Quick answer: The seven AI concierge platforms worth shortlisting for native healthtech integrations in 2026 are Lorikeet, Decagon, Sierra, Fin by Intercom, Ada, Salesforce Agentforce, and Cognigy, and Lorikeet ranks first because its concierge actions read and write to your systems of record on every plan, its ticket platform integrations (Zendesk, Intercom, HubSpot, Front, Salesforce) are included from the Start tier, and it publishes the pricing and HIPAA posture (BAA on Scale and above) that the other six leave to a sales call.
This page is the narrow cut. It ranks AI concierge platforms on one question only: how deep and how native are the integrations between the concierge and the infrastructure a healthtech actually runs on, meaning the helpdesk, the EHR or the identity and eligibility source in front of it, the scheduling system, billing and claims status, the patient portal, and the messaging channels. If you want the wider evaluation covering compliance, quality control, channels, and total cost, read the broader end-to-end guide at Best AI Concierge Platforms for End-to-End Healthtech Support (2026).
A healthtech concierge lives or dies on what it can reach. A member asking whether a visit is covered, a patient asking to move an appointment, a caregiver asking why a claim was denied: none of these are answerable from a help center. They require the agent to look something up in a system of record and, in most cases, change something there. Every demo shows an agent taking an action; the word native is doing a lot of work in those demos, and this guide unpacks it.
What Native Integrations Mean for a Healthtech Concierge
A native integration is a connection the vendor builds, maintains, documents, and supports as part of the product. It has a configuration surface your admin can see, an auth model your security team can review, and a failure mode the vendor has already handled. A custom integration is anything else: a connector a services engineer wrote for your deployment, a middleware hop, or a script that calls your API from inside a workflow. Most healthtech stacks need at least one custom integration. The distinction matters because native integrations survive vendor upgrades and staff turnover, and custom ones tend to become the reason a concierge quietly stops resolving a category of ticket six months after launch.
Three terms are used throughout this guide and are worth pinning down.
System of record: the database that holds the authoritative version of a fact. For a healthtech that is usually the EHR or practice management platform for clinical and appointment data, a membership or eligibility service for coverage, a billing engine for balances and claims, and the helpdesk for the ticket itself.
API and webhook: the two directions of an integration. The API is how the concierge reaches into your system to read or write. The webhook is how your system reaches out to the concierge when something changes, for example a claim status update that should trigger a proactive message.
Write-back: the concierge changing data in the system of record, rather than only reading it. Write-back is where authorization, audit, and guardrails become non-negotiable, and it is the capability that separates a resolution engine from a search box.
Lorikeet is an AI concierge platform built for complex, regulated companies including healthtechs, fintechs, and insurers. Its concierge resolves tickets end to end across chat, email, voice, and SMS, using natural-language workflows alongside deterministic structured workflows, with concierge actions that read and write to your systems of record. Custom data integrations, custom guardrails, and concierge actions are available on every plan, and ticket platform integrations for Zendesk, Intercom, HubSpot, Front, and Salesforce are included on Start. How Lorikeet works covers the workflow model in more detail.
Integration Matrix: The Seven Platforms Side by Side
The table records only what each vendor states in its own positioning. Where a vendor does not state a capability, the cell says Not stated, which is a prompt to ask for it in writing rather than a judgment that it is missing. Competitor pricing is listed as Not published because none of the six publish a rate card.
