to run than a Medicaid plan would pay will not score well on sustainability.
The foundation also reserves the right not to pick the highest scorer, and says it will select "the applicant deemed to be in the best overall interest of the Foundation." The full board participates, "with 100% of Board members engaged" in review, discussion, or final selection. Phase 2 may include presentations or demonstrations between February 15 and 22. Plan for a live demo.
The Fine Print Worth Reading Before You Start
Several provisions deserve attention before a team commits time:
The confidentiality clause. Applicants and their partners "shall not make any public statement, press release, social media post, marketing communication, or other disclosure regarding this RFP, their proposal, the evaluation process" or any award without written consent. Violations can lead to disqualification. Technology partners used to announcing pilots should be briefed.
Intellectual property is negotiated later. IP, licensing, commercialization, and data ownership "will be negotiated with the selected awardee." Vendors with proprietary models should decide in advance what they can concede, because the foundation's data access is part of the value it is offering.
Dates around the holidays. An information session is listed for December 10, 2026, though Section 9.1 says the foundation "may hold" one, with details "if scheduled." Questions go to Nicole Krider at Nicole.krider@caresource.com. Submission is through the foundation's Blackbaud Grantmaking portal, linked from the RFP page. With a January 6 deadline, most drafting will happen over the holidays.
Should You Apply?
The case for applying is a $1 million, two-year commitment, a real Medicaid plan as a learning partner, and a rare open door to a foundation that has otherwise gone invitation-only. The case against it is the odds: one award, an unrestricted national field, and a growing number of health AI startups looking for nonprofit partners.
The organizations most likely to win already have a deployed AI tool serving a Medicaid or dual-eligible population, outcome data from that deployment, a nonprofit lead with at least three years of standing, and a credible link to a CareSource market. Teams missing two or more of those pieces should still consider submitting the short concept proposal, since the Phase 1 effort is modest, but they should not build their 2027 budget around this award. For everyone else, the next step is to lock the partnership and write the readiness section first. Tools like Granted can help you draft character-limited concepts and test them against the scoring rubric before January 6.