Case study · AI Voice · Venture
FluxDial: AI telecalling that respects emerging-market unit economics.
Founder & Lead Architect · fluxdial.com
An agent-first voice-calling CRM for phone-first businesses in India and Nigeria — built because wrapping US voice APIs does not survive local telecaller math.
Context
Tanish founded FluxDial after seeing “AI voice” products that were thin wrappers around US platforms like Vapi and Bland. In India, a human telecaller can cost on the order of ~$0.02/minute equivalent, while many US AI voice APIs bill roughly $0.07–0.15/minute — inverting the ROI story. Nigeria became an early go-to-market beachhead with USD billing and local routing; India pilot workflows run in parallel. Production work includes live client deployments (including Nigerian legal/ops workflows) and hard engineering on realtime telephony paths.
Problem
Call-heavy businesses lose revenue when phones go unanswered, yet imported AI voice stacks are priced for US labor markets. Teams needed inbound reception and outbound qualification that feel human, speak Hindi/English naturally, log every outcome, and still make financial sense.
Approach
- —Treat telephony, speech, reasoning, and tool-calling as one owned stack — not a collage of opaque vendors you cannot cost-control.
- —Split product surfaces into Flux Desk (inbound receptionist) and Flux Hunter (outbound qualification) so pilots can start narrow.
- —Design for bilingual conversations and regional numbers; optimize for sub-500ms response feel on live calls.
- —Harden money and concurrency paths (e.g. wallet/ledger updates with row-level locking) so race conditions cannot double-spend credits.
- —Make the dashboard the system of record: transcripts, lead status, bookings, transfers, and next actions.
- —Ship Nigeria with clear USD plans and instant activation; run founder-led onboarding for custom call flows.
What shipped
- —Public FluxDial product site and signup path for Nigeria (USD via xPayCheckout) with Starter / Pro / Premium minute plans
- —Inbound + outbound AI telecaller flows with booking, warm transfer, and dashboard memory
- —Industry pilots framing for healthcare, real estate, education admissions, hospitality, and other phone-first verticals
- —Integration hooks: Google Calendar, webhooks, REST API for CRM/helpdesk follow-up
- —Production concurrency fixes on wallet/idempotency paths for prepaid minute economics
- —Roadmap products (FluxMed, FluxStay, Flux Shadow) sharing the same voice R&D thesis
Outcomes
- —Nigeria publicly live with local routing and USD subscription — teams can activate without a long enterprise sales cycle
- —Cost thesis made concrete: AI minutes must compete with ~$0.02/min human telecaller economics, not US SaaS pricing
- —Live client deployments proving the stack under real call traffic — not a demo reel
- —Sub-500ms response target as a product constraint, not a marketing claim
- —Established Tanish as a practitioner-founder on AI voice infrastructure — not a reseller of US APIs
Stack & surfaces
Hire for similar work
FAQ
- What was Tanish’s role on FluxDial?
- Founder and lead architect — product thesis, voice stack direction, and the emerging-market cost architecture.
- Why not just use Vapi or Bland?
- Those tools can be excellent in US cost structures. In India, AI minutes that cost several times a human agent destroy the business case. FluxDial starts from local economics.
- Can I hire Tanish for a similar system?
- Yes — see AI voice agent engagements or email mailtanish.me@gmail.com / call +91 93406 08328.