Platform | Ticket platforms | Custom data/API integrations | Webhooks | Writes back to system of record | Human handoff inside helpdesk | Published pricing |
|---|---|---|---|---|---|---|
Lorikeet | Zendesk, Intercom, HubSpot, Front, Salesforce on Start; all integrations on Scale and Enterprise | Yes, custom data integrations and concierge actions on all plans | Yes, alongside API integrations | Yes, concierge actions read and write to your systems of record | Yes, escalation to your team in the connected ticket platform | Yes: Start $2,100/mo, Scale $5,100/mo, billed annually; Signature custom |
Decagon | Not stated | Custom-built per deployment with embedded engineering | Not stated | Not stated (scoped as custom integration work) | Not stated | Not published (contact sales) |
Sierra | Not stated | Custom-built per customer | Not stated | Not stated | Not stated | Not published (outcome-based, contact sales) |
Fin by Intercom | Intercom; also works with Salesforce and HubSpot helpdesks | Custom actions to your APIs | Yes, via custom actions and webhooks | Not stated | Runs inside the Intercom helpdesk | Not published here (contact sales) |
Ada | Salesforce, Zendesk, and major helpdesks | Custom action integrations | Yes, into your APIs and systems | Not stated | Not stated | Not published (contact sales) |
Salesforce Agentforce | Salesforce CRM and Health Cloud (native) | MuleSoft and AppExchange connectors for external systems | Not stated | Within Salesforce data via flows; external systems not stated | Not stated | Not published (per-conversation plus platform license, contact sales) |
Cognigy | Not stated (contact-center and CCaaS integrations instead) | Connector library plus custom API integrations | Not stated | Not stated | Not stated | Not published (contact sales) |
The ticket platform column is where vendors are most specific, because helpdesk connectors are the oldest integrations in the category. The write-back column is where they are vaguest. Treat a Not stated there as the first item on your pilot checklist.
The 7 Best AI Concierge Platforms With Native Healthtech Integrations in 2026
1. Lorikeet
Lorikeet is the AI concierge built for complex, regulated companies, with healthtech as a core vertical alongside fintech and insurance. It ranks first on this list for a specific reason: its integration model is the product, not a services add-on. Concierge actions that read and write to your systems of record are available on every plan, custom data integrations and custom guardrails are available on every plan, and the ticket platform integrations most healthtechs already run on (Zendesk, Intercom, HubSpot, Front, Salesforce) are included from the Start tier. On Scale and Enterprise, all integrations are included.
The workflow layer is what makes those integrations usable by a support team. Lorikeet combines natural-language workflows, where you describe the process in plain English, with deterministic structured workflows for steps that must run in a fixed order, such as verifying identity before returning a balance. The concierge resolves tickets end to end across chat, email, voice, and SMS on one workflow engine, so an integration built for chat is the same integration the voice agent uses.
Key integration features
Ticket platform integrations on Start: Zendesk, Intercom, HubSpot, Front, and Salesforce. All integrations on Scale and Enterprise.
Custom data integrations, custom guardrails, and concierge actions on all plans, so the concierge can read a record from your identity, eligibility, scheduling, or billing source and write the result back to the system of record.
Lori MCP, which lets AI tools call Lorikeet.
Unlimited testing and simulations on all plans, which is what makes the pilot test in this guide practical: you can run an integration against hundreds of synthetic cases before a single patient touches it.
Dedicated engineering support for implementation on Start, and across the whole contract on Scale and Enterprise, with implementation fees included in the published price.
Four quality layers around every integration call: agent quality, simulations, runtime guardrails, and Coach QA. Coach reviews 100% of conversations, human or AI, and assigns a Ticket Quality Score of Good, Warning, or Critical. The Quality Guarantee refunds the AI portion of a badly scored interaction.
SOC 2 Type 2, ISO 27001, HIPAA with a BAA on Scale and above, and GDPR. Runs on Google Cloud with zero-data-retention inference and PHI handled on a minimum-necessary basis. The trust center is public.
Ideal for
Healthtechs whose hardest tickets require a read from one system and a write to another, whose compliance team needs to see the integration scope and the audit posture before go-live, and whose support leaders want to own the workflow logic after launch. The natural fit already runs one of the five Start-tier ticket platforms and has at least one internal API it wants the concierge to act on.
Pricing
Published on the pricing page. Start is $2,100 per month billed annually. Scale is $5,100 per month billed annually. Signature is custom. There are no per-seat charges, implementation fees are included, and you pay only for resolved tickets.
