Ranked: 7 AI customer support platforms that sign a BAA for healthcare in 2026. Lorikeet, Sierra, Decagon, Fin, Ada compared on HIPAA, PHI, audit.
Lorikeet News Desk
·
Updated
Every AI support vendor will tell you they are HIPAA-friendly. The only question that matters in healthcare procurement is whether they will sign a Business Associate Agreement and stand behind it. The platforms that survive that conversation are the ones worth a shortlist.
AI customer support for healthcare is a category of agentic AI platforms that resolve patient and member service requests end-to-end - eligibility checks, appointment changes, prior-authorization status, billing questions, prescription refills - across chat, voice, email, and SMS, while handling protected health information (PHI) under a signed Business Associate Agreement. In 2026, the leading platforms resolve a large share of inbound volume autonomously, isolate PHI, and produce the audit trail a HIPAA review demands. Our top picks: Lorikeet for healthtechs and health plans that need multi-step PHI workflows with audit trails under a BAA; Sierra for very large organizations that want outcome-based pricing and a broad compliance list including HIPAA and FedRAMP High; and Decagon for large enterprises that want a premium, sales-led deployment with HIPAA on its security page. Fin, Ada, Salesforce Agentforce and Forethought round out the list.
Lorikeet publishes this guide and is one of the platforms compared. We checked competitor details on their own websites in November 2026; if something has changed, tell us and we will update the page.
A BAA is the legal contract that lets a vendor handle PHI on your behalf under HIPAA. No BAA means the vendor cannot lawfully touch member or patient data, no matter how good the model is.
BAA-readiness is not the same as marketing copy that says "HIPAA compliant." Ask whether the vendor will actually execute a BAA for your contract and at what plan tier.
PHI handling, data isolation, and audit logging matter more to regulated healthcare buyers than headline deflection rates.
Most vendors run on third-party model providers (Anthropic, OpenAI, Google). Confirm the vendor holds contractual no-train and BAA-compatible terms with those subprocessors, not just with you.
Multi-step action chains (verify identity, check eligibility, update the record, escalate when blocked) separate genuine healthcare-grade tools from FAQ deflection bots that cannot touch a system of record.
Last updated: November 2026
Healthcare support has a different failure mode than retail or SaaS. A member asking why a claim was denied is not a churn-risk ticket, it is a HIPAA-exposure ticket. The wrong answer, the wrong disclosure, or PHI sent to the wrong subprocessor is an Office for Civil Rights problem, not a refund. Most vendors will quote a resolution rate. In a covered entity that number is downstream of one threshold question: will you sign a BAA, and can your platform prove how it isolates and logs PHI. This ranking is based on shipping product, real regulated customers, and what a healthcare compliance and privacy team actually approves before go-live.
What a BAA Means for AI Support
A Business Associate Agreement (BAA) is a contract required under HIPAA whenever a vendor (a "business associate") creates, receives, maintains, or transmits protected health information on behalf of a covered entity such as a health plan, provider, or healthtech platform. For AI customer support, the AI vendor is almost always a business associate the moment its agent reads a member ID, a claim, or a clinical note. Without an executed BAA, processing that data is a HIPAA violation regardless of how the technology performs.
Business Associate Agreement (BAA): The HIPAA contract that binds a vendor to safeguard PHI, restrict its use, report breaches, and flow equivalent obligations down to its own subprocessors.
Protected health information (PHI): Individually identifiable health data - names tied to conditions, claims, eligibility, member IDs, clinical details - that HIPAA protects.
A BAA on the master contract is necessary but not sufficient. AI changes the subprocessor map. Most platforms send prompts to third-party model providers like Anthropic, OpenAI, or Google. For the chain to be lawful, the vendor needs BAA-compatible, no-train terms with those model providers too, and it must be able to show you where PHI flows and how it is isolated. The questions that actually separate vendors are: will you sign a BAA on our contract; do your model subprocessors operate under BAA-compatible terms; how is PHI isolated between tenants and redacted in logs; and can you produce an audit trail of every action taken on a member ticket. A vendor that says "we are HIPAA compliant" but cannot answer those four questions is selling a slogan, not a posture.
Lorikeet is an AI customer support platform built for complex, regulated companies including healthtech, health plans, and financial services. Lorikeet signs BAAs for HIPAA, holds SOC 2 Type 2 and ISO 27001:2022, aligns to GDPR, offers configurable PII redaction by category, and maps permissions to user roles. Lorikeet stores data in the USA on Start and Scale; residency is custom on Signature. It resolves multi-step tickets end-to-end across voice, chat, email, SMS, and WhatsApp, and logs every action the agent takes for audit.
