Vriksha Voice

What it is

Vriksha Voice is a voice AI calling platform for teams that want outbound calls to behave like a paid media channel. It is not a general builder where you drag blocks onto a canvas. You define an audience, a business goal and a target return, and the platform runs voice agents against them, tests script variants and reports results as ROAS. The site is operated by Vriksha Technologies, and a comparison page describes the whitelabel product as built for agencies, founders and resellers in India and Southeast Asia. It belongs in the voice agent builders category because you assemble multilingual agents in its Creative Agent Studio, and in chatbot builders for the same reason: the agent is a scripted conversation you author, test and push into WhatsApp and a CRM.

What you build and how

The home page calls it "Five engines. One growth loop." Each engine is a screen in the product.

The Audience Engine is where you "segment and target like a media planner, not a dialler". The page describes reusable audience queries with business filters and structured attributes. A sample panel shows an audience match of 128,412 contacts, 24 live segments, a 91% match rate and a cost per lead of Rs. 184. These are figures shown in the product illustration, not a customer result.

The Creative Agent Studio is the builder. Its tagline is "Multilingual voice agents, A/B tested like ad creatives." The panel shows language chips for EN, HI, MR and TA and 3 variants under test. You author agent scripts and run several against each other. The experiment description reads: a hypothesis "written by AI from your transcripts", an experiment using Thompson sampling "across script, voice, timing, number, audience", and a decision that will "auto-promote at 95%, auto-kill at 5%". A sample test shows Script A at 52%, Script B at 31%, Script C at 12% and Script D at 5%, with a 96% probability that Script A is best.

The Campaign and Bid Manager launches, paces and retries calls from one command center. The panel shows pacing on track, a 62% connect rate and automatic retry. The FAQ says lists are scrubbed against TRAI DND, consent is recorded, and retry windows and calling hours are enforced per campaign.

Attribution and ROAS turns every call into "a conversion event you can defend". The sample panel shows a live ROAS of 4.6x and cost per lead down 32%. The omnichannel section ties voice to WhatsApp Business API, email follow-up through Gmail and Outlook, Meta and Google Ads campaign signals, Google Sheets lead sync, webhooks and API, and a CRM of your choice. A sample blended ROAS of 4.2x and a 38% connect-rate lift are shown there. Retargeting and CRM, the fifth engine, pushes qualified intent into WhatsApp, ads and the CRM.

The product also builds in a test-before-you-spend flow. You set a minimum test budget, the platform runs a statistically valid test and issues a verdict: "Green scales. Red saves your budget." The FAQ says a test usually reaches a verdict within 48 hours of launch, inside a minimum test budget of a few hundred calls. There is no setup fee, and you can cancel any campaign at the end of a test.

For agencies, a whitelabel layer lists custom branding, custom domains, client workspaces, configurable plans and white-label reporting. The comparison page says you can launch your own branded Voice AI calling platform in 48 hours.

What the pages say about pricing and proof

Prices are in rupees, with a currency toggle for INR and USD on the comparison page. The home page lists 3 models. Per minute is Rs. 5 per minute, "All upside stays with you." A tier labeled Plus optimizer is written "Rs. 5 + Rs. 2 per minute + per call" and covers experiments, verdicts and auto-promotion. The third, cost per lead, is "from Rs. 200 per delivered lead" for D2C and Rs. 180 for Healthcare B2B, with a contract cap you approve upfront and nothing owed for leads not delivered. Cost per lead pricing opens only after a green test verdict.

A separate pay-per-pulse band bills voice calling at Rs. 0.65 per 10-second pulse, rounded up, which works out to Rs. 3.90 per minute. It applies to 2,000 to 10,000 minutes a month with a minimum spend of Rs. 7,800 a month. A 16-second call bills 2 pulses, or 20 seconds. Telephony is billed at cost as a separate line, and prices exclude GST.

For lenders and insurers, a BFSI model sells consented leads. Packages come in 25, 100, 250 and 500 leads a month. The sample shows Rs. 999 per consented lead for personal loans at 100 leads a month. Non-qualifying calls bill Rs. 0, every lead ships with the recording, a transcript excerpt and captured variables, and the lead definition is agreed in writing. The page adds that lenient leads ship free as a nurture list. GST is invoiced in India.

The whitelabel plan on the comparison page starts at Rs. 4,999 a month plus Rs. 4.50 a minute, and the page says wholesale rates are bundled at Rs. 4.5 per minute.

On proof, the customer story section of the home page is still a placeholder: it says verified customer name, quote and metric are "to be added". So the figures above are product illustrations and the company's own claims, and the pages carry no customer count, security certification or open-source claim.

Who it suits and who it does not

Vriksha Voice suits a performance marketing team, a lending or insurance business in India, or an agency that wants to resell outbound voice AI under its own brand. The pay-per-lead and consented-lead models move dialing risk to the vendor, and the DND scrubbing and consent recording address Indian calling rules on the page's own terms.

It is a poor match for someone who wants a general no-code app or chatbot builder, a visual flow canvas or a support-line bot. The studio is about scripts and experiments, and the pages describe outbound and inbound voice as part of one channel without detail on building branching flows. Pricing is in rupees and per-pulse bands begin at 2,000 minutes a month, so very small pilots will want the per-minute or cost-per-lead route. And with no published customer stories yet, you will be taking the ROAS figures on trust until a test on your own list produces a verdict.