Trust

Voice operations you can audit, not just trust.

Agni is built for teams that need control over customer conversations, call data and who can see it. Every call is recorded, transcribed and reviewable, and the calling rules of your market are part of how campaigns run, not a problem left for after go-live.

Access controlsAudit-ready call recordsHuman handoffData handling controls
Operational controls

Controls your security team can verify.

A phone call made by AI should never be a black box. Admins get the permissions, records and review tools to see exactly what every agent said and did.

01
Workspace access and permissions

Decide who can build agents, launch campaigns and open call data. Production calling stays separate from experiments.

02
Complete call records

Every call leaves a recording, a full transcript and the session details: duration, outcome and why it ended. Nothing is summarised away.

03
Data flows you control

Choose where call data goes. APIs and webhooks deliver transcripts, recordings and outcomes into your own systems.

04
Human handoff for sensitive calls

Set the moments an agent must pass the call to a person, such as disputes or vulnerable customers, with the conversation so far attached.

05
Post-call review

Review transcripts, sentiment and outcomes call by call. Flag and correct agent behaviour before the next campaign goes out.

06
Test before you dial

Run test calls against real scenarios before a single customer is called, and check every script change the same way.

Calling compliance

Calling compliance, built into the workflow.

Most platforms leave calling rules to the customer once the contract is signed. We treat list scrubbing, calling hours, disclosure and records as part of how a campaign runs, set to the rules of the jurisdiction you call into.

01
Do-not-call scrubbing before dialing

Contact lists are checked against the do-not-call registries and suppression lists you are bound by before a campaign dials, not after a complaint.

02
Consent logging

The consent basis for each contact is captured and stored, so every dial traces back to a recorded permission.

03
Calling-window enforcement

Campaigns respect the calling hours you configure for each jurisdiction and time zone, including tighter windows for collections work, automatically.

04
AI-call disclosure

Agents can tell the customer at the start of the call that they are speaking with an AI assistant, in the customer's own language.

05
Opt-outs that stick

A customer who asks not to be called again is recorded as a structured outcome and suppressed from future campaigns.

06
Per-campaign audit trails

Who was called, when, under which script and what was said, ready for your internal review or a regulator's questions.

This is a description of product controls, not legal advice. Your obligations depend on where you call, your industry and your use case. Agni provides the controls; your compliance programme decides how they are set.

Data protection

Privacy by design, in every language you call in.

Privacy law differs across the US, the UAE, India and Canada, but the questions are the same: what was agreed, where the data lives, and how a customer changes their mind. The calling workflow is built so those answers are on record.

How we handle data
  • Purpose-scoped campaigns: each campaign is tied to the specific purpose a customer agreed to.
  • Withdrawal on the call: opt-outs are captured as structured outcomes and honoured in later campaigns.
  • Access and correction requests: recordings, transcripts and session details make them answerable.
  • Regional data residency: call data can be kept in the region your regulator or your policy requires.
  • Retention and deletion: terms are agreed per deployment during onboarding and written into your agreement.
  • GDPR and HIPAA: supported, with a data processing agreement available for enterprise contracts.
Consent in the customer's own language

A consent notice only means something if the customer understood it. Most platforms treat that as a translation chore. Agni speaks your market's languages and dialects natively, so the consent request, the AI disclosure and the opt-out all happen in the language the customer actually speaks, and all three are on the recording.

Which languages and dialects are production-ready for your customer base is confirmed in the demo on your own calls, not assumed here.

Certifications

Where we are on certifications, honestly.

We would rather tell you exactly where we stand than hang a badge we cannot back up. Here is the current position, and your security team can test every line of it in evaluation.

SOC 2 In process Controls are being built and evidenced to the SOC 2 standard. We will say so plainly when the report is issued, not before.
HIPAA Supported Deployment options for healthcare workflows that handle protected health information.
GDPR Supported Data processing agreement and data subject request support for customers calling people covered by GDPR.
Data residency Regional options available Call data held in the region your regulator or internal policy requires, scoped per deployment.
Uptime 99%+ SLA Service level agreed per deployment and written into your contract.
Responsible AI

Responsible AI calling, as policy.

Nobody should come off a call feeling tricked by a machine. These are our design defaults and the configuration we recommend to every customer.

AI disclosure

Agents can identify the call as AI-assisted from the first sentence, in the customer's language.

Human escalation

A path to a person can always be configured, and we recommend making it mandatory for sensitive workflows.

Recording transparency

Calls are recorded for quality and audit, and agents can say so wherever your policy or local law requires it.

A route for concerns

If a call powered by Agni raises a concern, it goes to the business that ran the campaign and into our own review as well. Raise it with our team.

Security FAQ

Questions security teams ask us.

Regional data residency is available for enterprise deployments. The exact posture depends on the deployment model and your telephony setup, so we scope it per engagement and document the architecture for your security team before anything goes live.

For as long as you decide. Retention is agreed per deployment at onboarding, records stay available for audit and review for that period, deletion requests are supported, and the terms are written into your agreement.

Yes, a limited set of infrastructure and telephony providers. The list that applies to your deployment, and how you are told about changes, is shared during due diligence.

Yes. It covers processing roles, sub-processors and support for data subject requests. If your organisation has its own standard paper, we are happy to review that instead.

Not yet. SOC 2 is in process, and we will not describe it any other way until the report exists. HIPAA and GDPR are supported today, regional data residency is available, and uptime is covered by a 99%+ SLA agreed per deployment.

Before a campaign dials, lists are scrubbed against the do-not-call and suppression lists you are bound by, calls outside your configured calling window are held, and anyone who opted out on a previous call is excluded. Every dial is logged against its consent basis, so the reason a person was called is always on record.

Please do. We complete it during evaluation and walk your team through the answers live. Book a demo and bring it along.

Bring your security questionnaire.

Bring your security lead and your compliance requirements. We walk through the controls live on the demo and answer the hard questions on the call.