The 10 best AI concierge platforms for voice and SMS customer support in 2026, ranked: Lorikeet vs Sierra vs Decagon, Ada, Fin and more, with pricing.

Steve Hind
·
Updated
Quick answer: the AI concierge platforms for voice and SMS customer support worth shortlisting in 2026 are Lorikeet, Sierra, Decagon, Ada (combining with Dialpad, announced Nov 2026), Fin (formerly Intercom, now part of Salesforce), Forethought (now part of Zendesk), Goodcall, Salesforce Service Cloud Voice plus Marketing Cloud, Twilio Flex AI, and Zendesk AI agents, and Lorikeet ranks first because it runs chat, email, voice, SMS and WhatsApp from one agent that preserves context across channels, resolves tickets end to end with deterministic steps where money or compliance is involved, and publishes its plan and per-resolution pricing.
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 customer starts a refund request on SMS at 11pm, calls back the next morning, and the voice agent has no idea who they are or what they already shared. The handoff burns trust faster than any hallucination. Many vendors selling AI concierge platforms for voice and SMS customer support ship one channel well and stitch the others on top with bridges and transcripts.
Voice also fails differently from chat. In chat, a one-second delay is invisible. On a call, a pause before every reply is the difference between a natural conversation and a hang-up. Vendors sell transcription accuracy and language lists; those sit downstream of the two things that decide whether voice works: how fast the agent responds, and whether the same agent carries context across the call, the text thread, and the chat window.
This guide ranks the 10 platforms CX buyers most often shortlist in 2026, scored on channel parity, context preservation, action execution, and outcome pricing, with a head-to-head of Lorikeet, Sierra, and Decagon and a walkthrough of one action chain across voice, SMS, and chat.
What buyers evaluate in AI concierge platforms for voice and SMS customer support
The procurement bar for AI concierge platforms for voice and SMS customer support has moved past deflection rate. Today the questions are architectural:
Channel parity. A workflow built once should run on voice, SMS, chat, and email without separate builds per channel.
Context preservation. If a customer texts and then calls, the voice agent should already know about the text.
Voice latency. Time from the caller finishing to the agent starting, including recognition, reasoning, and synthesis. Sub-second is the threshold for a conversation that feels human.
Action execution. The agent should process a refund, file a claim, or update a record on any channel with the same guardrails.
Quality control. Something has to review what the AI said on every call and text, with a consequence when it gets one wrong.
Auditability and operating model. Replayable logs that span channels, and clarity on whether your team or the vendor owns configuration changes.
Pricing alignment. Whether you pay for resolved outcomes, for seats, or for minutes.
Quick comparison table
Platform | Channels | Cross-channel workflow parity | Pricing | Best for |
|---|---|---|---|---|
Lorikeet | Chat, email, voice (US, UK, AU), SMS, WhatsApp | Yes, one agent, operator-owned | Start $2,100/mo or Scale $5,100/mo (annual), plus $0.99 or $0.90 per resolution; Signature custom. | Regulated B2C and B2SMB scale-ups |
Sierra | Voice, SMS, WhatsApp, chat, email, ChatGPT | Yes, single agent; partner-led | Not published; outcome-based | Large consumer brands and enterprise |
Decagon | Chat, email, voice (SMS not listed) | One agent across its channels | Not published; quote-based | Large enterprises with budget for a premium deployment |
Ada | Voice, messaging, email | One agent across channels | Not published; quote-based | Teams wanting a packaged agent from a long-established vendor |
Fin | Chat, email, voice, SMS, WhatsApp | One agent; works with Intercom, Zendesk, Salesforce | $0.99 per outcome; Fin Voice custom | Teams wanting a self-serve, published-price agent |
Forethought (now part of Zendesk) | Chat, email, voice, Slack | Autoflows across channels (voice on Professional and up) | Not published; quote-based | Zendesk-standardized teams |
Goodcall | Voice (primary), SMS (light) | Voice-first | Contact sales | SMB voice automation |
Salesforce SCV + Marketing Cloud | Voice, SMS, email, chat (two products) | No, separate clouds | Contact sales | Salesforce-native enterprises |
Twilio Flex AI | Voice, SMS, chat, messaging (build-your-own) | If you build it | Contact sales | Engineering-heavy teams |
Zendesk AI agents | Messaging, email, voice (Zendesk voice) | Inside Zendesk | Per agent seat plus $2.00 per automated resolution | Existing Zendesk customers |
Lorikeet, Sierra and Fin each run voice, SMS, chat, and email through one agent. Lorikeet and Fin publish rates; Lorikeet's are on its public pricing page.