Proof
Lorikeet's published customer results are in fintech and adjacent categories rather than healthtech, so read them as evidence of the integration model rather than of a healthtech deployment. Breeze reported that within 30 days the agent independently resolved 40% of complex support volume, covering KYC reviews and transaction status, built on the workflow builder integrated with existing tools. Flex reported 2x CSAT, 4x rent-week volume, and a 50% shorter median resolution, and its CX AI Product Lead, Lindsay Boland, said: "We tested AI solutions head-to-head and Lorikeet was a winner in every metric." Summ reported 97% faster resolutions during tax time. KYC review and transaction status are the fintech equivalents of eligibility and claim status: identity-gated reads from a system of record.
A real limitation
Lorikeet is not a swipe-a-card product. The Start tier is priced for teams that intend to integrate, and the value is highest when at least one system of record is connected. If your entire support load is answerable from a help center, a lighter tool will get you there for less, and Lorikeet's published customer results are not yet in healthtech.
2. Decagon
Decagon is an enterprise AI agent platform with white-glove implementation. For healthtech, its integration story is custom-built per deployment with embedded engineering. The advantage is an integration shaped around your systems; the cost is that you are buying a services relationship as much as a product, and the integration knowledge sits with the vendor's engineers.
Key integration features
Voice, chat, and email channels in one platform.
White-glove deployment with embedded engineering during launch, including custom integrations into systems of record.
SOC 2 posture; BAA available on enterprise engagements for PHI workflows.
Ideal for
Large healthtech and health-services enterprises with the budget and engineering appetite for a white-glove deployment and custom integrations into their EHR, eligibility, and billing systems.
Pricing
Not published. Contact sales for a custom enterprise contract.
3. Sierra
Sierra is an enterprise AI agent company whose hallmark is outcome-based pricing, and whose integration approach is custom-built per customer. The integration consequence is that every system-of-record connection is a custom engagement, which suits an enterprise with a dedicated integration team and is slower for one that wants a helpdesk and an eligibility API connected in the first month.
Key integration features
Outcome-only pricing: customers pay when the AI fully resolves a case; escalations cost nothing.
Voice, chat, and email channels.
Custom-built integrations and a branded AI Persona deployment model.
High-touch implementation with embedded Sierra staff.
SOC 2 posture; HIPAA handling on enterprise engagements with a BAA.
Ideal for
Large health enterprises that want billing aligned to successful resolutions and have the appetite for an enterprise engagement with custom integration work.
Pricing
Not published. Outcome-based, negotiated on enterprise contracts. Contact sales.
4. Fin by Intercom
Fin by Intercom is the AI agent layered on top of Intercom's messenger and helpdesk. Its integration reach for healthtech comes through the Intercom app ecosystem and custom actions and webhooks to your APIs, rather than purpose-built healthcare connectors. Fin has the most obviously native helpdesk integration of the seven because it lives inside the helpdesk; everything beyond it is a custom action you configure and maintain.
Key integration features
Integrations via the Intercom app store and custom actions and webhooks to your APIs.
Works with Salesforce and HubSpot helpdesks in addition to Intercom.
SOC 2; HIPAA support and a BAA available on qualifying plans.
Ideal for
Healthtechs already on Intercom that want a fast chat concierge for lighter workflows, and that can build deeper system integrations through custom actions.
Pricing
Contact sales. Not restated here; confirm the current rate card and any helpdesk seat requirements directly with Intercom.
5. Ada
Ada is one of the most established AI concierge vendors, with a long enterprise track record. It has expanded from chat into voice and email and supports custom action integrations and webhooks into your APIs and systems. Helpdesk integrations with Salesforce, Zendesk, and other major platforms are mature; depth on regulated system-of-record workflows depends on the custom integration work you commission.
Key integration features
Multi-channel: chat, voice, and email.