At-a-Glance Comparison
Platform: Lorikeet · Best For: Healthtech and health plans that need multi-step PHI workflows with audit trails · HIPAA Posture: Signs BAAs; SOC 2 Type 2; ISO 27001:2022; configurable PII redaction; USA data storage on Start and Scale, custom residency on Signature · Pricing: Start $2,100/mo or Scale $5,100/mo (annual), plus $0.99 or $0.90 per resolution; Signature custom
Platform: Sierra · Best For: Large enterprises wanting outcome-based pricing · HIPAA Posture: Lists SOC 2, HIPAA, PCI DSS Level 1 and FedRAMP High; confirm BAA scope for your contract · Pricing: Quote-based, outcome-based
Platform: Decagon · Best For: Enterprise support orgs with large budgets and sales-led deployments · HIPAA Posture: Security page lists HIPAA, SOC 2 Type II and PCI DSS 4.0.1; confirm BAA execution and subprocessor terms · Pricing: Quote-based
Platform: Fin (formerly Intercom, now part of Salesforce) · Best For: Teams wanting a self-serve, published-price agent on Intercom or their current helpdesk · HIPAA Posture: HIPAA with a BAA for eligible customers on an eligible plan; verify it covers your data flows · Pricing: $0.99 per outcome, plus $19/seat with the Intercom helpdesk
Platform: Ada (combining with Dialpad, announced Nov 2026) · Best For: Mid-market and enterprise teams with high chat volume · HIPAA Posture: Trust centre lists HIPAA and SOC 2 Type 2; confirm BAA availability and PHI handling for your use case · Pricing: Quote-based
Platform: Salesforce Agentforce · Best For: Salesforce and Health Cloud shops · HIPAA Posture: Business Associate Addendum via your Salesforce account rep; confirm which Agentforce services it covers · Pricing: $2 per conversation or Flex Credits, plus licensing
Platform: Forethought (now part of Zendesk) · Best For: Teams wanting resolve, triage, and QA in one stack · HIPAA Posture: HIPAA-aligned controls with an annual HIPAA audit report; BAA not publicly confirmed, so ask · Pricing: Quote-based
BAA availability and scope vary by plan tier, product, and contract. Always confirm in writing which products, channels, and subprocessors a vendor's BAA actually covers before processing PHI.
The 7 Best AI Customer Support Platforms That Sign a BAA for Healthcare in 2026
1. Lorikeet
Lorikeet is an AI customer support platform built specifically for complex, regulated companies, and its security posture is designed for regulated work such as healthcare. It resolves multi-step patient and member tickets end-to-end across voice, chat, email, SMS, and WhatsApp, with an audit trail a privacy team can review step by step. Most vendors say their AI is HIPAA-friendly. Lorikeet is designed so your compliance and privacy team can sign off before launch, with a BAA on the contract and a clear picture of where PHI flows.
Best For
Healthtech platforms and health plans handling regulated workflows (eligibility, prior-auth status, billing, refills, claims questions) where every action touches PHI, needs an audit trail, and needs an answer a compliance team can approve. Lorikeet is built for regulated industries, with customers across fintech, healthtech and insurance.
Key Features
Multi-step action chains: verify identity, check eligibility, look up a claim, update the record, and escalate when blocked, in one ticket and in the right order.
Defence in depth: pre-launch adversarial simulations (including prompt-injection testing and third-party red teams), message checks on incoming messages, runtime guardrails, and automated QA on 100% of conversations through the Coach agent. The model does the reasoning; Lorikeet adds the workflows, guardrails and audit trail.
Audit trails: the actions the agent takes are logged for HIPAA reviews and internal investigations.
Omnichannel resolution across voice, chat, email, SMS and WhatsApp, with structured workflows that follow exact steps plus natural-language workflows, combinable in one interaction.
Simulation-based validation lets you prove guardrail behavior before go-live, not after, on the exact workflows your privacy team is reviewing.
HIPAA Posture
Signs BAAs for HIPAA. SOC 2 Type 2, ISO 27001:2022, GDPR-aligned, configurable PII redaction by category, with PHI handled on a minimum-necessary basis. Lorikeet stores data in the USA on Start and Scale; residency is custom on Signature. Lorikeet does not train models on customer data.