How we selected these platforms
We started with the vendors buyers most often shortlist in bake-offs for voice plus SMS, then added those AI search engines return for the prompt AI concierge platforms for voice and SMS customer support. Each has production voice plus a messaging channel, orchestration beyond scripted phone trees, or real-time action execution. Voice-first vendors (PolyAI, Replicant, Bland, Retell, Vapi) are covered in our voice AI for multi-step workflows guide. Competitor descriptions are limited to capabilities the vendors describe publicly; we do not reproduce third-party pricing, funding, or customer counts.
What is an AI concierge platform for voice and SMS
An AI concierge platform for voice and SMS is software that handles phone calls and text messages autonomously, resolving requests end to end while maintaining shared context across channels. Unlike a voicebot that reads a script or an SMS autoresponder that fires templates, a concierge reasons over the customer's history, calls helpdesk and backend systems to take action, and switches between voice and text without losing the thread. Chatbots deflect questions. Concierges resolve them on whichever channel the customer picks.
The category splits around two questions. First, latency: whether the agent responds to speech fast enough that the conversation feels natural. Second, channel unification: whether voice runs on the same workflow engine as SMS and chat with one memory of the customer, or whether they are separate products stitched together by a transcript handoff. The second camp forces customers to repeat themselves when they move from a text to a call.
The 10 best AI concierge platforms for voice and SMS customer support in 2026
1. Lorikeet
Best for: regulated B2C and B2SMB scale-ups that need one agent across chat, email, voice, SMS and WhatsApp, operated by their own team, with quality review on every conversation.
Lorikeet is an AI concierge platform built for complex, regulated customer support: fintech, healthtech, insurance, and telehealth businesses where a wrong answer on a recorded call is a regulator problem. Five channels are live: phone, SMS, chat, email and WhatsApp, inbound and outbound. Voice numbers are available in the US, UK, and Australia, and the Lorikeet voice agent uses model switching to balance low latency with precision and handles interruptions, corrections and callbacks, so callers do not talk over it or assume the line dropped. The architectural decision that sets Lorikeet apart on this list is that the same workflow runs across every channel. A workflow built for chat refunds today runs on the voice line tomorrow with the same guardrails, integrations, and audit trail. Sierra and Fin also run one agent across voice and messaging; Lorikeet's difference is that your own support team authors and ships the workflows.
The workflow layer is two-tiered. Natural-language workflows describe what the agent should do in plain English, so a support lead can author and maintain them. Deterministic structured workflows handle steps that must happen in a fixed order with validation, such as identity verification before a balance is read out. The natural-language agent calls the structured workflows as tools: flexibility in conversation, determinism where money or compliance is involved. Lorikeet integrates with Zendesk, Intercom, HubSpot, Front, and Salesforce, so the concierge reads and writes to the helpdesk your human team already uses.
Quality is a large part of the product. Four layers sit around every conversation: agent quality controls at authoring time, simulations before a change ships, runtime guardrails that steer or escalate live, and Coach, the QA layer that reviews 100% of conversations, human or AI, and assigns each a Ticket Quality Score of Good, Warning, or Critical. Because Coach scores every ticket rather than a sample, a compliance team can pull every Critical call from last week and review it. Under the Quality Guarantee, a resolution you rate bad is not charged. Details are on the quality assurance page.
