All client work

How Sushilanand stopped losing 1 in 5 buyer calls

A construction-equipment manufacturer losing buyers to voicemail, because the enquiries arrived while the owner was on the shop floor. We built a voice agent that answers every one of them, in whichever language the caller speaks.

Construction lifting equipment Founded 2022 ~$180k–600k revenue Sells through a B2B marketplace India

The client

Sushilanand Equipments makes lifting and hoisting equipment for the construction trade — mini cranes, monkey hoists, tower hoists, rope-suspended platforms, material lifts, and passenger-cum-material hoists that carry 1,000 kg up to a hundred metres. Their buyers are builders, contractors and site owners across Maharashtra and north Karnataka.

The firm was founded in 2022 and sells largely through a storefront on IndiaMART — India's dominant B2B marketplace, roughly what Thomasnet is to American manufacturing. Prices are listed publicly, from a ₹70,000 (about $850) mini crane to a ₹16.5 lakh (about $20,000) passenger hoist, which means a buyer usually calls already knowing roughly what they want to pay. What they need to establish on the phone is whether the machine fits their site.

The problem

Sushilanand's own IndiaMART storefront publishes a 78% call response rate. Read plainly: about one buyer call in five was going unanswered — and in this trade, an unanswered call is a buyer who dials the next supplier on the results page.

The cause wasn't neglect. It was that the owner is the front desk. Anup Patil was on the shop floor, at a site, or driving between the two when the phone rang. Callbacks slipped to the evening or the next morning. Quotes that went out and went quiet were never chased, because no system existed to notice they'd gone quiet. Everything — every buyer, every requirement, every promise — lived in one person's phone.

  • One in five inbound calls unanswered, against a public benchmark on their own listing.
  • Callbacks measured in hours or overnight, when the competing supplier answers on the first ring.
  • No follow-up on silent quotes — no step in the process, ad hoc at best.
  • No CRM. No record of what a buyer asked for last month.

What we built

A voice agent that answers the phone — not a chatbot, not a form. It handles the entire first conversation the way Anup would, and stops exactly where he'd want to take over.

  • Answers in the caller's own language — here Marathi, Hindi or English — detecting it and switching mid-call when the caller does. The language set is configuration, not architecture: the same build handles any language the voice layer supports.
  • Qualifies properly: which machine, what capacity, what lift height, where the site is, what timeline, what budget.
  • Quotes the listed catalog price and nothing else. If it isn't certain, it says Anup will confirm — it never invents a number.
  • Sends the spec sheet and price on WhatsApp during the call, while the buyer is still listening.
  • Books a callback in Anup's Google Calendar, or warm-transfers to him if he's free.
  • Writes every lead to the CRM sheet with a one-line summary and the full transcript.
  • Calls back new IndiaMART "Get Best Quote" enquiries within two minutes, automatically, before the buyer has finished comparing suppliers.

Just as important is what it refuses to do. No discounts. No custom specifications. No delivery-date promises. Anything above an agreed value goes to a human. The agent is built to be trusted with the first five minutes, not the negotiation.

IndiaMART Lead Manager webhook Indian virtual number + forwarding ElevenLabs voice (SIP) Claude — reasoning & tool use WhatsApp Business API Google Calendar Google Sheets CRM

How we did it

  1. Discovery

    We sat with Anup, mapped the call flow, listened to real buyer calls, and listed the twenty questions that actually come up. Then we agreed — in writing — what the agent is allowed to do and what it must never do.

  2. Knowledge & data setup

    We structured the product catalog and had Anup correct it line by line, then wired the IndiaMART webhook, the virtual number, WhatsApp Business API, the calendar and the CRM sheet.

  3. Build

    Multilingual script, qualification branches for each machine class, tool wiring, the hard rules, and full logging of every call.

  4. Test

    Over fifty internal role-play calls — the Marathi-only contractor, the rambler, the discount-hunter, the caller who wants a product they don't make, the one on a bad line.

  5. Shadow mode, then launch

    The agent took missed and after-hours calls first, with Anup reading transcripts daily — pure upside, nothing at risk. Then speed-to-lead on IndiaMART, then all inbound. A WhatsApp digest lands every evening.

Results

MetricBeforeNow
Inbound calls answered~78% (IndiaMART public stat)Every call picked up
First callback on an IndiaMART leadHours, often next morningUnder 2 minutes
Languages handledWhoever was free to talkMarathi, Hindi, English
Owner's time on first-contact callsAll of it, mid shop-floor workOnly qualified callbacks
Follow-up on silent quotesAd hoc or noneEvery quote chased
Record of a buyer's requirementOne person's phoneCRM row + full transcript

Half our enquiries come in while I'm on the shop floor. Now an agent answers every call in Marathi, Hindi or English, quotes from our catalog, and books the callback.

Anup Patil CEO, Sushilanand Equipments

Missing calls while you're on the floor?

Book a free 30-minute audit. We'll map one workflow live and tell you honestly whether it's worth automating — even if the answer is no.

Book a free audit