Saathi

Note: This analysis is based solely on the self-reported project description provided by the authors. No independent verification, archived data, or third-party sources are available. All claims are treated as stated by the author and not proven.

The description states that Saathi is an AI patient advocate platform with a six-step workflow:

1. Intake
2. Document Intelligence
3. Provider Research
4. Coverage Verification
5. Recommendation
6. Care Plan

It uses structured data from AI agents (via OpenAI Agents SDK) and supports manual review steps for accuracy.

• Evidenced: The product is described as a monorepo with FastAPI backend, Next.js frontend, and Supabase database.
• Inferred: That it functions end-to-end in a real-world setting is not evidenced; only that it was built to be tested end-to-end during development.

The authors position Saathi as:

• A guide for navigating complex insurance systems
• An assistant that simplifies the process of finding affordable care
• A tool that reduces time spent decoding insurance and increases clarity around costs

They claim it addresses a common pain point: "millions of people face the same confusion when navigating a new diagnosis, surgery, or specialist referral."

• Evidenced: The tagline and write-up describe Saathi as an AI patient advocate.
• Inferred: That this is a scalable solution or has been adopted by users beyond the team is not evidenced.

The target customer is:

• Patients undergoing new diagnoses, surgeries, or specialist referrals
• Individuals who are overwhelmed by insurance complexity and need guidance on where to go and how much it will cost

• Evidenced: The inspiration story focuses on a pregnant woman needing prenatal care.
• Inferred: That the platform targets specific demographics or health conditions is not detailed.

There is no evidence of pricing, monetization strategy, or business model in the description.

• Not evidenced: No mention of how Saathi would make money or whether it’s intended for consumers, insurers, or providers.

The project was built as a monorepo with:

• FastAPI backend
• Next.js/React frontend
• Supabase for DB/auth/storage
• OpenAI Agents SDK (GPT-5.6)
• Regex-based extraction and rule-based scoring as deterministic fallbacks
• A durable job queue with retries, exponential backoff, and dead-letter handling

• Evidenced: The technical stack and infrastructure components are listed.
• Inferred: That this architecture supports production-scale use is not evidenced.

The project was submitted to a hackathon (OpenAI 2026) and described as:

• A working prototype with full six-step flow
• Tested end-to-end in development
• Used internally by the team to test their own sister’s case

• Evidenced: The internal use case and testing are mentioned.
• Not evidenced: No external users, customer feedback, or real-world adoption data.

No mention of competitors or market positioning beyond general healthcare navigation tools.

• Not evidenced: No competitive landscape analysis, pricing comparisons, or differentiation from existing solutions.

Key risks include:

• Lack of traction or revenue
• Unverified claims about AI performance and reliability
• No evidence of real-world usage or customer validation
• The project is described as a hackathon submission — not a commercial product

• Inferred: That the platform can scale beyond prototype-level functionality without further development is not evidenced.

1. What is the actual user base or pilot program, if any?
2. How does Saathi handle edge cases in insurance documents that don’t conform to standard formats?
3. Is there a plan for monetization or revenue generation?
4. Has the platform been tested with real patients outside of internal use cases?
5. What are the key assumptions about AI accuracy and trustworthiness that underpin the product?