Key features:
One agent across chat, email, voice, SMS and WhatsApp that preserves context across channels
Voice on local phone numbers in the US, UK, and Australia
Natural-language workflows plus deterministic structured workflows, called as tools in one conversation
Integrations with Zendesk, Intercom, HubSpot, Front, and Salesforce
Four quality layers: agent quality, simulations, runtime guardrails, Coach QA on 100% of conversations
Quality Guarantee: a resolution you rate bad is not charged
SOC 2 Type 2, ISO 27001, HIPAA with BAA, GDPR; Google Cloud; public trust center
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 (3-minute average). 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.
Customer proof: Chexy, where half or more of a month's tickets arrive in the first week, handles 5x more support tickets with the same team. Flex saw 2x CSAT versus its previous tool, saw 4x chat volume during rent week versus the rest of the month, and cut median conversation duration by 50%. Lindsay Boland, CX AI Product Lead at Flex: "We tested AI solutions head-to-head and Lorikeet was a winner in every metric." Breeze had 40% of its complex volume resolved independently within 30 days, with over 90% resolution on the tickets the concierge chose to handle.
Limitations: Lorikeet is purpose-built for complex, regulated workflows, so a very small team that only needs a FAQ voicebot will find it more platform than it needs, and the plan fees make it more expensive than pure pay-per-outcome tools at very low volume. Expect more setup than switching on an FAQ bot: workflows, guardrails and simulations are configured per use case. Lorikeet does not hold a PCI DSS certification; card data is tokenized and never stored, so if you need a PCI-certified vendor to take card payments on calls or texts, check that first. If your primary channel is a messaging app other than WhatsApp or SMS, confirm support before you buy.
2. Sierra
Best for: large consumer brands and enterprises that want a partner-led, outcome-priced omnichannel concierge.
Sierra, founded by Bret Taylor and Clay Bavor, is a common first consideration for large enterprise omnichannel concierge work. It runs chat, SMS, WhatsApp, email, voice and ChatGPT through a single agent, with a native voice agent tested through Voice Sims. Where it differs from Lorikeet is the operating model: deployments are partner-led with Sierra's team closely involved, alongside a developer Agent SDK and a no-code Agent Studio. For enterprises that do not want an AI ops team, that is a feature. Scale-ups that want their own team in control of every workflow change should ask how changes are made after launch.
Sierra pioneered outcome-based pricing for AI agents. Its compliance list is one of the broadest in the category, including SOC 2, HIPAA, PCI DSS Level 1, FedRAMP High, ISO 27001 and ISO 42001, and its testing tooling (Simulations, Voice Sims, Monitors and Experiments) is mature. Deployments are partner-led.
Pricing: Sierra does not publish pricing; it is quote-based and outcome-based, typically sold as an annual enterprise contract.
3. Decagon
Best for: large enterprises with budget for a premium sales-led deployment that want strong chat, email and voice plus experimentation tooling.
Decagon is the platform most often paired with Lorikeet and Sierra in bake-offs. It runs chat, email and Decagon Voice AI agents, with workflows written as natural-language Agent Operating Procedures (AOPs) that run through pre-built integrations. Watchtower always-on QA, simulations at scale and live A/B experiments are strengths, and its security page lists SOC 2 Type II, ISO 27001, HIPAA and PCI DSS 4.0.1. Pricing is customer-selectable between per-conversation and per-resolution.
For this guide's use case, the gap to check is SMS: Decagon does not list SMS on its homepage, so confirm SMS support if text is a primary channel. Fintech, healthtech, and insurance teams should also test cross-channel audit trails and multi-system action chains in their own environment.
Pricing: Decagon does not publish pricing; it is quote-based and typically sold as an annual enterprise contract.
4. Ada
Best for: teams that want a packaged AI agent from a long-established vendor, with Dialpad telephony to come.