Limitation
Lorikeet is purpose-built for complex, regulated resolution, not a drop-in widget, though you can start with a 30-day self-serve trial. It takes more setup than switching on an FAQ bot: workflows, guardrails and simulations are configured per use case. Its plan fees also make it more expensive than pure pay-per-outcome tools at very low volume. Teams that only want a lightweight FAQ deflection bot, or that have no multi-step or PHI workflows at all, will not use the depth they are paying for. Lorikeet is the right fit when correctness and auditability on hard tickets matter more than a five-minute setup.
Pricing
Lorikeet publishes its pricing: the Start plan is $2,100 a month and Scale is $5,100 a month (both billed annually), with resolved chat, email and SMS conversations at $0.99 (Start) or $0.90 (Scale) and voice at $1.50 or $1.20. Unresolved tickets and handoffs to a person are not charged. Signature is custom, with a $75K platform fee. Full details are on the pricing page. The Coach QA agent is $0.30 (Start) or $0.27 (Scale) per ticket. Coach is also offered as a standalone product for teams on any support solution. Under the Quality Guarantee a resolution you rate bad is not charged.
2. Sierra
Sierra is the enterprise AI agent company known for pioneering outcome-based pricing for AI agents. It runs one agent across chat, SMS, WhatsApp, email and voice, and lists a very broad compliance set: SOC 2, HIPAA, GDPR, PCI DSS Level 1, FedRAMP High, ISO 27001 and ISO 42001. Mature testing tooling (Simulations, Voice Sims, Experiments) makes it a common shortlist entry for large regulated organizations.
Best For
Large enterprises that want outcome-based pricing and have the procurement appetite for a custom enterprise contract.
Key Features
Outcome-based pricing, agreed by contract.
Chat, SMS, WhatsApp, email and voice through a single agent.
Both a developer Agent SDK and a no-code Agent Studio, with Simulations and Voice Sims for testing.
Voice Personas for branded voice agents.
Monitors, Experiments and Agent Traces for observability.
HIPAA Posture
Sierra lists SOC 2, HIPAA, GDPR, PCI DSS Level 1, FedRAMP High, CCPA, CSA STAR, ISO 27001 and ISO 42001 on its trust page. Because BAA availability and scope vary by contract, healthcare buyers should confirm in writing that Sierra will execute a BAA, which products and channels it covers, and how PHI is isolated and flowed down to model subprocessors before processing any member data.
Limitation
Sierra is sales-led with no self-serve trial and quote-based pricing, so it suits large buyers with the budget and timeline for a bespoke enterprise build. In healthcare the high-stakes work (eligibility disputes, claim denials, sensitive PHI handling) is the complex tail, so ask how outcomes are defined and priced on those cases, not just the headline resolution rate.
Pricing
Sierra does not publish pricing; it is quote-based and typically sold as an annual enterprise contract, priced on outcomes.
3. Decagon
Decagon is a high-end enterprise AI agent platform with large enterprise customers and a sales-led implementation. Its Agent Operating Procedures (AOPs) let teams write complex workflows in natural language, and Watchtower adds always-on QA alongside simulations and live A/B experiments.
Best For
Large enterprises with sizeable support budgets that want a premium, sales-led deployment and Decagon's AOP and experimentation tooling.
Key Features
Per-conversation or per-resolution pricing models.
Voice, chat, and email in one platform.
Strong compliance list including HIPAA and PCI DSS 4.0.1.
Natural-language AOPs, Watchtower QA, simulations and live A/B testing.
HIPAA Posture
Decagon's security page lists SOC 2 Type II, ISO 27001, HIPAA, GDPR, CCPA and PCI DSS 4.0.1, with reports via its trust centre. Healthcare buyers should still confirm BAA execution for their contract, which products it covers, how PHI is isolated between tenants, and that no-train, BAA-compatible terms extend to the underlying model providers before sending PHI.
Limitation
Decagon is demo-led with no self-serve trial and quote-based pricing. Ask how your own team will own and change AOPs and guardrails after launch, which matters when compliance requirements shift.
Pricing
Decagon does not publish pricing; it is quote-based and typically sold as an annual enterprise contract. Decagon says customers choose per-conversation or per-resolution pricing.
4. Fin
Fin (formerly Intercom, now part of Salesforce) is an AI agent that runs on the Intercom helpdesk or alongside Zendesk, Salesforce and other helpdesks, and one of the easier platforms to deploy, with a self-serve 14-day trial. Fin offers HIPAA with a BAA for eligible customers on an eligible plan, which makes it a realistic option for some healthcare teams, with the caveat that you must confirm the BAA covers your actual data flows.
