A solar lead arrives with one useful fact in it: the electricity bill. Everything after that is your team retyping the same twelve answers about subsidy slabs, shadow-free roof area and where the net meter has got to. Voxdonna builds an AI employee that owns the whole chain. It sizes the system from the bill, books the survey, files and tracks the subsidy and net-meter stages, and tells the customer where things stand before they ask you twice.
Every one of them is answerable from data you already hold. None of them is answerable at 11:30 PM by a person who has gone home.
"My bill is about this much a month, is solar worth it?" is the entire brief you get. Sizing it, explaining the subsidy slab and getting to a survey slot costs a trained person most of half an hour, and that message arrives all day, every day, from people who have not decided anything yet.
Who qualifies, how much for 1 kW versus 3 kW, whose name has to be on the connection, whether a rented roof counts, when the money actually lands. Your sales team answers these instead of selling, and one wrong answer puts an application in the rejection pile.
Panels went up three weeks ago and the customer is calling you about a net meter you do not control. Name mismatches across Aadhaar, bill and bank account, pending load enhancement and bidirectional meter stock are the usual causes, and none of them are visible to the customer unless somebody tells them.
Import gets billed without the export being adjusted, the bill lands far above near-zero, and the customer concludes the system does not work. That call comes to you, not to the DISCOM, and you have no evidence at hand unless the generation data is already pulled.
Most red lights are grid-side faults the inverter cleared itself. Without remote diagnostics in the first message, you dispatch a technician for a voltage excursion that resolved before he reached the gate.
Generation drifts down because the panels have not been cleaned in five weeks, the customer never sees a generation report, and the referral you were owed goes nowhere. The best salesperson you have is a customer holding a near-zero bill, and you never ask them to show it.
Sales, subsidy and service are one continuous relationship in rooftop solar. The agent treats them that way.
The calling side runs on the same stack described in how the voice layer works, and the messaging side on WhatsApp agents for enquiry and service. For the commercial deals, SalesDonna, for what happens in the C&I meeting captures the room, and your closers can get real-time coaching on live sales calls. On the buying side there is procurement intelligence for module and BOS sourcing and TenderCraft, for government and C&I solar tenders.
Each owns a stage of the same customer relationship. Most EPCs begin at lead-to-survey, because the volume is obvious and the arithmetic never changes.
Reads the electricity bill, sizes the system, applies the right PM Surya Ghar slab, and books the site survey without a human doing the arithmetic.
Pre-checks the documents that actually cause rejections, files the National Portal and DISCOM applications, and reports every stage change before the customer asks.
Owns the gap between installation and commissioning, escalates a slipped DISCOM queue, and keeps the customer informed through the wait you do not control.
Checks the first post-solar bills against generation data, catches the unadjusted-export billing error, and files the correction with evidence attached.
Triages inverter faults from a voice note, separates grid events from hardware failures, books only the technician visits that are actually needed, and sends the monthly generation report.
Directional, because we publish what we can point at. The magnitude is yours to measure on your own pipeline and your own dispatch log.
Nothing new to buy. The agent sits on the systems that already hold the answer and stops your team retyping it.
The application, the reference number and the stage are the spine of the whole post-sale experience.
Per state, per circle. The four-stage tracker lives or dies on reading status back from here.
Where bill photos, voice notes, stage updates and generation reports actually land in India.
Growatt, SolarEdge, Enphase, SolarMan and the OEM portals. Fault codes and generation data are what make the service agent useful instead of apologetic.
Lead, sized system, subsidy eligibility, survey slot, case status.
The majority of residential solar inbound, and the reason the 11:30 PM response matters.
Surveyor and technician calendars, so a booked slot is a real slot.
Reading the electricity bill turns a photo into units, sanctioned load, connection name and tariff category in one step.
Quotations, milestone payments and EMI paperwork, so the quote the agent explains matches the one finance issued.
Escalation alerts for slipped DISCOM queues, and C&I leads routed to the commercial desk.
These are demo builds, not deployments. The figures inside them are illustrative, and each one takes about a minute.
Sizing, subsidy eligibility, survey booking, net-meter stages, first-bill audit and the monthly generation report, in one thread.
Tara, the rooftop solar WhatsApp concierge →
An outbound line that explains the scheme, answers eligibility questions and logs interest for follow-up.
Suraj, the Hindi PM Surya Ghar outreach voice agent →
The pattern behind the O&M desk: symptom captured, severity judged, engineer routed with the context written down.
emergency field-service dispatch →
Monitoring data turned into a scheduled visit instead of a breakdown, which is what generation-drift alerts become.
predictive maintenance dispatch →
What the utility's own connection desk sounds like when it is automated. Useful context for the queue you are waiting in.
the DISCOM new-connection agent →
Voice and WhatsApp agents across industries, all playable without a signup.
all 40+ live Voxdonna demos →
Rooftop solar is one of every industry Voxdonna builds for. The enquiry-to-appointment chain is the same shape as AI agents for real estate developers, the appointment and walk-in side rhymes with AI agents for jewellery retail, and if you make the modules or the inverters rather than install them, the page you want is AI agents for manufacturers.
Yes, and it is the fastest qualification signal you have. In our Tata Power Solar WhatsApp demo the agent takes the monthly bill amount, converts it to monthly units at the local tariff, recommends a kW band, states the net cost after the applicable PM Surya Ghar slab and gives a payback range. It then books the survey, where the engineer confirms shadow-free roof area and issues the real quote. Bands from the agent, final numbers from your engineer.
It works from a locked knowledge base with the current slabs, and it pre-checks the three things that cause rejections: the connection has to be in the applicant's name, the roof has to be theirs, and the connection has to be residential. Name mismatches across the bill, Aadhaar and bank account are the most common failure, so it catches those before filing, not after.
That is the point of it. The demo tracks four stages, application, feasibility, meter inspection and physical swap, and pushes an update the moment one moves. When the DISCOM queue slips, the agent says so and records the escalation. Customers stop calling your office about a queue neither of you controls, because the status arrives before the question does.
Mostly. A red light on the inverter is usually a grid-side voltage excursion the inverter self-protected against, and it resumes when the grid stabilises. The agent pulls live system data, returns the fault code and what it means, and books a visit only when the data says a visit is needed. Our demo runs that triage from a Hindi voice note.
All three, with different paths. Residential runs the subsidy path. Housing societies run common-area systems on group net metering, which needs an RWA resolution and six months of bills. C&I gets routed to a commercial desk with a CAPEX versus OPEX comparison on a real 12-month load profile rather than brochure numbers, because a CFO is not buying a subsidy, they are buying a tariff.
It runs in 22 Indian languages in one agent, and it switches when the customer does. Voice notes in Hindi come in and answers go back out as voice notes, which matters because the customer who sends one is usually standing in front of the inverter. Technical terms customers use in English, kW, net meter, inverter, subsidy, stay in English.
Fourteen days to a working line, using the channels you already run. No new CRM, no portal rebuild. The agent reads from your existing system catalogue, subsidy rules and monitoring platform, and writes back to the CRM you already pay for. Start with one workflow, usually lead-to-survey, and add the subsidy tracker once it is earning.
Bring one week of inbound enquiries and one month of service tickets. We will tell you which chain pays for itself first, and what it takes to stand it up.