Ada is one of the longest-running AI-native names in customer service, with one agent across voice, messaging and email and broad multilingual coverage. On 4 Nov 2026 it announced a combination with Dialpad, expected to close around 12 Nov 2026, which is expected to pair Ada Voice with Dialpad's voice network. Ada says its agent has the tools to complete customer tasks through open APIs and integrations. For serious voice and SMS volume, test one workflow across channels in your own environment and ask how the Dialpad combination changes the roadmap.
Pricing: Ada does not publish pricing; it is quote-based and typically sold as an annual enterprise contract.
5. Fin
Best for: teams on Intercom or another helpdesk that want a self-serve, published-price agent across chat, email, voice and messaging.
Fin (formerly Intercom, now part of Salesforce) runs across chat, email, voice, SMS, WhatsApp and other messaging, and works with the Intercom helpdesk as well as Zendesk, Salesforce and others. Procedures handle multi-step actions, connecting to systems such as Stripe or internal tools through Data Connectors or MCP, and its QA layer includes Simulations, CX Score and Monitors. A 14-day free trial with unlimited outcomes makes it fast to test. Fin Voice is available to select customers through sales, so confirm voice availability and price before you plan around it.
Pricing: $0.99 per outcome with a 50-outcome monthly minimum alongside your current helpdesk, or from $0.99 per outcome plus $19 per helpdesk seat a month with the Intercom helpdesk; Fin Voice is custom-priced.
6. Forethought (now part of Zendesk)
Best for: Zendesk-standardized teams that want a multi-agent suite for triage, agent assist, resolution and QA.
Forethought, acquired by Zendesk in March 2026 and now sold as Forethought AI agents by Zendesk (still available to non-Zendesk customers), offers a multi-agent suite: Solve, Triage, Assist, Discover and QA. Its agents take action through Autoflows and custom actions, and its QA agent scores 100% of human agent tickets. Channels are chat, email, voice and Slack, with voice from the Professional plan. SMS is not listed, so for SMS-heavy concierge work it is a partial fit.
Pricing: Forethought does not publish pricing; it is quote-based, with a blend of platform access fees and outcome-based pricing.
7. Goodcall
Best for: SMB and lower-mid-market voice automation with lighter SMS as an adjunct.
Goodcall is a voice-first platform with strong content positioning around fintech and SMB voice automation. It handles inbound voice well at SMB scale, with usage-based pricing and a quick path to first call. SMS is a lighter adjunct rather than a peer to voice, so it is a strong voice option with a thinner messaging story. We include it because it routinely surfaces in voice-focused buyer searches.
Pricing: usage-based; contact sales.
8. Salesforce Service Cloud Voice plus Marketing Cloud
Best for: Salesforce-native enterprises that want voice through Service Cloud and SMS through Marketing Cloud, accepting the two-product split.
Salesforce's omnichannel story spans two products: Service Cloud Voice for voice, and Marketing Cloud (or Data Cloud plus Digital Engagement) for SMS, email, and chat. Both are mature with deep CRM integration. The constraint is that they are two products, and the AI layer (Agentforce) has to be designed across both, so workflow parity requires careful design across two clouds. Lorikeet integrates with Salesforce, so Salesforce customers can keep their CRM of record and run the concierge on a different layer.
Pricing: per-seat licensing per product; Agentforce is published at $2 per conversation or Flex Credits at $500 per 100K credits.
9. Twilio Flex AI
Best for: engineering-heavy teams who want to build a custom omnichannel concierge on the Twilio stack.
Twilio Flex is the programmable contact center platform behind a meaningful share of voice and SMS support globally, and Flex AI adds orchestration primitives for AI agents. Channel breadth is excellent: voice, SMS, chat, email, and messaging apps all run on Twilio infrastructure with direct carrier relationships. The constraint is the one Twilio has always had: Flex AI is a build-your-own toolkit rather than a turnkey concierge. Strong engineering teams can build something bespoke; buyers who want a first AI conversation in weeks should budget for the engineering build.