Custom action integrations and webhooks into your APIs and systems.
Mature integrations with Salesforce, Zendesk, and major helpdesks.
SOC 2; HIPAA support available for qualifying deployments.
Ideal for
Mid-market and enterprise healthtechs with high inbound chat volume that prefer a long-track-record vendor and will commission custom integrations into their EHR, eligibility, or billing systems.
Pricing
Not published. Custom annual contracts; contact sales.
6. Salesforce Agentforce
Salesforce Agentforce is Salesforce's agentic layer, native to its CRM and Health Cloud data model. For healthtechs already standardized on Salesforce Health Cloud, the integration advantage is real: the agent reaches member, care, and case objects without middleware, and its actions are grounded in Salesforce data and flows. Reaching systems outside the Salesforce estate, such as a separate EHR or an eligibility clearinghouse, still requires MuleSoft or AppExchange connectors and integration work.
Key integration features
Native to Salesforce CRM and Health Cloud objects, with no middleware for data already in Salesforce.
Agent actions grounded in Salesforce data and flows.
MuleSoft and the AppExchange for reaching external systems.
SOC 2; HIPAA support on Health Cloud with a BAA.
Ideal for
Healthtechs and payers already running Salesforce Health Cloud that want an agent grounded in their existing CRM data and are comfortable doing external-system integration through MuleSoft and connectors.
Pricing
Not published as a standalone figure. Per-conversation pricing layered on a Salesforce platform license; contact sales.
7. Cognigy
Cognigy is an enterprise conversational AI platform with particular strength in contact-center voice and IVR integrations. For healthtechs running large phone-based member services, its telephony and CCaaS connector depth is a genuine differentiator, and its connector library plus custom API integrations reach back-end systems. Deep read-write integration into clinical or billing systems of record is integration work you scope and build.
Key integration features
Deep telephony and contact-center integrations (IVR, major CCaaS platforms).
Voice and chat with enterprise orchestration and routing.
Connector library plus custom API integrations into back-end systems.
SOC 2; HIPAA handling on enterprise terms with a BAA.
Ideal for
Healthtechs and health-services organizations with large phone-based member services that need enterprise-grade voice and contact-center integration and will build system-of-record connectors to match.
Pricing
Not published. Custom enterprise pricing; contact sales.
Across all seven, helpdesk integrations are native almost everywhere and system-of-record integrations are custom almost everywhere; Agentforce inside Salesforce and Lorikeet are the exceptions. See Lorikeet run an integration against your systems.
The five integrations a healthtech concierge actually needs
Vendors list integrations by count; buyers should list them by job. There are five jobs a healthtech concierge has to do against external systems, and for each there is a specific meaning of native and a specific way to test it. If a vendor cannot pass the test for a job, that job stays with your human team, whatever the demo showed.
1. Helpdesk (the ticket platform)
What native should mean. The concierge reads the ticket, replies inside the same thread, updates ticket fields and tags, and hands off to a human agent inside the helpdesk with the full transcript and the reasoning attached. Native here means the concierge is a first-class participant in the ticket and the human agent never switches tools to pick up where the AI left off.
How to test it. Open a ticket in your real helpdesk (Zendesk, Intercom, HubSpot, Front, Salesforce, or whatever you run), let the concierge take it, then force an escalation. Check that the ticket lands in the right queue, that the transcript and any actions taken are visible to the agent, and that the agent's reply goes back to the customer through the same channel.
2. Identity and eligibility source
What native should mean. Before the concierge answers anything specific to a person, it verifies who it is talking to against your identity source and checks what that person is entitled to against your eligibility or membership record. That source might sit behind the EHR, a membership platform, or a payer clearinghouse. Native means the verification step is a deterministic gate in the workflow and the eligibility read is a live API call rather than a cached export.