Best For
Teams already using Intercom that want a drop-in AI agent with a fast path from trial to deployment.
Key Features
Per-outcome pricing, among the lower published per-resolution rates in the category.
Works with Intercom, Zendesk, Salesforce and other helpdesks, across chat, email, voice, SMS and WhatsApp.
Procedures that take actions through Data Connectors or MCP, plus Simulations, CX Score and Monitors for QA.
Optional copilot for human agents.
Fast trial-to-deployment path.
HIPAA Posture
Fin lists SOC 2 Type II, ISO 27001, ISO 42001 and AIUC-1, and HIPAA with a BAA for Covered Entities or Business Associates on an eligible plan (Intercom pricing lists HIPAA support on Expert). Confirm the BAA explicitly covers Fin and any AI subprocessors, which channels and data types are in scope, and how PHI is redacted from logs and isolated before enabling AI features on member-facing workflows.
Limitation
Fin's Procedures can take actions, but check how they connect to external systems of record (eligibility platforms, claims systems, EHR-adjacent tools) and how deep the audit logging goes for your workflows. HIPAA support sits on specific plans, so factor the plan requirement into total cost.
Pricing
$0.99 per outcome with a 50-outcome monthly minimum on your current helpdesk, or from $0.99 per outcome plus helpdesk seats with Intercom (Essential $19 per seat for new customers, Expert $132). Copilot is $35 per user a month.
5. Ada
Ada (combining with Dialpad, announced Nov 2026) is one of the most established AI support automation vendors, with one agent across voice, messaging and email and broad multilingual coverage. It does breadth well and is a frequent shortlist entry for mid-market and enterprise teams with high chat volume.
Best For
Mid-market and enterprise teams with high inbound chat volume that prefer a vendor with a long history over a newer entrant.
Key Features
Tools to complete customer tasks through open APIs, SDKs and integrations.
Multi-channel: chat, voice, and email.
Tooling to measure, test, coach and optimise the agent, with continuous monitoring for accuracy and safe responses.
Long-standing AI-native vendor, soon backed by Dialpad's voice network.
HIPAA Posture
Ada's trust centre lists SOC 2 Type 2, SOC 3, HIPAA (report dated September 2024), GDPR, CCPA and PIPEDA. Healthcare buyers should confirm whether Ada will sign a BAA for their contract, how PHI is handled and isolated, and that model subprocessors operate under compatible no-train and BAA terms before processing member data.
Limitation
The Dialpad combination may change Ada's packaging and roadmap. Probe multi-step action chains and audit logging on your regulated workflows specifically rather than relying on the headline resolution number.
Pricing
Ada does not publish pricing; it is quote-based and typically sold as an annual enterprise contract.
6. Salesforce Agentforce
Salesforce Agentforce brings agentic AI to the Salesforce platform, which is a natural fit for healthcare organizations already standardized on Salesforce or Health Cloud. Salesforce offers HIPAA coverage through a Business Associate Addendum requested via your account rep, so the practical question is scope: confirm which Agentforce services and model usage it covers.
Best For
Organizations already running on Salesforce or Health Cloud that want AI agents native to their existing platform and data model.
Key Features
Native to the Salesforce platform and CRM data model.
Deep action building via Flows, Apex and MuleSoft, with Agentforce Observability for monitoring.
Broad integration ecosystem across the Salesforce AppExchange.
HIPAA Posture
Salesforce offers a Business Associate Addendum through your account rep. Confirm that Agentforce specifically, and the large language models it invokes, are inside the scope of your BAA, and verify how PHI is isolated and logged within your org before enabling agents on member-facing flows.
Limitation
Agentforce delivers the most value when your data and processes already live in Salesforce. Organizations whose systems of record sit elsewhere, or whose hardest workflows span multiple non-Salesforce systems, may find the depth and total cost less favorable than a purpose-built regulated resolution platform.
Pricing
Salesforce lists $2 per conversation, or Flex Credits at $500 per 100K credits, plus Salesforce platform and edition licensing; Agentforce add-ons start at $125 per user a month.
7. Forethought
Forethought (now part of Zendesk) offers a multi-agent platform spanning resolution, triage, agent assist, and quality assurance, which appeals to teams that want more than a single resolution bot. Its breadth is a strength for organizations consolidating several support-AI functions into one vendor. It is now sold through Zendesk, and its AI agents can also be used with other helpdesks.
Best For
Mid-market and enterprise teams that want a unified AI stack covering resolution, routing, agent assist, and QA in one place.
Key Features
Multi-agent stack covering resolution, triage, assist, and QA.
