AI Calling Agent Setup Time India: Timeline

Use one named milestone at a time. Calling all four of these "live" makes every setup-time comparison misleading.

The subject described in this section: This page uses accepted go-live for the point at which both workflow acceptance and the required channel test are complete. Record that definition in the project notes before recording a duration. A vendor that measures only from payment to dashboard access is measuring a different event from a business that measures until its production number handles the first accepted call.

Use a dependency model instead of a universal day count:

Time to accepted go-live = configuration time + dependency wait time + test-and-fix cycles.

• Configuration time is the work of entering business facts, instructions, actions and escalation rules.
• Dependency wait time covers things outside the agent editor, such as access to a calendar, KYC and number activation at the business's phone company, or approval and connection state on a third-party social account.
• Test-and-fix cycles run until the agreed conversations produce the accepted outcome.

The three parts can overlap, so keep a dated event log rather than blindly adding calendar days. For example, web testing can start while number activation at the phone company is pending. Mark the clock stopped only when a named dependency is outside the implementation team's control, and publish both elapsed time and active working time if the distinction matters.

Dvaarik's current direct product is self-serve. The business uses the dashboard and APIs to configure the facts and workflow that the agent is allowed to use:

1. Business identity, operating hours and contact details.
2. Services or products, prices and the boundaries of what may be promised.
3. Staff, locations, availability and booking rules when appointments are involved.
4. Approved knowledge and the response for an unknown or sensitive question.
5. The agent profile, the connected speech and AI accounts, the selected voice and the languages that will actually be tested.
6. The action path: collect details, book, send a configured payment link, log an outcome or escalate.
7. The required channels, the phone company account and one accountable owner for each external account.

The separate AI receptionist onboarding checklist owns the detailed input list. This setup-time page owns the elapsed-time model and acceptance boundary, so it does not turn the same checklist into a second competing guide.

The agent workflow and the delivery channel are separate dependencies.

Do not turn the slowest channel dependency into a claim that the AI itself takes that many days to configure. Equally, do not call the system production-ready because a web test worked while the phone or social path remains untested.

Build the test set from the business's common enquiries and its expensive failure cases. Each row needs an expected outcome, not just a transcript that sounds fluent.

Keep failed cases in the test set after they are fixed. A production acceptance suite is more useful than a one-time demo because the same cases can be rerun after knowledge, prompt, integration or voice changes.

Delay usually comes from an unresolved business decision or external dependency, not from waiting for a model to generate text.

• Prices, operating hours or escalation ownership are still disputed internally.
• Calendar, CRM, payment, speech or AI provider credentials are missing or belong to the wrong account.
• KYC or number activation is still pending at the business's phone company.
• An existing number cannot yet forward to the number in the connected phone company account.
• The WABA or Instagram professional account has an ownership, permission or approval problem.
• Nobody defined the safe answer for an unknown, regulated or sensitive question.
• The test set checks only the happy path and finds failure cases late.
• The connected speech models do not cover the language that the business assumed they did.
• Concurrent-call expectations exceed the phone company account's simultaneous-call capacity.

Put an owner and next action beside each dependency. “AI setup delayed” is not actionable; “clinic owner must approve the cancellation rule” and “number activation pending at the phone company” are.

Use the current product boundary rather than an old implementation quote:

• Dvaarik is self-serve SaaS with no setup fee, no monthly fee and no minimum volume. Signing up is free with no card, and a new account gets ₹50 of free credits.
• Dvaarik charges ₹0.50 for each call minute, in 30-second steps. That is Dvaarik's fee, not the whole cost of a call: the phone company and the speech and AI providers the business connects bill it directly. The current rates are on the pricing page.
• Language coverage follows the speech models the business connects, and a model's language inventory does not prove arbitrary mid-call switching.
• Dvaarik supplies no phone numbers. The business connects its own Plivo, Telnyx or Twilio account and buys or ports its number there, and that account sets how many calls can run at once. KYC and number activation follow that company's process, so this page deliberately does not publish an activation duration.
• WhatsApp and Instagram come with every account and run on the business's connected accounts. Each AI text reply is ₹0.05, and Meta bills WhatsApp message charges to the business directly.
• A result such as a booking or payment link is available only when that exact action is configured and passes acceptance testing.

The subject described in this section: These are planning inputs, not a promise that a business with incomplete data or a blocked external account will be live by a particular date.

A short dated record makes setup time comparable across projects:

1. Start event and accepted-go-live definition.
2. Required channels, languages, connected phone, speech and AI providers, and concurrent-call count.
3. Named owner for business facts, integrations and every external account.
4. Dependency opened, dependency cleared and whether the time was active or waiting.
5. Acceptance-test version, failures, fixes and final approver.
6. First production observation and the rollback or human fallback path.
7. Changes made after go-live and the tests rerun after each change.

Report the result as a measured project fact: “configuration began on X; workflow acceptance passed on Y; phone acceptance passed on Z.” Do not turn one project into a universal setup-time claim for every business, channel or vendor.