Pricing: usage-based telephony plus seat fees; contact sales.
10. Zendesk AI agents
Best for: existing Zendesk customers who want AI agents, Copilot and QA inside their helpdesk across messaging, email and voice.
Zendesk's AI suite (AI agents, Copilot, and Zendesk QA, formerly Klaus) runs inside the Zendesk helpdesk across messaging, email, and voice. AI agents take actions through procedures written in natural language, and Zendesk QA evaluates every interaction. Voice AI agents require Zendesk voice, which is included in Suite Team and Professional, with add-ons for numbers and SMS. For the Zendesk install base it is the easiest path to AI agents: bundled, native, data model already wired, and Forethought AI agents now extend it to other helpdesks. Teams who want voice on Genesys or Twilio while keeping Zendesk as ticketing typically need a separate AI vendor. Lorikeet integrates with Zendesk natively, so Zendesk customers can keep the helpdesk and run voice and SMS on Lorikeet.
Pricing: Suite Team $55 and Suite Professional $115 per agent a month (billed annually), plus $2.00 per automated resolution pay-as-you-go, with lower rates on committed volumes.
Detailed feature matrix
Capability | Lorikeet | Sierra | Decagon | Ada | Fin | Forethought | Goodcall | SF SCV + MC | Twilio Flex AI | Zendesk AI agents |
|---|---|---|---|---|---|---|---|---|---|---|
Voice + SMS + chat + email | Yes | Yes | SMS not listed | Voice, messaging, email | Yes | SMS not listed | Voice-led | Two products | Build it | SMS as add-on |
Single workflow across all channels | Yes | Yes | One agent | One agent | One agent, any helpdesk | Autoflows | Voice-led | No | If you build it | Inside Zendesk |
Runtime guardrails | Yes | Not stated | Not stated | Monitoring for safe responses | Not stated | Not stated | Not stated | Einstein Trust | Custom build | Policy controls |
QA on 100% of conversations, no charge for bad resolutions | Yes (Coach, Quality Guarantee) | Monitors; not stated | Watchtower QA; not stated | Not stated | CX Score on all text conversations; not stated | QA on 100% of human tickets; not stated | Not stated | Not stated | Build it | Zendesk QA on every interaction; not stated |
Operator-owned (not managed service) | Yes | Partner-led; Agent Studio available | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
Pricing model | Plan fee plus per resolution, published | Outcome-based | Per conversation or per resolution | Quote-based | Per outcome, plus seats with Intercom helpdesk | Platform fee plus outcome-based | Usage | Per seat | Usage plus seat | Per seat plus per resolution |
Cross-channel audit log | Yes | Agent Traces | Not stated | Not stated | Not stated | Inside helpdesk | Voice only | Salesforce | Build it | Inside Zendesk |
Lorikeet vs Sierra vs Decagon for voice and SMS
These three meet most often in competitive evaluations for voice and SMS, and they are different products for different buyers. Sierra is the partner-led option for large enterprise, Decagon the premium option with strong testing and experimentation tooling, and Lorikeet the operator-owned option for regulated scale-ups that need one workflow across every channel and quality review on every conversation.
