Who this voice bot guide is for
If your BPO in Noida, Mohali, Ahmedabad or Kolkata runs a US process at night, a voice bot is probably already on your list of things to try. This guide is for the process owner or ops head who has to make it work: what a voice bot actually is, inbound against outbound, how one sets up on your dialer, and what to test in a pilot before you take results to your client.
We are a Pakistani company and we sell into the same kind of floors in India, so the shift, the dialers and the client pressure will be familiar. Nothing here depends on buying from us.
What is a voice bot? The three kinds you will be offered
"Voice bot" covers three very different things, and vendors do not always say which they sell. The quickest way to tell is to interrupt their demo on your own phone: a clip bot carries on or plays the wrong line, while a conversational bot stops, answers and goes back to where it was.
- IVR. Press 1 for this, press 2 for that. Useful for routing inbound calls, useless for qualifying a cold lead.
- Recorded-clip bots. A soundboard of pre-recorded lines triggered by keywords. Cheap, and they break as soon as the caller says something the clip list did not expect.
- Conversational AI voice bot. It listens, understands the answer, replies in a natural voice and can be interrupted. This is the kind that can actually front or verify on a US campaign.
Inbound vs outbound: which does your process need?
An inbound voice bot answers calls: order status, appointment changes, first-line support, routing to the right team. An outbound voice bot works the other direction. It takes calls your dialer connects to leads, reads the opener, qualifies and transfers to a human, or verifies a sale on a recorded line.
If you run both kinds of process for the same client, start with the outbound one. Outbound fronting has a narrow script and a clear success measure, billable transfers, which makes a pilot easy to judge. Inbound calls vary more, so the script takes longer to agree and the results are harder to compare.
Most US lead-generation processes run out of Indian BPOs (Medicare, ACA, final expense, auto, solar, debt relief) are outbound, and that is what our bots are built for: a fronter that qualifies and warm-transfers, and a verifier that confirms details and writes them into your CRM. If your process is inbound or unusual, that is a custom bot conversation rather than an off-the-shelf one.
Whichever you choose, judge it by the same measure your client uses. A bot that sounds impressive but produces fewer billable transfers than your agents is not a win, however much the client enjoys the demo.
How a voice bot sets up on your dialer
On a well-built setup, the bot fits into your dialer the way a new agent does. Nothing about your carrier or your numbers has to change.
Insist on the last step in particular. A bot that has not been tested on your own trunk, with your caller ID and your transfer queue, is still a demo, however good it sounded on the vendor's line. The usual steps look like this.
- The bot gets an agent login on ViciDial, a virtual agent slot on Convoso, or a route on your own SIP trunk
- You assign it to the campaign and list it should work, with your existing calling hours
- You share the script your agents read today, or two recordings of good calls if the script lives in people's heads
- Your disposition codes are mapped, so the bot writes the same outcomes your team does
- The transfer queue or extension is set, so qualified callers reach your closers
- You run test calls on your own numbers and approve the flow before it dials real leads
Mapping dispositions and getting the client's sign-off
Dispositions are where pilots get judged, so get them right on day one. If your US client pays on billable transfers and reads your dialer reports, the bot has to use the same codes. Our ViciDial setup notes and the page on disposition write-back cover how that mapping works.
Then there is your client. Many US clients write call handling into the contract, so check before a bot touches their leads. Expect questions on the disclosure wording, whether the bot says it is automated, how recordings are kept, and who approves script changes. Get the answers, and the approval, in writing. A pilot that surprises the client can cost you the process, however good the numbers look.
It also helps to agree with the client what counts as a good result before you start. If they judge you on billable transfers, report billable transfers. If they care about complaints or chargebacks, track those from day one. A pilot measured on the wrong number can look like a failure when it is not, or the other way round.
Language: English today, Spanish coming
For a US process the bot needs to speak American-sounding English, and that is what our bots do today. Spanish is coming, but it is not live yet. If a share of your list prefers Spanish, keep those leads with your bilingual agents for now and mention it when you sign up, so we can tell you when it is ready.
Watch for one practical trap. If your list mixes English and Spanish speakers, the bot needs a clear rule for a caller who answers in Spanish: end the call politely with its own disposition, or transfer to a bilingual agent if one is free. Either is fine. Leaving it undefined is not, because those calls will otherwise be logged as hang-ups and skew the pilot.
There is a side benefit on the voice itself. Accent coaching is one of the slowest parts of training new agents for a US process. A bot's voice is chosen, not trained, from a large library of natural voices, or cloned from one of your own agents with their consent.
A two-week voice bot pilot test plan
Run the pilot on a slice of one list, not the whole campaign. Our guide to piloting bots on a slice of your list covers sizing. This is the plan we suggest for the two weeks themselves. Agree the measures with your client before day one, and keep the human slice running the whole time so both are compared on the same leads, hours and closers.
- Days 1 and 2. Test calls only, on your own phones. Interrupt the bot, answer out of order, ask if it is a robot, ask to be removed. Fix the script.
- Days 3 to 5. Live on a small slice of fresh leads, alongside your agents on a matched slice. Listen to twenty calls a day, including the ones that did not transfer.
- Days 6 and 7. Check AMD: count live people hung up on and voicemails that reached a closer. Adjust.
- Week 2. Widen the slice. Track billable transfers per hour, closer complaints, DNC requests and disqualifications by question.
- End of week 2. Compare bot and human slices on billable transfers, not raw transfers. Share the results and recordings with your client before scaling.
The IST night shift with bots on it
The late hours, roughly 1 a.m. to 3:30 a.m. IST on an Eastern campaign, are where bots earn their place. Human transfer rates fall, supervisors are stretched, and transport for the last batch of agents is already booked. Bots hold their pace, so the question moves to whether you have enough closers awake to take the transfers. Plan closer staffing and callback rules together.
Supervision changes as well. Instead of walking the floor to catch shortened disclosures, a team lead can watch live calls and transcripts on a screen and step in on the few that go wrong. That is a different skill, and it is worth training one or two people in it before the pilot starts rather than during it.
Our page for Indian call centers has more on how we work with floors on IST hours. B3 Voice runs on US shift hours too, so support is awake when your floor is. If you want to try a bot on your own process, every account starts with a 7-day free trial.