How to test it. Run three scenarios with synthetic records: a member whose coverage is active, one whose coverage lapsed last week, and one who fails identity verification. The concierge should answer the first, explain the second accurately, and refuse the third without leaking any detail about the record. Then time out the API and confirm the concierge says it cannot verify right now rather than guessing.
3. Scheduling system
What native should mean. The concierge reads real availability from the scheduling system, offers slots, and writes the booking, reschedule, or cancellation back to the system of record so the change shows up in the clinic's calendar and the patient portal. Native means write-back with a confirmation from the scheduling system rather than a note on the ticket asking a human to make the change.
How to test it. Book, move, and cancel an appointment through the concierge using a test patient, then look at the scheduling system directly. The slot should be held and released correctly, the appointment should carry the right provider and location, and the patient portal should reflect the change without a manual sync.
4. Billing and claims status
What native should mean. The concierge reads a balance, a payment history, or a claim status from the billing engine and explains it in plain language, and where authorized it takes the next action: sends a payment link, sets up a plan, or logs a dispute. Native means the read is scoped to the verified person and the write is limited to the actions you have explicitly allowed. This is where least-privilege matters most; the guide on least-privilege API and webhook design for AI support covers the permission model in depth.
How to test it. Ask the concierge for a balance on a test account, then ask it to do something it should not be allowed to do (refund, write off, change a claim). It should decline the unauthorized action cleanly and log the attempt.
5. Messaging channel
What native should mean. The concierge is the same agent, with the same workflows and the same integrations, whether the patient arrives on chat, email, voice, or SMS. Native means one workflow engine across channels, so an eligibility check built once works everywhere, and a conversation that starts on SMS and continues by phone carries its context. Webhooks earn their keep here: a claim status change in the billing engine should trigger an outbound message without a human in the loop.
How to test it. Run the same eligibility scenario on chat and on voice and compare the transcripts. The answers should match and both should show the same API calls. Then trigger a webhook from a test system (for example, a mock claim status update) and confirm the concierge sends the right message on the right channel to the right person.
How to test an integration claim in a pilot
Every vendor on this list will say yes to the question of whether they integrate with your systems. The pilot is where yes becomes specific. The five steps below fit a two-to-four week pilot and leave a written record for security and compliance. Lorikeet includes unlimited simulations on every plan; ask other vendors what testing volume their pilot allows.
Write the integration in one sentence per job. For each of the five jobs above, state the system, the read, the write, and the guardrail. Example: reads eligibility from the membership API for a verified member, writes nothing, refuses on verification failure. If the vendor cannot map each sentence to a specific integration or concierge action in the product, the claim is a services promise.
Get the auth model in writing. Ask how the concierge authenticates to each system of record, what scopes it holds, whether credentials are per-customer or shared, and how they are rotated. Native integrations have a documented answer; custom ones often have an engineer's answer, which changes when the engineer does.
Run simulations before real traffic. Generate at least a few hundred synthetic conversations per job, covering the happy path, the edge cases (lapsed coverage, double-booked slot, disputed claim), and the adversarial cases (a caller trying to read someone else's record). Do not go live on a job whose simulation pass rate you have not seen.
Break every system on purpose. During the pilot, make each integration fail: time out the eligibility API, return a malformed scheduling response, revoke a billing credential. The concierge should degrade honestly, tell the person what it cannot do, and hand off with context. A concierge that invents an answer when the system of record is unavailable is a compliance incident in waiting.
Review 100% of pilot conversations, not a sample. Integration failures hide in the long tail. Use whatever QA the vendor offers (Lorikeet's Coach reviews every conversation and scores it Good, Warning, or Critical) and read every Warning and Critical against the integration that produced it. Then decide, job by job, which integrations are ready for production and which stay with humans.
The output of the pilot is a table with five rows (the jobs), each marked native, custom, or not ready, with the evidence attached. That table is your integration contract, and it is the step most procurement processes skip.