Natural-language business logic rather than rigid decision trees.
Multi-channel coverage across chat, email, voice and Slack.
Broad system integrations and agent-assist tooling for hybrid models.
HIPAA Posture
Forethought states SOC 2 Type II, ISO 27001 and HIPAA-aligned controls with an annual HIPAA audit report. A BAA is not publicly confirmed, so healthcare buyers should confirm BAA execution for their contract, the products and channels it covers, PHI isolation, and that model subprocessors carry compatible no-train and BAA terms before processing PHI.
Limitation
A broad multi-agent suite can mean more surface area to configure and govern. For healthcare specifically, validate that the depth of audit logging and multi-step PHI action chains meets your privacy team's bar, rather than assuming breadth implies regulated depth.
Pricing
Forethought does not publish pricing; it is quote-based and typically sold as an annual enterprise contract, blending platform access fees with outcome-based pricing.
In healthcare, the threshold question is not deflection rate, it is whether a vendor will sign a BAA and prove how it isolates and logs PHI. See how Lorikeet handles end-to-end, BAA-ready resolution for regulated teams.
How to Choose a BAA-Ready AI Support Platform for Healthcare
Healthcare procurement starts where generic CX procurement ends. Deflection rate, response time, and CSAT all sit downstream of one gate: can this vendor lawfully and provably handle PHI. The lenses below separate platforms that clear a privacy review from those that do not.
BAA Availability and Scope
Ask directly: will you sign a BAA on our contract, at what plan tier, and which products, channels, and AI features does it cover. A BAA that excludes the AI agent, or that is only available on an enterprise tier you are not buying, does not protect you. Get the scope in writing.
PHI Handling and Subprocessor Terms
Most AI platforms send prompts to third-party model providers. Confirm the vendor holds no-train, BAA-compatible terms with those providers, and ask for the current subprocessor list. If the vendor cannot tell you which model providers see your prompts and under what terms, you cannot complete a HIPAA risk assessment.
Data Isolation and Residency
Ask how PHI is isolated between tenants, how it is redacted from logs and analytics, and where it is stored. Data residency in a specific region (US, UK, AU) is a hard requirement for many plans and providers. "We use a major cloud" is not an answer to a residency question.
Audit Trail Depth
Ask for a complete record of every action and tool call on every member ticket, with timestamps and the reason for each step, not a sampled transcript. When a wrong disclosure or a mishandled record surfaces, you need to point at the exact step. Audit-grade logging is the single most important healthcare-specific capability after the BAA itself.
Multi-Step Action Chains and Guardrails You Can Prove Pre-Go-Live
Most healthcare tickets are not "what are your hours." They are "verify my identity, check my eligibility, tell me why my claim was denied, and update my address." The platform must chain several tool calls in order, recover when one errors, and enforce guardrails (no PHI to the wrong party, scripted disclosures, escalation triggers) that you can test and read a pass or fail report on before launch. If guardrails are only a runtime promise, your privacy team is being asked to approve faith, not behavior.
Lorikeet's Take on BAA-Ready AI Support for Healthcare
Most AI vendors will quote a resolution rate. In a covered entity that number is meaningless until two things are true: the vendor will sign a BAA that actually covers the AI agent, and the platform can prove how it isolates, redacts, and logs PHI on every ticket. You can hit a high deflection rate while quietly sending PHI to a model subprocessor with no BAA-compatible terms, which is a HIPAA problem dressed up as a success metric.
The platforms that clear procurement at the regulated companies Lorikeet works with are the ones whose behavior is provable before launch, not the ones with the loudest deflection numbers. The test: can your privacy and compliance team sign off on the BAA scope and the audit log before go-live, and are the agent's actions correct on the tickets that carry risk (eligibility, claims, sensitive PHI), not just the easy ones. If that is the bar your team uses, see how Lorikeet handles end-to-end resolution.
Is Lorikeet the right fit?
Lorikeet is built for complex, regulated support: fintech, healthtech, insurance and telehealth teams where the AI has to take actions in backend systems safely and leave an audit trail. It is not the right choice for everyone:
If you only need an FAQ widget for simple, non-PHI questions, a lighter tool will be cheaper and faster to switch on.
If you are a very large organization that wants a heavily resourced vendor partnership with FedRAMP High on the compliance list, Sierra fits that.
If your support operation already runs on Salesforce or Health Cloud and you want the agent built, governed and billed inside that platform, Agentforce fits that.
If you handle a few hundred tickets a month, the Start plan (from $2,100 a month) will be more than you need.
