For companies · AI Agent

An agent that speaks your brand and resolves on the first turn.

Chat and voice, one brain — account-aware, seven native languages, grounded in your knowledge base, and gated where law or judgment demands a human. One line of JavaScript on your site; updated centrally so you never re-deploy.

Try it on your site Join Waitlist

Paste your URL and watch it answer from your own content — no signup.

Good morning

How can I help?

🇺🇸 English🇪🇸 Español🇫🇷 Français🇩🇪 Deutsch🇧🇷 Português🇮🇹 Italiano🇯🇵 日本語

¿Pueden ayudarme con mi pedido?

¡Claro! ¿Me das el número de pedido o el correo asociado?

What it does

Resolution, not deflection.

First-turn resolution.

The agent has the visitor's context before they finish typing. Issues resolve on turn one — not after the third hand-off.

Always on, every language.

Chat and voice, 24/7, seven native-fluency languages. No off-hours queue, no “leave a message” fallback.

Voice and chat, one brain.

Same agent, same knowledge, both surfaces. Visitors switch mid-conversation without losing the thread.

Branded, never generic.

Your name, your voice, your tone. The agent speaks as your brand — not as a third-party widget pretending to be you.

In real life

Three conversations it wins.

The pharmacy call that never hits a hold queue

A patient calls at 2 a.m. about a refill. The agent recognizes the caller, checks the prescription status, and answers in their language — while anything regulated routes to a human pharmacist by rule, not by guess. The routine work resolves instantly; the judgment calls reach people.

How it works

Plug in your stack. Your agent speaks fluent customer.

Teleperson pulls signal from your contact center, CRM, help desk, and any other system you can name. The agent on your site recognizes the visitor, knows their history, and resolves intent in one turn — not fifteen.

Contact Center

Five9 · Genesys · Zendesk

CRM

Salesforce · HubSpot

Help Desk

Zendesk · Freshdesk · Intercom

Other

Plaid · custom APIs

Plug in your stack

See Starter pricing Join Waitlist

Integrations live at our Starter tier and above.

Everything under the hood

Eleven capabilities, one agent.

Good morning

How can I help?

🇺🇸 English🇪🇸 Español🇫🇷 Français🇩🇪 Deutsch🇧🇷 Português🇮🇹 Italiano🇯🇵 日本語

¿Pueden ayudarme con mi pedido?

¡Claro! ¿Me das el número de pedido o el correo asociado?

Speaks seven languages natively

Not English with a translation layer. Each supported language has a native voice for text-to-speech and a language-pinned transcriber for speech-to-text, so Spanish audio comes back as Spanish text and the reply sounds like a fluent Spanish speaker — not an English speaker reading words.

English, Spanish, French, German, Portuguese, Italian, Japanese
Per-language voice override (ElevenLabs Multilingual v2)
Visitor-selectable in the chat panel header

Live voice calls in the same agent

One assistant, one knowledge base, one set of tools — same brain for chat and voice. The visitor clicks a button and starts talking. No separate voice product to license, no separate agent to maintain.

Real-time WebRTC voice via Vapi
Voice-mode visualizer + interrupt-friendly turn-taking
Visitor can pause, resume, or switch back to chat mid-call

Good morning

Voice concierge

Listening…

Good morning

Adapts to your industry

Show me Jessica for…

All industries🏦 Banking🛍️ Retail⚕️ Healthcare🛡️ Insurance☁️ SaaS

How does Teleperson handle disputes?

For banking: I read the transaction, gather the dispute type (fraud / merchant / billing) and the SAR-relevant details, then route to your fraud queue with a pre- populated case. Want to walk through a sample dispute?

Adapts to your industry

Pick a vertical from the in-panel switcher (or let the visitor self-identify) and the chatbot shifts vocabulary, example flows, and the integrations it foregrounds. Banking gets balance lookups + dispute flows. Retail gets order status + returns. Healthcare gets HIPAA-aligned guardrails.

Six built-in verticals: banking, retail, healthcare, insurance, SaaS, travel
Per-vertical prompt addendum + suggested chip prompts
Auto-detect from visitor's self-introduction via `set_vertical` tool

Books meetings on the spot

Demos, design-partner discovery calls, sales follow- ups — the bot pulls Cal.com availability mid- conversation, narrates options, and books the slot without ever leaving the panel.

Live Cal.com side-drawer embed
Pre-filled with visitor name + email when known
Confirms time zone before booking (no double-booking surprises)

Good morning

Books meetings live

Can we set up a 30 min demo Tuesday?

Tuesday · April 23

10:0010:3011:0013:3014:0015:00

Good morning

Cites your research

What does your white paper say about A2A?