Lorikeet vs Decagon vs Ada on healthtech integrations
These three are the most common finalists for a healthtech that wants an agent to take action, and they represent three different integration philosophies. The comparison below is limited to integrations; the end-to-end guide compares them on everything else.
Integration philosophy
Lorikeet treats integrations as product features with plan entitlements: named ticket platform integrations on Start, all integrations on Scale and Enterprise, and custom data integrations plus concierge actions on every plan. Your team configures the workflow that calls the integration, and dedicated engineering support is included for implementation. Decagon treats integrations as an embedded-engineering engagement: custom-built per deployment with white-glove implementation. Ada treats integrations as custom actions and webhooks on top of mature helpdesk connectors, with system-of-record depth as work you commission.
Ticket platform coverage
Lorikeet names its Start-tier ticket platforms publicly: Zendesk, Intercom, HubSpot, Front, and Salesforce. Ada states mature integrations with Salesforce, Zendesk, and major helpdesks. Decagon does not state ticket platform coverage in its public positioning, so ask for it explicitly.
Write-back to systems of record
Lorikeet states that concierge actions read and write to your systems of record and includes them on every plan. Decagon scopes write-back inside custom integration work, so it depends on what your engagement covers. Ada supports custom actions and webhooks into your APIs and systems; whether a given action writes back depends on how the action is built. In a pilot, this is the column to test first for all three.
Testing and quality around integrations
Lorikeet includes unlimited testing and simulations on every plan, reviews 100% of conversations through Coach with a Good, Warning, or Critical Ticket Quality Score, and refunds the AI portion of badly scored interactions under its Quality Guarantee. Decagon and Ada do not state simulation volume or a per-conversation QA score publicly, so ask both what a pilot includes.
Compliance and pricing
All three state SOC 2. Lorikeet adds ISO 27001, HIPAA with a BAA on Scale and above, GDPR, zero-data-retention inference on Google Cloud, and a public trust center; Decagon offers a BAA on enterprise engagements; Ada offers HIPAA support for qualifying deployments. On price, Lorikeet publishes Start at $2,100 per month and Scale at $5,100 per month, billed annually, with no per-seat charges and implementation included. Decagon and Ada are contact-sales.
Where each one wins
Choose Lorikeet for integrations as plan features your own team can configure, published pricing, and QA on every conversation. Choose Decagon for an enterprise, white-glove engagement where the vendor's engineers build and own the integrations. Choose Ada for high chat volume on a supported helpdesk, with custom actions commissioned for the system-of-record work.
Key Takeaways
The five integrations that matter for a healthtech concierge are the helpdesk, the identity and eligibility source, the scheduling system, billing and claims status, and the messaging channel. Test each one by job, not by count.
Helpdesk integrations are native almost everywhere on this list. System-of-record write-back is custom almost everywhere. Lorikeet and Agentforce (inside Salesforce) are the exceptions.
Lorikeet is the only vendor of the seven that publishes both its pricing and its plan-level integration entitlements, and it includes unlimited simulations on every plan for the pilot test above.
A signed BAA on the plan you are buying is the precondition for any of this. Confirm it before the pilot starts.
Conclusion
All seven platforms on this list can take action against a healthtech system, given enough integration work. The question is how much of that work is product and how much is services, who owns it after launch, and whether you can prove it works before a patient depends on it. Lorikeet ranks first on this integrations-only cut because its concierge actions, custom data integrations, and ticket platform connectors are plan features with published pricing, backed by unlimited simulations and 100% QA coverage to test them. For the wider criteria, see the guide to AI support for healthcare in 2026. The other six are credible depending on your stack: Agentforce if you live in Health Cloud, Fin if you live in Intercom, Cognigy if your volume is on the phone, Decagon or Sierra if you want an enterprise engagement, and Ada if chat breadth on a mature helpdesk connector is the priority.
If you are evaluating on integration depth, book a Lorikeet demo and bring one real system of record. We will scope the read, the write, and the guardrail, and run it in a sandbox before you sign.
