Web chat agents

Web chat that closes the loan, not the ticket.

AI chat agents grounded in your knowledge base. They resolve real work in the portal, closing-day questions, wire instructions, document checklists, and escalate with full context when a person needs to step in.

Norpath Mortgage

OverviewClosingDocumentsPayments

Mara ReyesMR

LoanNMT-2024-08471 30y refi · 6.125%

Cash to close$4,217.18

Closing · Sept 21All milestonesDisclosures

Sept 7Rate locked · 6.125%, 60-day lockComplete

Sept 8Loan Estimate delivered (RESPA)Complete

Sept 12Appraisal received · $498,000Complete

Sept 17Closing Disclosure issued · TRIDComplete

Sept 21Signing · First American Title · 2 PMToday

Norpath Mortgage

Online · responds instantly

Type a message

Same brain. Every channel.

Voice, iMessage, SMS, and chat run the same agent and share the same memory. Customers don’t repeat themselves when the channel changes. The next one picks up where the last one stopped.

Chat is where the work finishes.

Voice opens the file. iMessage moves it along. Chat is where the customer signs, sees the artifact, and confirms, with the same agent that started the conversation.

Drop-in embed

One snippet. Loads after interactive, respects your CSP, scopes itself to the host page. Zero impact on Core Web Vitals.

Knowledge-base grounded

Answers come from your docs, runbooks, and policy library, not a model hallucination. Cite the source, then offer the next step.

Same memory as voice and iMessage

The customer doesn’t re-explain. The chat agent already knows the rate they locked on Monday and the disclosure that went out Wednesday.

Human escalation with context

When a thread needs a person, the handoff includes transcript, intent, and live state. Your agent picks up mid-conversation, not from zero.

Or write it the way you write everything else.

REST, JavaScript, the CLI, or MCP for any AI editor. Same API surface across every channel.

1
2
3
4
5
6
7
8
curl -X POST https://api.bland.ai/v1/calls \
  -H "Authorization: $BLAND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "phone_number": "+14155552120",
    "task": "Qualify refi lead",
    "voice": "june"
  }'

Built to pass your security review.

Chat handles regulated transactions in the same trust posture as voice. Encryption, residency, identity, audit all wired in.

CSP-safe embed

No third-party cookies. No off-host data exfiltration.

SSO / SAML

Agent handoff with verified identity. RBAC across orgs.

MFA enforced

On every production system, every privileged action.

Encryption

AES-256 at rest. TLS 1.3 in transit. HSM-backed keys.

KB scoping

Per-tenant document isolation. Sources cited in responses.

PII handling

Redaction in transcripts. Configurable retention.

HIPAA + SOC 2

BAA available. SOC 2 Type II audited annually.

Audit trail

Every escalation logged with full context.

Visit Trust & Security Portal

Frequently asked questions

The widget loads lazyOnload and is CSP-safe. No render-blocking scripts, no layout shift, and zero impact on Lighthouse scores.

It refuses to fabricate. Out-of-scope questions trigger a structured escalation with full context instead of a guess.

Yes. Bidirectional sync with Zendesk, Intercom, Salesforce Service Cloud, and any REST API. Tickets created automatically when the thread needs a human.

Same agent across channels. Customer info ties to the contact, not the channel. The chat thread on Friday remembers the call on Monday.

Build AI that people actually want to talk to

Deploy production-grade voice agents in 30 days. Book a demo to get started today.

Book a demo