$100K+
Additional revenue from AI-handled conversations
80%
Conversations handled fully autonomously
35%
Conversion rate from inquiry to order
About the company
Tokoparts
Tokoparts is Indonesia's first B2B automotive spare parts e-commerce platform, founded in 2020 as a joint venture between Mitsubishi Corporation and Putra Mandiri Jembar. Operating across four branches in Jakarta, Bandung, Surabaya, and Semarang, the platform carries 50,000+ SKUs spanning 50+ genuine and aftermarket brands including Mitsubishi, Toyota, Honda, Daihatsu, Suzuki, Hino, and Fuso. It connects parts suppliers directly with workshops, garages, and fleet operators: buyers who know something is wrong with their vehicle but rarely arrive with a part number. Getting the right part matched to the right application, accurately and fast, is the entire commercial proposition.
The Challenge
Spare parts are not like most products. Customers do not browse. They describe a problem and expect an answer.
01
Buyers describe symptoms, not part numbers.
Most inquiries start as symptom descriptions, not part numbers. Translating "my brakes have been grinding since last week" into a specific, compatible SKU requires product depth that no rep can hold across 50,000 items.
02
Catalog scale made accurate human handling impossible.
At the volume Tokoparts operates, a rep fielding dozens of inquiries daily cannot maintain accuracy across the full catalog. One wrong recommendation means a return, a wasted courier run, and a customer who does not come back.
03
Bulk orders arrived as images, spreadsheets, and handwritten lists.
Customers sending fleet or workshop orders did not use forms. They sent a photo of a handwritten list, a Google Sheet, or a WhatsApp forward. A sales rep had to manually parse 50+ line items one by one, cross-check each SKU, and compile a quote. A single bulk order could consume most of a rep's day.
50+ items per bulk order
Bulk orders arrived as photos, handwritten lists, or informal messages — no structured format, each requiring full manual cross-referencing.
30–60 min per order
Every complex parts request required a human to look up compatibility across 50,000+ SKUs and 47 brands before confirming anything.
3–10 hour response time
Orders that needed catalog lookups or bulk pricing waited hours. No after-hours coverage meant overnight queues for workshop clients.
Manual invoice generation
Payment links and invoices were produced manually after every order was confirmed — adding a step after an already manual process.
What Covena Built: A Technical Sales Agent That Knows the Full Catalog
Covena handled the build, training, and deployment end-to-end. Tokoparts did not write a single prompt.
Symptom-to-SKU Natural Language Matching
Buyers describe a problem and the AI returns the correct part number across 50,000+ SKUs in seconds. No form, no dropdown, no rep.
Vehicle Compatibility Verification Before Recommendation
The AI cross-checks vehicle model, year, and trim before surfacing any SKU. The wrong part never gets recommended, and returns drop to near zero.
Autonomous Invoice and Payment Link Generation
Once the part is confirmed, the AI generates the invoice and sends the payment link in the same thread. One conversation, one transaction, no rep required.
Bulk Multi-Vehicle Order Coordination
Garages ordering parts for multiple vehicles in a single thread get each item handled in sequence, with compatibility checks on all, and a combined invoice for the full order.
Specialist Escalation With Full Thread Context
Inquiries that need a parts specialist route with the complete conversation already attached. The rep steps in at the decision point, not square one.
The AI in Action
See how the agent navigates a 50,000-SKU catalog from first symptom description to paid order.
Upload image for "Symptom-to-SKU: Pajero Sport Brake Job"
How We Got to Live in 12 Weeks
Weeks 1–2
Discovery & Catalog Integration
50,000+ SKUs and 47 brands mapped, vehicle compatibility matrix scoped, and bulk order workflow documented with Tokoparts' catalog and sales teams.
Weeks 3–6
Agent Build & SKU Matching Engine
Symptom-to-SKU natural language matching built, vehicle compatibility verification integrated, and invoice generation wired to Tokoparts' order system.
Weeks 7–9
WhatsApp Integration & Payment System Connection
Connected to Tokoparts' WhatsApp Business account and integrated with the invoice and payment link generation pipeline.
Weeks 10–11
Testing & Accuracy Validation
Real parts queries tested at scale, bulk order scenarios validated, and compatibility accuracy confirmed with Tokoparts' catalog team.
Week 12
Go Live & Iteration
Full inbound traffic handed to the agent, with both teams monitoring edge cases in real time.
The Impact
From manual, rep-dependent handling to a scalable AI sales channel.
$100K+
Additional Revenue per Quarter
From AI-handled conversations with no sales rep in the thread.
80%
Conversations Autonomous
End-to-end from inquiry to invoice without human involvement.
35%
Conversion Rate
From inquiry to confirmed order.
0
Additional Headcount
50,000+ SKUs, same team size.
Metric
Before
After
Catalog coverage per conversation
Limited by human recall
50,000+ SKUs across 47 brandsFull catalog
Response time
10 min to 3 hours
Under 30 seconds↑ 99%+
Bulk order handling
Manual line-by-line (50+ items)
Fully automated
Wrong part recommendations
Common
Near zero (compatibility verified first)↓ Significantly
Specialist escalation
Rep starts from scratch
Full thread context passed on handoff
Wrong part recommendations
[To confirm]
Near zero↓ Significantly
What's Next
The AI grows with the catalog: every new SKU or brand added is immediately searchable and matchable. The sales team focuses on fleet accounts and relationships, while the agent handles the transaction volume underneath.
Operating a large SKU catalog that your team cannot navigate at speed?
See how Covena builds AI agents that turn complex inventory into a scalable sales channel: from first symptom description to paid order.