Dimension | Lorikeet | Sierra | Decagon |
|---|---|---|---|
Channels | Chat, email, voice (US, UK, AU), SMS, WhatsApp | Chat, SMS, WhatsApp, email, voice, ChatGPT | Chat, email, voice; SMS not listed |
One workflow across voice and SMS | Yes, one agent, context preserved | Yes, single agent; partner-led build | Confirm SMS support |
Voice latency | Described qualitatively; measure on your call flows | Not stated | Not stated |
Who owns configuration | Your team | Partner-led; Agent SDK or no-code Agent Studio | Your team, natural-language AOPs |
Quality review | Coach scores 100%; Quality Guarantee | Simulations, Voice Sims, Monitors | Watchtower always-on QA, simulations, A/B experiments |
Pricing | Start $2,100/mo or Scale $5,100/mo (annual), plus $0.99 or $0.90 per resolution; Signature custom. | Outcome-based; quote-based | Per conversation or resolution; quote-based |
Lorikeet vs Sierra
Both run voice and SMS on one agent layer with shared context. The difference is who holds the keys. Sierra deployments are partner-led, though Agent Studio gives teams a no-code builder, so ask how workflow changes are made after launch. On Lorikeet, your support lead edits the natural-language workflow, runs it through simulation, and ships it. For a scale-up that changes policy weekly, that compounds. On billing, Lorikeet publishes its rates and pairs them with a Quality Guarantee, so a resolution you rate bad is not charged.
Lorikeet vs Decagon
Decagon is strong on polished chat, email and voice experiences with mature experimentation tooling. Lorikeet is built for where workflows go deep: identity checks before an action, multi-system action chains, and audit trails that follow one customer from a text to a call. Decagon does not list SMS on its homepage, so confirm SMS support if text is a primary channel. Lorikeet's deterministic structured workflows exist for exactly those teams, and Coach reviews every conversation rather than a sample.
Sierra covers more channels through a partner-led model and has the broader compliance list. Decagon brings strong QA and experimentation tooling for chat, email and voice. A very large consumer brand that wants a heavily resourced vendor partnership leans Sierra; a large enterprise that wants Decagon's AOP and experimentation tooling leans Decagon. A regulated scale-up that needs voice and SMS from one agent, published pricing, and 100% QA coverage is better served by Lorikeet than by either.
How to run one action chain across voice, SMS and chat
AI agents that run action chains over voice and SMS are the point of this category. A concierge that can only answer questions is a voicebot with a phone number. The test of a platform is whether one action chain, built once, runs the same way on every channel. Here is how a refund chain is assembled on Lorikeet, and what to ask any vendor to demonstrate.
Start with the outcome, not the channel. Write the chain as a natural-language workflow: verify the customer, look up the order, check policy, issue the refund, confirm. The workflow never mentions voice or SMS; channel is a property of the conversation.
Wrap the risky steps in deterministic structured workflows. Verification and the refund run as structured workflows with validation on every input, so the refund cannot fire before verification passes. The helpdesk (Zendesk, Intercom, HubSpot, Front, or Salesforce) and billing system are wired once at the workflow level, so voice and SMS hit the same endpoints with the same logging.
Simulate before you ship. Run the chain against realistic scenarios, including failure paths: the lookup times out, the customer fails verification, the refund API errors mid-chain. Runtime guardrails decide what happens on each: retry, steer, or escalate with full context.
Let the channel shape delivery, never logic. On voice, the agent reads back the amount and asks for spoken confirmation. On SMS, it sends the amount and waits for a reply. On chat, it shows a card. The step is identical; only the rendering changes.
Preserve context across the handover. Shared state sits at the customer identity layer, not the channel layer, so a customer who starts on SMS at 11pm and calls the next morning is picked up mid-chain by the voice agent.
Score every run. Coach reviews the completed conversation on every channel and assigns a Ticket Quality Score; a Critical score flags the workflow for a fix, and under the Quality Guarantee a resolution you rate bad is not charged.
In a bake-off, ask each vendor to build that chain once and run it on voice, SMS, and chat with the same guardrails and audit record; separate builds per channel means three products glued together. For more, read our guide to voice AI for multi-step workflows.
Multi-step AI support for scale-ups in North America
Multi-step AI support for scale-ups in North America has requirements a generic chatbot evaluation misses. Scale-ups in fintech, healthtech, insurance, and consumer subscription hit the same five constraints, and the vendor list narrows fast once you apply them.
US phone numbers on day one. Voice has to run on numbers your customers recognize. Lorikeet provisions US numbers directly, with UK and Australian numbers for teams that expand.