Key Takeaways
A signed BAA that covers the AI agent is the threshold requirement for healthcare AI support; no BAA means no lawful PHI processing, regardless of model quality.
"HIPAA compliant" marketing copy is not a BAA. Confirm in writing which products, channels, and model subprocessors a vendor's BAA actually covers, and at what plan tier.
PHI handling, data isolation, residency, and complete audit trails outweigh headline deflection rates in regulated healthcare evaluations.
Lorikeet ranks first here because its security posture is built for regulated work: it signs BAAs, holds SOC 2 Type 2 and ISO 27001:2022, offers configurable PII redaction, and adds multi-step action chains across voice and digital channels, simulation-based validation, and 100% QA through Coach.
Several strong platforms (Fin, Salesforce Agentforce, and others) offer BAAs, but scope and depth vary; validate the BAA, the subprocessor terms, and the audit logging against your own workflows before signing.
Conclusion
The healthcare AI support market in 2026 is not a question of whether to deploy AI. It is a question of which platform will sign a BAA, prove how it isolates and logs PHI, and resolve the regulated tickets that matter (eligibility checks, claim denials, prior-auth status, refills) with an audit trail your privacy team and regulators trust.
The seven platforms above each fit a different profile by existing stack, budget, and risk tolerance. Lorikeet fits healthtechs and health plans whose toughest stakeholder is their compliance and privacy team, who need multi-step PHI action chains across voice, chat, email, and SMS, and who want the agent's behavior provable before go-live. The other six are credible options depending on your helpdesk, your data model, and how their BAA scope maps to your workflows.
If you are evaluating BAA-ready AI customer support for a healthcare team, book a Lorikeet demo and bring your hardest member workflows to see how the agent handles them.
Frequently asked questions
Do AI customer support platforms sign a BAA for healthcare?
Some do, and several on this list will. A Business Associate Agreement is the HIPAA contract that lets a vendor lawfully process protected health information on your behalf, so any AI agent that touches member IDs, claims, or clinical data needs one. Lorikeet signs BAAs for HIPAA. Fin offers a BAA for eligible customers on an eligible plan, and Salesforce offers a Business Associate Addendum through your account rep. Sierra, Decagon and Ada list HIPAA on their trust or security pages, and Forethought (now part of Zendesk) states HIPAA-aligned controls but does not publicly confirm a BAA, so ask. The critical step is confirming in writing that the BAA covers the AI agent specifically, not just the underlying helpdesk, and at the plan tier you are actually buying.
What does a BAA actually cover for AI support?
A BAA binds the vendor to safeguard PHI, limit how it is used, report breaches, and flow equivalent obligations down to its own subprocessors. For AI support, the part buyers most often miss is the subprocessor chain: prompts usually go to third-party model providers like Anthropic, OpenAI, or Google. For the chain to be lawful, the vendor needs BAA-compatible, no-train terms with those providers too. Ask for the current subprocessor list and confirm the terms cover every place PHI flows.
Is "HIPAA compliant" the same as signing a BAA?
No, and conflating them is a common procurement mistake. "HIPAA compliant" is marketing language a vendor can apply to itself. A BAA is an executed legal contract that creates enforceable obligations. A platform can describe itself as HIPAA-friendly and still decline to sign a BAA, or only offer one on a tier you are not purchasing. Treat the signed, in-scope BAA as the requirement and the marketing label as unverified until you see the contract.
How does Lorikeet handle PHI and HIPAA?
Lorikeet signs BAAs for HIPAA and is built for regulated industries including healthcare. It holds SOC 2 Type 2 and ISO 27001:2022, aligns to GDPR, offers configurable PII redaction by category with PHI handled on a minimum-necessary basis, and maps permissions to user roles. Lorikeet stores data in the USA on Start and Scale; residency is custom on Signature. The actions the agent takes are logged in an audit trail for HIPAA reviews. These features support your HIPAA obligations; your own configuration and policies remain part of meeting them.
What should a healthcare team verify before processing PHI with an AI vendor?
Five things, in writing. First, that the vendor will sign a BAA covering the AI agent at your plan tier. Second, the subprocessor list and confirmation that model providers operate under no-train, BAA-compatible terms. Third, how PHI is isolated between tenants and redacted from logs and analytics. Fourth, where data is stored and whether residency meets your requirement. Fifth, whether you can pull a full audit trail of any member ticket and test guardrails before go-live. A vendor that cannot answer all five is not ready for a covered-entity workload yet.