From The Architecture of Agent-to-Agent Commerce: consumer agents will transact directly with brand agents under aligned incentives. Three protocol pieces matter…

The Architecture of A2A Commerce

12 min read · cited in this conversation

Cites your knowledge base

Upload your product docs, FAQs, help-center articles in the dashboard. We chunk and embed them with OpenAI. On every conversation, the chatbot retrieves the top-K most relevant excerpts and grounds its answer there. The visitor experiences it as “the bot just knows our business.”

Sentence-aware chunking, OpenAI text-embedding-3-small
pgvector cosine-similarity retrieval (sub-50ms at scale)
Manage docs from /agent-builder/dashboard/kb — no re-deploy on update

Emails the transcript on request

After ~60 seconds of engagement, the chatbot offers to email the conversation to the visitor. Powered by Resend; arrives with a clean HTML format the visitor can forward to their team.

Audience-aware blurb (consumer / company / general)
Always-on archival to your admin dashboard regardless
Reply-to routes back to a real human inbox

Good morning

Emails the transcript

Want a copy of this conversation?

Drop your email and we'll send the transcript when you're done.

you@yourcompany.com

Good morning

Recap with next steps

Recap

Walked through deflection metrics + a banking-specific pilot scoped to Q2. Discussed SOC 2 + GLBA fit.

Next steps

· Confirm Tuesday 10:30 demo with Sarah
· Share the white-label spec
· Loop in your CX lead before kickoff

Recap + next steps in every transcript email

On top of the raw transcript, every recipient gets an LLM-generated recap — a 1-2 sentence summary, three to five “what we covered” bullets, two to three specific next steps tailored to THIS conversation, and one suggested CTA. Forwardable artifact for the buyer’s internal team.

Claude Haiku 4.6 from the full transcript (~$0.001 per email)
Suggested CTA href allowlisted to your domain
Renders as an accent-tinted card above the message log

Captures CSAT after every voice call

When a voice call ends, the chatbot replaces the “Start a call” button with a five-star rating prompt plus an optional feedback textarea. You see aggregate ratings + raw feedback in your admin dashboard, with each rating linked back to the conversation it came from.

Submitted ratings persist to Supabase with engagement_id
CSAT by locale segmentation (language tagged on every rating)
Soft-fail UI — visitor sees ✓ Thanks regardless of network

Good morning

Rate your experience

"Loved how it adapted to our brand voice."

Good morning

Remembers across visits

Welcome back, Jesse

Last time we talked about your white-label pilot. Want to pick up where we left off?

Yes — where did we leave off?

You were comparing Growth vs Scale tier. We covered 5,000 vs 25,000 conversation caps, and you mentioned your run rate is ~12k/mo.

Remembers visitors across visits

Opt-in device memory keeps the visitor’s name, track, and recent actions for 90 days. Returning visitors get a personalized greeting; the bot picks up where the prior conversation left off without making them re-introduce themselves.

Default OFF — explicit consent dialog before any write
Stored in the visitor's browser only, never on our servers
“Clear memory” button always available

Adapts to where the visitor is — privacy-aware

Jessica picks up the visitor’s country automatically (legitimate-interest framing for regulation + currency hints) and unlocks city / timezone personalization only after the visitor flips on “Remember me.” That same consent gates persistent memory, so one opt-in covers everything personal we touch.

Country-level always on — used silently for GLBA, GDPR, PIPEDA, LGPD, APPI framing
City + timezone unlock with explicit consent (Article 6(1)(a))
Schedules offered in the visitor's local timezone once unlocked
Visitor can override with the `set_visitor_location` tool — visitor-confirmed > IP geo
No server-side persistence — passes through to Vapi for the session only

Good morning

Adapts to your region

Detected · Austin, TX · UTC −6

Country-level always on for regulatory context. City + timezone unlock with your consent (Remember me).

Can we set up a demo this week?

Sure — I’ll suggest times in Central. Wednesday 2 PM or Thursday 10 AM CT both work. Which?

Good morning

Call Jessica directly

Call Jessica

+1 (701) 645-4298

Same agent as the chat — just over the phone.

Call directly — same agent, just phone

Click the phone icon in the chatbot header and you’ll see the number to call. Same Jessica, same knowledge base, same brand voice — just over the phone for visitors who prefer to speak with a real-time agent without typing.

Tap-to-dial on mobile, copy-to-clipboard fallback on desktop
Voice agent answers with your tenant’s brand identity
Optional SMS surface coming soon (A2P 10DLC pending)

Trust & control

Autonomy you can put in front of a regulator.

Bounded authority.

The agent acts within scoped, explicit permission — refund up to a limit, change within policy — and nothing more.

Human gates by rule.

Regulated, identity-sensitive, and payment actions route to a human by design, not model discretion.

Everything on the record.

Every consequential action is logged and reproducible — the audit trail is built in, not bolted on.

Ready when you are.

Three transparent tiers, no per-resolution surprises, full self-service dashboard, cancellable anytime.

See pricing Speak with our team