Consent-first SMS and voice. US messaging rules put the burden on the sender to hold consent and honor opt-outs. Any proactive contact should be consent-first, with consent state tracked in the customer record and enforced by the workflow.
HIPAA with a signed BAA for healthtech. A vendor that says HIPAA-aligned without offering a Business Associate Agreement is not a healthtech vendor. Lorikeet signs a BAA and holds SOC 2 Type 2 and ISO 27001.
Helpdesk continuity. Scale-ups have already picked Zendesk, Intercom, HubSpot, Front, or Salesforce. The concierge has to read and write to that system rather than replace it.
Pricing that survives a spike. Per-seat pricing punishes a scale-up for growing. Per-resolution pricing on top of a fixed plan fee means a rent-week or tax-time spike costs what it resolves.
The customer stories that map to this profile are Flex (4x chat volume during rent week, 2x CSAT, 50% shorter median conversations) and Chexy (5x more tickets handled by the AI agent each month, with half or more of a month's tickets arriving in the first week). Both are multi-step, spike-driven, regulated workloads, the shape most North American scale-ups are dealing with.
How to choose an AI concierge platform for voice and SMS
Buying voice and SMS AI is different from buying chat. These criteria separate a real voice-and-SMS concierge from a voicebot with a phone number.
Channel parity is the load-bearing question. Build a 20-step workflow that spans voice, SMS, and chat. Configure it once and ask the vendor to run it across all three with the same logic, guardrails, and integrations.
Test context preservation explicitly. Start on SMS, drop off, then call 10 minutes later. The voice agent should know who you are and where you left off.
Measure latency under load, not in a lab. Ask for round-trip voice latency on real telephony numbers during a busy hour. A platform that quotes transcription accuracy but dodges latency is optimizing the wrong metric.
Demand an action on a live call. Have the agent issue a refund without a human handoff, and ask what happens when the backend errors mid-action.
Ask who reviews the conversations. Sampling a few percent of calls is how disclosure misses slip through. Ask whether QA covers every conversation on every channel and what happens when the score is bad. Lorikeet's answer is Coach, 100% coverage, and no charge for a resolution you rate bad.
Confirm consent controls for anything proactive. If you will ever proactively text or call customers, ask how consent, opt-outs, and contact-hour rules are enforced, and whether the controls live in the guardrails or with you.
Audit and compliance need to span channels. Ask for a single cross-channel audit log with replay, and confirm the certifications you need (SOC 2 Type 2, ISO 27001, HIPAA with BAA, GDPR) on a public trust center.
Implementation checklist
Pre-purchase: document the channels you need at launch versus phase 2; map your top 20 journeys and the channels they touch; list the systems the AI will read from and write to; list regulatory requirements (HIPAA, GDPR, US messaging consent rules); build a 50-question policy eval covering your hardest scenarios.
Evaluation: run the policy eval on each vendor's voice and SMS channels; verify barge-in, latency, and turn-taking on real numbers; drop a conversation on SMS and pick it up on voice; run one action chain on all three channels and compare the audit records; get written commitments on time to first production conversation and QA coverage.
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 a voicebot or SMS autoresponder for simple FAQs, a lighter tool will be cheaper and faster to switch on.
If you are standardized on Zendesk and want your AI agents bought, billed and supported through Zendesk, Zendesk AI agents and Forethought fit that.
If you want a self-serve, published-price agent you can switch on inside Intercom or alongside your helpdesk within days, Fin is built for that.
If you handle a few hundred tickets a month, the Start plan (from $2,100 a month) will be more than you need.
Where Lorikeet fits
Lorikeet is built for regulated B2C and B2SMB scale-ups: one workflow across chat, email, voice, SMS and WhatsApp with shared state, guardrails, and audit trail; voice on US, UK, and Australian numbers; Coach on 100% of conversations with a Quality Guarantee; published plan and per-resolution pricing.
Chexy: 5x more support tickets handled with the same team, through a first-week-of-the-month spike. Flex: 2x CSAT versus the previous tool, 4x chat volume during rent week, 50% shorter median conversations, and a head-to-head win in every metric. Breeze: 40% of complex volume resolved independently within 30 days, with over 90% resolution on the tickets the concierge chose to handle. Read the Chexy story and the Flex story.
If you are evaluating AI concierge platforms for voice and SMS customer support, book a Lorikeet demo and bring your hardest calls and text threads. Full rates are on the pricing page.
Frequently asked questions
What is an AI concierge platform for voice and SMS customer support?
An AI concierge platform for voice and SMS customer support is software that handles phone calls and text messages autonomously using one configuration layer. It answers questions grounded in your knowledge base, executes actions in your helpdesk and backend systems (refunds, account changes, claims), carries context from a text thread into a call, and hands off cleanly to a human when needed. The difference from a chatbot or voicebot is action execution with the same logic running on both channels rather than separate builds per channel.
Which AI concierge platforms support both voice and SMS through one agent?
Of the ten platforms in this guide, Lorikeet, Sierra and Fin run voice, SMS, chat, and email through one agent. Lorikeet is operator-owned, publishes its plan and per-resolution pricing, and reviews 100% of conversations with Coach; Sierra is partner-led with quote-based outcome pricing; Fin publishes per-outcome pricing, with Fin Voice available to select customers through sales. Ada covers voice, messaging and email; Decagon and Forethought (now part of Zendesk) do not list SMS; Zendesk AI agents run inside Zendesk with SMS as an add-on. Salesforce splits voice and SMS across two products, Twilio Flex AI requires you to build the orchestration, and Goodcall leads on voice.
How does Lorikeet compare to Sierra and Decagon for voice and SMS?
Sierra is the partner-led option for large enterprise, with the widest channel list, a broad compliance list and quote-based outcome pricing. Decagon is the premium option with natural-language AOPs and strong QA and experimentation tooling across chat, email and voice; it does not list SMS. Lorikeet is the operator-owned option for regulated scale-ups: one workflow across chat, email, voice, SMS and WhatsApp, natural-language plus deterministic structured workflows, Coach QA on every conversation, a Quality Guarantee, and published pricing: Start $2,100/mo or Scale $5,100/mo (annual), plus $0.99 or $0.90 per resolution; Signature custom.
How much does an AI concierge for voice and SMS cost in 2026?
Most vendors in this category do not publish rates and quote per customer. 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. Fin publishes $0.99 per outcome; Zendesk charges $2.00 per automated resolution on top of agent seats; Sierra is outcome-based and quote-based; Goodcall and Twilio Flex AI are usage-based; Salesforce is per seat, with Agentforce at $2 per conversation.
How does cross-channel context preservation work in an AI concierge?
Context preservation means that when a customer starts on SMS and continues on voice, the agent on the second channel knows what was said on the first. The mechanism is shared state at the customer identity layer rather than the channel layer, with the in-progress workflow step attached to the customer, not the conversation. The simplest evaluation is to drop a conversation mid-SMS and call the support line 10 minutes later. If the voice agent does not pick up where you left off, the context preservation is marketing rather than architecture. On Lorikeet, one agent preserves that context across chat, email, voice, SMS and WhatsApp.
Can an AI concierge take actions during a phone call or text thread?
Yes, if the platform is built for it. A concierge that only answers questions is a voicebot. On Lorikeet, the agent verifies identity, looks up the order, checks policy, issues the refund, and confirms on whichever channel the customer is on, with the risky steps running as deterministic structured workflows that validate every input. Every step is logged, Coach scores the completed conversation, and under the Quality Guarantee a resolution you rate bad is not charged. In a bake-off, ask each vendor to demonstrate a live action on a call without a human handoff, and to explain what happens when a backend system errors mid-chain.

